*ATP2A2* **SINE Insertion in an Irish Terrier with Darier Disease and Associated Infundibular Cyst Formation**

**Monika Linek 1,**†**, Maren Doelle 1,**†**, Tosso Leeb 2,3,\*,**†**, Anina Bauer 2,3, Fabienne Leuthard 2,3, Jan Henkel 2,3, Danika Bannasch 2,4, Vidhya Jagannathan 2,3 and Monika M. Welle 3,5**


Received: 14 April 2020; Accepted: 27 April 2020; Published: 28 April 2020

**Abstract:** A 4-month-old female Irish Terrier presented with a well demarcated ulcerative and crusting lesion in the right ear canal. Histological analysis revealed epidermal hyperplasia with severe acantholysis affecting all suprabasal layers of the epidermis, which prompted a presumptive diagnosis of canine Darier disease. The lesion was successfully treated by repeated laser ablation of the affected epidermis. Over the course of three years, the dog additionally developed three dermal nodules of up to 4 cm in diameter that were excised and healed without complications. Histology of the excised tissue revealed multiple infundibular cysts extending from the upper dermis to the subcutis. The cysts were lined by squamous epithelium, which presented with abundant acantholysis of suprabasal keratinocytes. Infundibular cysts represent a novel finding not previously reported in Darier patients. Whole genome sequencing of the affected dog was performed, and the functional candidate genes for Darier disease (*ATP2A2*) and Hailey-Hailey disease (*ATP2C1*) were investigated. The analysis revealed a heterozygous SINE insertion into the *ATP2A2* gene, at the end of intron 14, close to the boundary of exon 15. Analysis of the *ATP2A2* mRNA from skin of the affected dog demonstrated a splicing defect and marked allelic imbalance, suggesting nonsense-mediated decay of the resulting aberrant transcripts. As Darier disease in humans is caused by haploinsufficiency of *ATP2A2*, our genetic findings are in agreement with the clinical and histopathological data and support the diagnosis of canine Darier disease.

**Keywords:** *Canis lupus familiaris*; dog; dermatology; skin; desmosome; acantholysis; calcium; animal model; veterinary medicine; precision medicine

#### **1. Introduction**

The skin is the largest organ of the human body and forms an essential barrier to protect the body from fluid loss and harmful agents of the environment. The epidermis representing the outermost layer of the skin consists of a stratified epithelium with keratinocytes as its major cell type. Keratinocytes proliferate in the basal layer and subsequently undergo a highly coordinated differentiation program while they move upwards through the spinous and granular layers until they finally reach the stratum

corneum, from which they are continuously shed [1]. The barrier function of the epidermis requires tight adhesion between keratinocytes, which is mainly mediated by desmosomes. Ca2<sup>+</sup> signaling is essential for epidermal differentiation and intraepidermal cohesion [2–5]. Several inherited disorders of the skin involving variants in calcium pumps have been recognized [6].

In humans, Darier disease (MIM #124200), also called Darier-White disease or keratosis follicularis, is inherited as an autosomal dominant trait and caused by heterozygous variants in the *ATP2A2* gene encoding the endoplasmic/sarcoplasmic reticulum Ca2<sup>+</sup>-ATPase 2 (SERCA2) [7,8]. Darier disease typically starts before the third decade and is clinically characterized by warty papules and plaques in seborrheic areas (central trunk, flexures, scalp, and forehead), palmoplantar pits, and distinctive nail abnormalities [7,9]. Secondary infection is common. Neuropsychiatric abnormalities have been described in a small fraction of the patients with Darier disease [9].

Hailey-Hailey disease (OMIM #169600), also called benign chronic pemphigus, is another autosomal dominant skin disorder caused by heterozygous variants in the *ATP2C1* gene encoding a Ca2+-ATPase expressed in the membrane of the Golgi apparatus [10]. Hailey-Hailey disease usually becomes manifest in the third or fourth decade of life with erythema, vesicles, and painful erosions involving the body folds, particularly the groin and axillary regions [11]. Both diseases are characterized histologically by the breakdown of intercellular contacts between suprabasal keratinocytes (acantholysis) with variable dyskeratosis. Differential diagnosis is based on the skin lesion types, their distribution on the body, and subtle histological differences [9,11,12].

Many independent genetic variants in *ATP2A2* and *ATP2C1* in human patients with Darier disease or Hailey-Hailey disease have been described. Variations in the clinical and histological phenotypes may at least partly correlate with the different specific genetic variants [8]. Nonetheless, both diseases are inherited as autosomal dominant traits and are due to haploinsufficiency of the encoded calcium pumps [6].

Dermatoses affecting desmosomes in domestic animals have been summarized in a comprehensive review [13]. In one report, clinical, histological, immunohistological, and ultrastructural findings in a male English Setter and two of its female offspring were initially reported as Hailey-Hailey disease [14,15]. A subsequent study found depletion of the ATP2A2-gated stores in cultured keratinocytes from one of these dogs and suggested that these dogs had Darier disease and not Hailey-Hailey disease as previously reported [16]. To the best of our knowledge, the underlying causative genetic variant was not reported in these cases and no further cases in dogs have been reported in the scientific literature.

In the present study, we describe the clinical and histological phenotype and the genetic analysis of an Irish Terrier, which all together enabled the diagnosis of canine Darier disease. In addition to the epidermal lesions, this dog presented with multifocal infundibular cysts with suprabasal acantholysis, a feature that has never been described with Darier disease, neither in humans nor in dogs. The successful management of the skin lesions with repeated diode laser ablation is outlined.

#### **2. Materials and Methods**

#### *2.1. Ethics Statement*

All dogs in this study were privately owned and samples were collected with the consent of their owners. The collection of blood samples from control dogs was approved by the "Cantonal Committee For Animal Experiments" (Canton of Bern; permit 75/16; Approval date: 11 July 2016). The collection of samples from the affected dog was performed for diagnostic or therapeutic reasons and did not constitute an animal experiment in the legal sense.

#### *2.2. Clinical Examinations and Management*

A 4-month-old, intact female Irish Terrier with 9.5 kg body weight was initially presented with skin lesions in the outer ear canal in August 2016. The dog was clinically monitored for general growth, general health and skin lesion development over a period of three years. Cytology swabs were taken from crusting lesions and fine needle aspirates from nodules, respectively. Skin lesions in the outer ear canal were visualized and punch biopsies were taken via video otoscopy (Tele Pack Vet X Led, Carl Zeiss, Germany, Tuttlingen) under general anesthesia with endotracheal intubation. Nodules were excised in toto. Tissue samples for histological evaluation were fixed in 10% buffered formalin immediately.

Blood was taken twice for complete blood count and genetic testing. Tear production was assessed by Schirmer's tear test, as recommended during vitamin A therapy. For laser ablation, an MLT Type 109 classic diode laser (Medizinische Laser Technologie GmbH, Ingelheim, Germany) with 4.0 W wave mode in continuous contact mode was used.

#### *2.3. Histopathology*

Biopsies were evaluated from the two plaque-like, partially eroded lesions of the external ear canal, one nodular lesion from the hind leg, and two large nodules from the neck. Tissue was processed routinely and stained with hematoxylin and eosin.

#### *2.4. Whole Genome Sequencing*

Genomic DNA was isolated from EDTA blood of the affected dog with the Maxwell RSC Whole Blood Kit using a Maxwell RSC 48 instrument (Promega, Madison, WI, USA). An Illumina TruSeq PCR-free DNA library (Illumina, San Diego, CA, USA) with ~350 bp insert size of the affected dog (IT390) was prepared. We collected 269 million 2 × 150 bp paired-end reads on a HiSeq 3000 instrument (32 × coverage). Mapping and alignment were performed as described [17]. The sequence data were deposited under the study accession PRJEB16012 and the sample accession SAMEA104283467 at the European Nucleotide Archive.

#### *2.5. Variant Calling*

Variant calling was performed using the Genome Analysis Toolkit (GATK) HaplotypeCaller [18] in gVCF mode as described [17]. To predict the functional effects of the called variants, SnpEff [19] software, together with NCBI annotation release 105 for the CanFam 3.1 genome reference assembly, was used. For variant filtering we used 655 control genomes, which were either publicly available [20,21] or produced during other projects of our group [17] (Table S1). Structural variants were identified by visual inspection of the Illumina short read alignments in the Integrated Genome Viewer (IGV) [22]. The genotypes at the *ATP2A2* SINE insertion (Chr26:8,200,944\_8,200,945ins205) were also derived by visual inspection of the short read alignments in IGV. Samples were genotyped as homozygous ref/ref, if they did not show any signs of a structural variant at this position and had at least 4 reads aligning from Chr26:8,200,929-8,200,945, thus spanning the 15 nucleotide duplication at the insertion site.

#### *2.6. Gene Analysis*

We used the CanFam 3.1 dog reference genome assembly and NCBI annotation release 105. Numbering within the canine *ATP2A2* gene corresponds to the NCBI RefSeq accession numbers NM\_001003214.1 (mRNA) and NP\_001003214.1 (protein).

#### *2.7. RT-PCR and Sanger Sequencing*

Total RNA was extracted from skin tissues using the RNeasy mini kit (Qiagen, Hilden, Germany). The tissue was first finely crushed in TRIZOL (Thermo Fisher Scientific, Waltham, MA, USA) using mechanical means, chloroform was then added and the RNA was separated by centrifugation. The RNA was cleared of genomic DNA contamination using the Quantitect Reverse Transcription Kit (Qiagen). The same kit was used to synthetize cDNA, as described by the manufacturer. RT-PCR was carried out using primer ATP2A2\_Ex14\_F, TCCTCCAAGGATTGAAGTGG, located in exon 14 and primer ATP2A2\_Ex16\_R, TGTCACCAGATTGACCCAGA, located in exon 16 of the *ATP2A2* gene.

After treatment with exonuclease I and alkaline phosphatase, cDNA amplicons were sequenced on an ABI 3730 DNA Analyzer (Thermo Fisher Scientific) using the forward primer ATP2A2\_Ex14\_F as sequencing primer. Sanger sequences were analyzed using the Sequencher 5.1 software (GeneCodes, Ann Arbor, MI, USA).

PCR on genomic DNA was performed using AmpliTaqGold360Mastermix (Thermo Fisher Scientific) and primers ATP2A2\_Ex14\_F (same as above) and ATP2A2\_Ex15\_R, TCAGGGCAGGAG CATCATTC. Genomic PCR products were also sequenced using the forward primer ATP2A2\_Ex14\_F as sequencing primer.

#### *2.8. Whole Transcriptome Analysis (RNA-seq)*

RNA libraries were prepared from total RNA of lesional and non-lesional skin of the affected Irish Terrier using the Illumina TruSeq Stranded mRNA Library Kit according to the manufacturer's instructions. The libraries were sequenced with 2 × 50 bp paired-end sequencing chemistry on an Illumina NovaSeq 6000 instrument. The reads were mapped with STAR aligner version 2.6.0 [23] to the CanFam3.1 reference genome assembly. The sequence data were deposited under the study accession PRJEB33508 and sample accessions SAMEA6800286 and SAMEA6800287 at the European Nucleotide Archive. The read alignments of the affected Irish Terrier were visually compared to a skin RNA-seq dataset from a healthy control dog (ENA project accession PRJEB33508, sample accession SAMEA6800283).

#### **3. Results**

#### *3.1. Clinical Examination and Management*

During the first consult in August 2016, a 4-month old intact female Irish Terrier presented with several confluent, well demarcated, proliferative, crusted, and partially eroded to ulcerated plaques at the concave pinnae of the right ear extending into the medial aspect of the vertical ear canal (Figure 1A). These lesions had been present since at least 4 weeks prior to the examination. Culture swabs taken by the referring veterinarian revealed *Staphylococcus pseudintermedius* sensible to most antibiotics. The presence of a foreign body had been excluded by an ear flush. At the time of presentation, the dog received amoxicillin/clavulanic acid at a dosage of 26 mg/kg body weight (BW) twice daily (Synulox®, Zoetis, Berlin, Germany) and prednisolone 5 mg/kg BW once daily (Prednisolon 5 mg®, CP Pharma mbH, Burgdorf, Germany). Several commercially available ear cleansers and eardrops had been applied before without any improvement.

**Figure 1.** Clinical phenotype. (**A**) Concave pinnae of the right ear showing well demarcated crusting, eroded and ulcerated skin plaques. (**B**) Medial aspect of the right ear canal with well demarcated ulcerated lesions visualized via video otoscopy after crusts had been flushed away. (**C**) Same aspect of the ear canal: Intact, slightly erythematous skin after repeated laser ablation.

The general and dermatological examination did not reveal any abnormalities except the moderately painful and mildly pruritic lesions of the right pinna and ear canal. Cytology showed clusters of acantholytic keratinocytes, non-degenerated neutrophils, and numerous cocci.

Video otoscopy showed intact eardrums and normal horizontal ear canals in both ears. The medial aspect of the right vertical ear canal revealed well demarcated, ulcerated lesions covered with thick crusts confluent with the lesions of the concave pinna (Figure 1B). Waiting for the biopsy results, the dog was treated with squalene ear cleanser every other day, twice daily topically with Triamcinolone Acetonide cream (Volon®A Haftsalbe 1 mg/g, Dermapharm AG) and sulfadiazine creme (Flammizine®, Alliance Pharmaceuticals Limited, Chippenham, UK), changed to customized eardrops of 1% fluoroquinolone in saline solution (Baytril®5%, Bayer AnimalHealth GmbH, Leverkusen, Germany) for easier handling. After the preliminary diagnosis of canine Darier disease, treatment with vitamin A 10,000 IU (Vitamin-A-saar®, Cephasaar, Ingbert, Germany) orally for two weeks daily; then, every other day was initiated and maintained for 3 months.

As no improvement was noticed after 3 months, we decided to ablate the lesional epidermis with a diode laser to remove the defect skin and provoke secondary healing from the periphery. This procedure was partially successful the first time. All lesions healed without crusts after repeated laser ablation of the affected tissue another three times, two, five, and 12 months apart (Figure 1C). On the pinna, small nodules of 1–2 mm remained. They were clinically and cytologically diagnosed as comedones.

In the following three years after the first presentation, the dog developed three well-demarcated, dermal nodules ranging from 2.5 cm to 4 cm in diameter on the dorsal neck, the left side of the neck, and on the right knee. Fine needle aspirates of all nodules revealed clusters of nucleated round to oval keratinocytes with mild anisocytosis and anisocaryosis and two to three nucleoli in the nucleus. These nodules were fully excised and submitted for histopathology. At the time of writing, no further lesions or nodules had developed.

#### *3.2. Histopathology*

Biopsies from the external ear canal revealed focally extensive epidermal hyperplasia with severe acantholysis affecting all suprabasal layers of the epidermis and resulting in in the formation of multiple small clefts and lacunae (Figure 2). Acantholytic keratinocytes were frequently dyskeratotic forming "corps ronds" (e.g., round bodies characterized by small pyknotic nuclei, a perinuclear clear halo and eosinophilic cytoplasm) or "grains" (cells with elongated nuclei present mainly in the stratum corneum and the granular layer). The epidermis was covered by compact orthokeratotic or parakeratotic keratin intermingled with dyskeratotic acantholytic cells. In areas of abundant acantholyisis, keratin extended as prominent focal plugs into the epidermis.

**Figure 2.** Histopathology. (**A**) Infundibular cyst underneath a focal area of hyperplastic epidermis with abundant suprabasal acantholyisis (rectangle). (**B**) Higher magnification of the focal area of hyperplastic

epidermis with suprabasal acantholysis overlying the cyst wall (arrow). The epidermal plaque is characterized by severely irregularly hyperplastic epidermis with abundant suprabasal acantholytic and dyskeratotic keratinocytes forming the "corps ronds" typical for Darier disease. Keratotic plugs composed of parakeratotic keratin and grains extend into the clefts resulting from abundant acantholysis. The hyperplastic plaque is overlying an infundibular cyst composed of squamous epithelium with abundant suprabasal acantholysis. The cyst is filled with parakeratotic keratin and numerous "corps ronds". (**C**) Higher magnification of the lesions already presented in (**A**,**B**). Note the abundant suprabasal acantholyis of dyskeratotic keratinocytes forming "corps ronds", "grains" and parakeratotic keratin (arrow). (**D**) Hyperplastic plaque from the outer ear canal. Within the severely hyperplastic epidermis, numerous acantholytic and dyskeratotic keratinocytes forming "corps ronds" (arrow) and causing small clefts are present.

In all biopsies from haired skin, one or multiple infundibular cysts measuring between 0.8 × 0.5 × 0.5 cm up to 3.5 × 3.0 × 1.2 cm were extending from the upper dermis to the subcutis. The cysts were lined by squamous epithelium, which presented with abundant acantholysis of suprabasal keratinocytes. The cysts were filled with parakeratotic keratin and numerous acantholytic and dyskeratotic cells. In one biopsy from the neck, the epidermis overlying the cyst presented with severe hyperplasia and suprabasal acantholysis comparable to the findings described for the outer ear canal. Similar findings were also present in the infundibular epithelium of some hair follicles.

#### *3.3. Identification of a Candidate Causative Variant*

We sequenced the genome of the affected dog at 32 × coverage and called single nucleotide and small indel variants with respect to the reference genome. The variants were compared to whole genome sequence data of 8 wolves and 647 control dogs from genetically diverse breeds and searched for private protein-changing variants in the two functional candidate genes *ATP2A2* and *ATP2C1*. This analysis of small variants did not identify any likely candidate causative variants for the phenotype (Table S2).

We then visually inspected the short read alignments in *ATP2A2* and *ATP2C1* to search for structural variants that would have been missed by our automated variant detection pipeline. Several truncated read alignments at the end of intron 14 of the *ATP2A2* gene indicated a potential insertion event including the duplication of 15 nucleotides flanking the insertion site. The inserted sequence represented a tRNA derived SINE (Figure 3A,B).

**Figure 3.** SINE insertion into intron 14 of the *ATP2A2* gene. (**A**) IGV screenshot illustrating the structural variant. The case shows an increased coverage over 15 nucleotides spanning from Chr26:8,200,930-8,200,944 (CanFam3.1 assembly). The sequence at the intron/exon boundary is given with the duplicated bases underlined. Capital letters represent the first 3 bases of exon 15. Several read alignments are soft-clipped at the left or right boundary of the duplicated 15 nt region. Colored reads indicate that their mates map to other chromosomes. These features are characteristic for an insertion of a repetitive element into the genome of the affected dog. (**B**) Schematic representation of the SINE insertion. A ~205 bp canine SINE-tRNA insertion was found in heterozygous state in the affected Irish Terrier. (**C**) Experimental genotyping of the SINE insertion by fragment size analysis. We amplified the intron 14/exon 15 boundary of the *ATP2A2* gene by PCR in the affected dog and a control and separated the products by capillary gel electrophoresis.

The genotypes at the SINE insertion site were investigated in the 655 control genomes. A total of 592 genomes had at least four reads spanning the insertion site and were genotyped as homozygous wildtype. In the remaining 63 control genomes, we did not see any indication for an insertion event. However, due to low coverage and/or short read lengths, the genotypes in these samples could not be reliably determined (Table S1). PCR amplification with flanking primers on a genomic DNA sample from the affected dog provided independent confirmation of the presence of a ~205 bp insertion in heterozygous state (Figure 3C).

#### *3.4. Analysis of the ATP2A2 mRNA*

We next investigated whether the SINE insertion into intron 14 had any effect on the expressed *ATP2A2* mRNA. Initial RT-PCR experiments on RNA from skin of the affected dog with different primer combinations yielded products of the expected size and sequence and did not indicate any obvious qualitative defects in mRNA splicing.

As the genomicinsertionwas only presentin a heterozygous state, thewildtype allelewas still expected to give rise to the normal *ATP2A2* transcript. Consequently, a potential splicing defect leading to nonsense mediated decay in transcripts from the mutant allele or transcriptional silencing of the mutant allele would not have been detected by our qualitative analysis of RT-PCR bands. We therefore additionally investigated the allele-specific expression of *ATP2A2* transcripts. This analysis demonstrated that ~85% of the detected transcripts were derived from the wildtype allele with an almost complete absence of transcripts from the mutant allele (Figure 4A,B).

**Figure 4.** Splicing defect and allele-specific expression of the *ATP2A2* mRNA. (**A**) Genomic organization of the *ATP2A2* gene. The affected dog was heterozygous at the silent c.2091A>G variant located at the end of exon 14 and heterozygous for the SINE insertion in intron 14. (**B**) A Sanger electropherogram obtained from a genomic PCR product shows the expected equal ratio of the two alleles at c.2091A>G. In contrast, a Sanger electropherogram obtained with the same sequencing primer from a cDNA amplicon showed a strong bias towards the A-allele (arrow). This semi-quantitative analysis suggests that the transcripts from the mutant *ATP2A2* allele are degraded, possibly by nonsense mediated decay or another mechanism of the cellular quality control. (**C**) RNA-seq analysis from skin of the affected dog confirmed the strong allelic bias in the transcripts. Only very little functional transcripts are produced from the G-allele. The majority of the transcripts from the G-allele contain an aberrant exon and a premature stop codon. Further details of the RNA-seq analysis are shown in Figure S1.

To gain further insights into possible splicing defects, we performed an RNA-seq experiment and whole transcriptome analysis in skin of the affected dog. Visual inspection of the short-read alignments in the region of the *ATP2A2* gene confirmed the strong allelic bias of the transcripts. Furthermore, this experiment revealed the presence of rare transcripts containing an additional, aberrantly spliced exon. This 139 nt exon was derived from genomic sequence a short distance upstream of the SINE insertion (Chr26:8,200,774-8,200,912). The aberrant exon contained an early premature stop codon. The variant designation of the predicted protein from transcripts containing the aberrant exon is NP\_001003214.1:p.(Thr700Valfs\*6). Only a small proportion of the transcripts from the mutant allele was correctly spliced and had the correct coding sequence (Figure 4C and Figure S1).

#### **4. Discussion**

In this study, we describe the clinical and histological phenotype and the genetic analysis of an Irish Terrier with canine Darier disease. The dog developed two different types of clinical lesions over a follow up time of 3 years. One lesion type presented as demarcated, proliferative, crusted and eroded to ulcerated and was present on the right concave pinna and in the ear canal. This lesion type was overlying an infundibular cyst, which represented the second type of lesion.

The crusted lesions were more severe, painful, and pruritic than the lesions described in the previously published cases [14,15]. Essentially the published cases describe one seven-month-old, male, intact English Setter that exhibited a peculiar crusting lesion on the ventral chest and two of his six living offspring that were intentionally bred by mating the affected English Setter to a normal laboratory Beagle. The two Setter-Beagle crossbred dogs developed similar lesions as the sire with alopecia, erythema, and hyperplasia on the lateral knee (one dog) or dorsal head (second dog) at the age of four and seven weeks, respectively. The lesions slightly enlarged and worsened during adolescence but remained static thereafter and did not require therapy. The histopathology and ultrastructural findings were similar in all three dogs and initially considered to represent Hailey-Hailey disease (benign familiar chronic pemphigus) [14,15]. In a subsequent study, cultured keratinocytes from one of these dogs were investigated and a depletion of ATP2A2-gated Ca2<sup>+</sup> stores was found. This finding suggested that these dogs had Darier disease rather than Hailey-Hailey disease [16].

Considering the clinical presentation of focal hyperplastic skin lesions in these dogs [14,15], their early age of onset, and the histology with severe acantholysis with prominent dyskeratosis and the formation of corps ronds and grains also suggests that they had Darier disease and not Hailey-Hailey disease as previously reported. In Hailey-Hailey disease, prominent suprabasal acantholyisis is also a feature, but loss of keratinocyte cohesion is not as complete as in Darier disease and detached keratinocytes still form clusters. Dyskeratosis is milder than in Darier disease [12].

The specific molecular mechanisms that lead to the multifocal hyperproliferation, dyskeratosis and acantholysis of epidermal keratinocytes have not yet been identified. It is well known that extracellular calcium plays a crucial role in regulating differentiation and adhesion of cultured keratinocytes [6,16]. Low levels of Ca2<sup>+</sup> induce keratinocyte proliferation while physiological levels induce cell-to-cell adhesion and keratinocyte differentiation. It has been shown that changes in the intracellular calcium homeostasis in Darier disease impair processing, transport, and assembly of calcium-dependent desmosomal proteins. Desmosomes provide strong adhesive bonds between neighboring cells by correct assembly of their intercellular and intracellular proteins and disturbance of this process results in acantholysis [6,13,16]. An alternative hypothesis is that the defective calcium homeostasis leads to delayed exit of keratinocytes from the cell cycle, which may promote secondary mutations that lead to acantholysis [16]. Furthermore, Ca2<sup>+</sup> levels in the endoplasmic reticulum play a key role in post-translational modification of proteins. Disturbances of Ca2<sup>+</sup> homeostasis may result in an accumulation of unfolded proteins in the endoplasmic reticulum and subsequent apoptosis. It has been suggested that the "corps ronds" in Darier disease may be the result of such an impaired protein folding [6].

In our patient, the crusting lesions did not enlarge, and only one similar lesion developed on other parts of the body. However, over a course of 3 years, the dog developed three infundibular cysts where the Darier specific acantholysis with dyskeratosis was seen within the cyst walls. In the epidermis overlying one cyst on the neck, a similar epidermal lesion as described for the ear developed. To the best of our knowledge, infundibular cysts have not yet been described in humans or dogs with Darier disease. In humans, several clinical variants of Darier disease including vesicobullous, hypopigmented, cornifying, zosteriform or linear, acute, and comedonal subtypes have been described [9]. Comedonal Darier disease is a very rare variant with severe follicular involvement and characterized by open or closed comedones with central keratotic plugs and the presence of greatly elongated epidermal protrusions at the base of the comedones [24,25]. However, in the canine case presented here, no comedones but true infundibular cysts without the described elongated papillary projections at the base of the cysts were present. Thus, this presentation of our case is new and has never been described in humans or dogs.

The dermal nodules were successfully excised and no recurrence was noted at the site of excision. The hyperplastic and ulcerated lesions on the pinna and the ear canal required treatment, as they were painful and prone to secondary infection at any time-point.

In human medicine, numerous therapeutic options have been described, including systemic or topical retinoids, cyclosporine, vitamin A, systemic or topical corticosteroids, topical 5-fluorouracil, keratolytics with urea, or interventional treatment, like dermabrasion, laser ablation, and excision [26–30].

The age of the dog, the difficulty in the application of topical treatments, and financial and psychological restrictions of the owner limited the treatment options in our patient. Systemic and topical glucocorticoids, as well as systemic vitamin A, were not successful. As other medical treatments were denied, we treated the lesions in the ear with a diode laser. Carbon dioxide (CO2) laser or yttrium aluminum garnet (YAG) laser ablation in Darier disease and Hailey- Hailey disease has been reported as a successful treatment option in human medicine [28–30]. We chose a diode laser, as this device can be used via the working channel of the video otoscope and allowed us to ablate the lesional skin of the medial aspect of the vertical ear canal under visual control. The concept of laser therapy in Darier disease is the ablation of the defective epidermis and the follicular infundibulum, which might be the focus of recurrence. In the described laser-treated human patients, remodeling of normal skin, as well as cicatrization, occurred. In our canine patient, the treatment was very well tolerated and led to full resolution of the lesion after several interventions. The ear canal tissue provides only a thin layer of dermis over the underlying cartilage, which is prone to necrosis if damaged. Therefore, our inventions had most likely not been aggressive enough to completely destroy the affected tissue and the follicular infundibulum in one session. In less vulnerable areas of the skin, a laser treatment might be more favorable.

Darier disease in human patients is caused by heterozygous genetic variants in *ATP2A2*. These include missense, nonsense, frameshift, and splice site variants [7,8,31]. An intronic 18 bp insertion, 12 nucleotides upstream of exon 3, caused Darier disease in one human family. This insertion altered splicing and resulted in an aberrant transcript with 6 additional codons, which could be detected as in-frame insertion that did not lead to nonsense mediated decay [31].

Our genetic analysis revealed a heterozygous SINE insertion in intron 14 of the *ATP2A2* gene, which was exclusively found in the affected dog, but not in 592 controls. The functional analysis at the mRNA level indicated nearly mono-allelic expression of *ATP2A2* transcripts in skin of the affected dog. RNA-seq showed that the SINE insertion led to the activation of cryptic splice sites in intron 14 and the inclusion of an aberrant exon containing a premature stop codon. The observed allelic imbalance of the transcripts can be plausibly explained by nonsense-mediated decay [32] of the transcripts with the premature stop codon.

A limitation of our genetic analysis is the lack of family data. We hypothesize that the SINE insertion in the affected dog is the consequence of a *de novo* mutation event. Thus, the parents of the dog are assumed to be phenotypically and genotypically wildtype. If the homozygous wildtype genotype were confirmed in both parents, this would provide proof for the hypothetical *de novo* mutation event and another strong supporting argument for the pathogenicity of the SINE insertion. Unfortunately, we did not have access to the parents of the affected dog.

Given the extensive functional knowledge on *ATP2A2* and the role of *ATP2A2* variants in human Darier disease, we nonetheless think that our data strongly suggest that the SINE insertion may be considered a candidate causative variant for the phenotype in the affected dog.

#### **5. Conclusions**

We provided a comprehensive clinical, histopathological and genetic characterization of an Irish Terrier with Darier disease. The genetic analysis revealed an intronic SINE insertion into *ATP2A2* as a candidate causative genetic variant leading to aberrant splicing and degradation of aberrant transcripts.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/11/5/481/s1, Figure S1: RNA-seq data confirming the aberrant splicing. Table S1: accession numbers of 648 dog and 8 wolf genome sequences. Table S2: private variants in the affected Irish Terrier.

**Author Contributions:** Conceptualization, M.L., T.L., M.M.W.; Data curation, V.J.; Investigation, M.L., M.D., T.L., A.B., F.L., J.H.; Supervision, M.L., T.L.; Visualization, M.L., M.D., T.L., M.M.W.; Writing—original draft, M.L., T.L., M.M.W.; Writing—review & editing, M.L., M.D., T.L., A.B., F.L., J.H., D.B., V.J., M.M.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by a grant from the Swiss National Science Foundation (CRSII3\_160738/1).

**Acknowledgments:** The authors are grateful to all dog owners who donated samples and shared health and pedigree data of their dogs.We thank Eva Andrist, Nathalie Besuchet Schmutz, and Sabrina Schenk for expert technical assistance, the Next Generation Sequencing Platform of the University of Bern for performing the high-throughput sequencing experiments, and the Interfaculty Bioinformatics Unit of the University of Bern for providing high performance computing infrastructure. We thank the Dog Biomedical Variant Database Consortium (Gus Aguirre, Catherine André, Danika Bannasch, Doreen Becker, Brian Davis, Cord Drögemüller, Kari Ekenstedt, Kiterie Faller, Oliver Forman, Steve Friedenberg, Eva Furrow, Urs Giger, Christophe Hitte, Marjo Hytönen, Vidhya Jagannathan, Tosso Leeb, Hannes Lohi, Cathryn Mellersh, Jim Mickelson, Leonardo Murgiano, Anita Oberbauer, Sheila Schmutz, Jeffrey Schoenebeck, Kim Summers, Frank van Steenbeek, Claire Wade) for sharing whole genome sequencing data from control dogs. We also acknowledge all researchers who deposited dog or wolf whole genome sequencing data into public databases.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **A Deletion in** *GDF7* **is Associated with a Heritable Forebrain Commissural Malformation Concurrent with Ventriculomegaly and Interhemispheric Cysts in Cats**

**Yoshihiko Yu 1,2,**†**, Erica K. Creighton 1,**†**, Reuben M. Buckley 1, Leslie A. Lyons 1,\* and 99 Lives Consortium** ‡


Received: 15 May 2020; Accepted: 16 June 2020; Published: 19 June 2020

**Abstract:** An inherited neurologic syndrome in a family of mixed-breed Oriental cats has been characterized as forebrain commissural malformation, concurrent with ventriculomegaly and interhemispheric cysts. However, the genetic basis for this autosomal recessive syndrome in cats is unknown. Forty-three cats were genotyped on the Illumina Infinium Feline 63K iSelect DNA Array and used for analyses. Genome-wide association studies, including a sib-transmission disequilibrium test and a case-control association analysis, and homozygosity mapping, identified a critical region on cat chromosome A3. Short-read whole genome sequencing was completed for a cat trio segregating with the syndrome. A homozygous 7 bp deletion in *growth di*ff*erentiation factor 7* (*GDF7*) (c.221\_227delGCCGCGC [p.Arg74Profs]) was identified in affected cats, by comparison to the 99 Lives Cat variant dataset, validated using Sanger sequencing and genotyped by fragment analyses. This variant was not identified in 192 unaffected cats in the 99 Lives dataset. The variant segregated concordantly in an extended pedigree. In mice, *GDF7* mRNA is expressed within the roof plate when commissural axons initiate ventrally-directed growth. This finding emphasized the importance of *GDF7* in the neurodevelopmental process in the mammalian brain. A genetic test can be developed for use by cat breeders to eradicate this variant.

**Keywords:** feline; *Felis catus*; brain malformation; BMP12; neurodevelopment; genetics; genomics; mendelian traits; genome-wide association study; whole genome sequencing

#### **1. Introduction**

Congenital brain malformations in humans are caused by genetic variants, in utero infection, or other environmental factors. Dogs and cats are also occasionally diagnosed with congenital brain malformations (reviewed in [1]), which are noted as breed predispositions, familial aggregations, or sporadic cases, especially in dogs [2–6]. Congenital hydrocephalus is common in toy and brachycephalic dog breeds, such as the Maltese, Yorkshire terrier, Chihuahua, toy poodle and pug dogs [7]. Widespread in Cavalier King Charles Spaniels, Chiari-like malformation is a common cause of foramen magnum obstruction, and results in the secondary syringomyelia in dogs, characterized by the mismatch of size between the brain and the skull [8].

Similarly, high grades of brachycephaly in cats are also associated with malformations of the calvarial and facial bones, as well as dental malformations or respiratory abnormalities [9–12]. A familial craniofacial malformation with meningoencephalocele has been recognized in Burmese cats [13], which is caused by *ALX Homeobox 1* (*ALX1*) variant [14]. However, feline brain malformations with (suspected) idiopathic nature are mostly reported as sporadic events [15–20]. Overall, the genetic factors contributing to brain (mal) formation and structural congenital brain disease in dogs and cats are largely unknown.

In an effort to develop a breed of cats having similar phenotypes to a tiger, including a small rounded ear, a mixed breed cat derived from the Oriental cat breed was discovered to have small rounded ears and hence, was used as a foundation sire for a breeding program. Outcross and backcross breeding indicated the phenotype was autosomal recessive [21]. However, a magnetic resonance imaging (MRI) examination of a kitten with the desired ear phenotype, which had an accidental head injury from a fall, indicated the presence of congenital hydrocephalus. Additional MRIs of the breeding stock suggested cats with the ear phenotype had congenital brain malformations. These cats have small rounded ear pinnae and doming of the head (Figure 1). This extended family of mixed-breed cats derived from the Oriental breed has been characterized clinically and histopathologically with forebrain commissural malformation concurrent with ventriculomegaly and interhemispheric cysts [21]. The forebrain malformations include dysgenesis of the septum pellucidum, interthalamic adhesion, and all the midline commissures, excluding the rostral white commissure, as well as hippocampal hypoplasia. Clinical symptoms include mild generalized ataxia when walking, and mild to marked postural reaction deficits, although cranial nerve examination and segmental reflexes are within normal limits. All the cats with neurological signs have midline and limbic structure abnormalities, dilated ventricles and hemispheral cysts with or without a suprapineal cyst. These findings resemble a mild variant of holoprosencephaly (HPE) in human (OMIM: 236,100 and others). Although variations in the severity of the forebrain commissural malformation were seen, most affected cats are hydrocephalic. No chromosomal abnormalities are noted in a karyotypic analysis of the cats. Segregation analysis suggests an autosomal recessive mode of inheritance; however, the causal variant remained unknown [21].

As a result of the potentially harmful impacts associated with the trait, the breeder promptly discontinued the breeding program and altered subsequent cats. However, some carriers for the trait had already been adopted for other breeding programs. A group of affected cats were presented to the researchers for pathological and genetic studies. Sample collection from the cats in the owner's breeding program and cats from controlled breeding within the university colony supported the genetic investigation of the abnormal brain development and mode of inheritance.

Genome-wide association studies (GWAS), using a sib-transmission disequilibrium test (sib-TDT) and a case-control analysis, and homozygosity mapping were conducted to detect an associated genomic region for the syndrome using genotypes from a feline single nucleotide polymorphism (SNP) DNA array [22]. Whole genome sequencing (WGS) was conducted on a cat trio segregating for the syndrome to define the location and identify candidate variants.

#### **2. Materials and Methods**

#### *2.1. Sampling and Pedigree*

All procedures were performed with an approved University of Missouri (MU) Institutional Animal Care and Use Committee protocol (ACUC protocol # 8292). Four affected and two carrier cats were donated and housed at the MU colony for controlled breeding. Additional buccal swab and cadaver samples from an external breeding program were provided voluntarily by the breeder/owner (*N* = 129). DNA samples were extracted using DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA). The quality of the DNA samples was visualized and confirmed by agarose gel electrophoresis. DNA samples whose concentration was insufficient were whole genome amplified, using the REPLI-g Mini Kit (Qiagen). The relationship of the ascertained cats was confirmed using short tandem repeat

(STR) markers, as previously described [23]. Parentage analysis was performed using the computer program COLONY [24,25]. Clinical and histopathological features of the syndrome were characterized previously [21]. Although some cats were phenotyped based on MRI and/or histopathology, most cats were assumed to have the brain malformation based on the ear morphology, since clinically healthy cats had elongated (normal) ears and clinically affected cats had the small, rounded ear type [21] (Figure 1). Images or cadavers of cats were not always available.

**Figure 1.** Domestic cats with heritable forebrain commissural malformation. Note the abnormal presentation of the pinnae used to determine affection status. (**a**) Frank—affected sire (left). (**b**) Camilla—carrier dam. (**c**) Bobble—affected offspring. These three cats (**a**–**c**) were whole genome sequenced. (**d**) Transverse plane of T2-weighted magnetic resonance imaging of an affected cat at the level of the thalamus. Severe ventriculomegaly, thinning of the cerebral parenchyma and midline structure deficits are seen. A part of the parietal lobe is deficient. (**e**) Mid-sagittal plane of T2-weighted magnetic resonance imaging of an affected cat (the same cat as (**d**)). Midline structure deficits are recognized. Note that the spinal cord is formed normally. Interhemispheric cysts are also seen at the rostrotentorial region and the quadrigeminal cistern. Due to the presence of cysts, cerebellar herniation is seen. (**f**) Gross dorsal view of the dissected head at necropsy. The skin was removed, and the skull was exposed. (**g**) Transverse sections of formalin-fixed brain tissue at the level of frontal lobe and thalamus. Severe ventriculomegaly, thinning of the cerebral parenchyma and midline structure deficits are seen. Note that a cat whose magnetic resonance imaging of (**d**) and (**e**) are presented here is different from cats whose gross pathological pictures of (**f**) and (**g**) are provided here.

#### *2.2. DNA Array Genotyping*

Fifty-two genomic DNA samples (~600 ng each) were submitted to GeneSeek (Neogene, Lincoln, NE, USA) for SNP genotyping on the Illumina Infinium Feline 63K iSelect DNA Array (Illumina, San Diego, CA, USA) [22]. The original SNP positions were based on an early assembly of the cat genome [26], and have been since relocalized to the latest feline genome assembly, Felis\_catus\_9.0. The SNP positions based on the Felis\_catus\_9.0 assembly were used for the analyses and the required map file is available. [27]. Quality control of the SNP data was performed using PLINK (v1.07) [28]. The following criteria were applied: (i) individuals with genotyping success rate of <80% were removed (–mind 0.2); (ii) SNP markers with a genotyping rate <80% were removed (–geno 0.2); and (iii) SNPs with a minor allele frequency of 0.05 or less were removed (–maf 0.05). Furthermore, SNPs that were previously reported to have missing ≥10% of genotypes and Mendelian errors [22], and that remained after quality controls were excluded.

#### *2.3. Genome-Wide Association Studies*

After the SNP pruning described above, GWAS were conducted using PLINK. Sib-TDT [29] was performed using the DFAM procedure in PLINK (–dfam). This method implements sib-TDT and also includes unrelated individuals in the analysis. A case-control association analysis was performed (–assoc). The genomic inflation factor was calculated using the function (–adjust). Multi-dimensional scaling (MDS) analysis was conducted (–genome) and MDS plots were generated to visualize the population stratification, using PLINK and R software (version 3.3.3; R Foundation for Statistical Computing, Vienna, Austria), respectively. A quantile-quantile (QQ) plot was created using R. Genome-wide significance for both analyses, which was determined using 100,000 permutations (–mperm 100000). Manhattan plots from the sib-TDT, case-control association and permutation analyses were generated using R. The MDS plot was used to reselect cats to minimize stratification between cases and controls for the secondary case-control association analysis, by visual interpretation.

#### *2.4. Haplotype Analysis*

An approximately 6 Mb region surrounding highly associated SNPs was extracted, including 81 SNPs, from SNP chrA3.163737349 at chromosome position A3: 123,014,546 to SNP chrA3.156620632 at chromosome position A3: 128,837,125. The haplotype boundaries were visually confirmed using Haploview (version 4.2) [30]. Linkage disequilibrium (LD) blocks were identified using the solid spine of LD method in Haploview. Haplotype sequences are estimated using an accelerated EM algorithm, as implemented in Haploview. When analyzing LD blocks and haplotypes, SNPs with MAF of 0% were allowed and included, because most cases showed the consistent genotypes at each SNP.

#### *2.5. Homozygosity Analysis*

Homozygosity analysis was performed using PLINK. SNPs within a 1000 kb window, containing at least 25, were investigated for runs of homozygosity (–homozyg-window-kb 1000, –homozyg-snp 25). In each window, five missing genotypes (20%) and a single heterozygote (2%) were tolerated (–homozyg-window-missing 5, –homozyg-window-het 1). The threshold of homozygosity match was set as 0.99 (–homozyg-match 0.99). A homozygous block was characterized by five SNPs (~200–250 kb). Consensus homozygosity blocks were identified as overlaps between individual homozygosity blocks (–consensus-match, –homozyg-group).

#### *2.6. Whole Genome Sequencing*

A trio of cats including an affected sire, a carrier dam and an affected offspring was selected for WGS as part of the 99 Lives Cat Genome Sequencing Initiative (http://felinegenetics.missouri.edu/99lives). These cats were produced at the MU colony; thus, the parentage was known. DNA extraction and library preparation were conducted as previously described [31]. A minimum of 4 μg genomic DNA was submitted for WGS to the MU DNA Core Facility. Two PCR-free libraries with insertion sizes of 350 bp and 550 bp were constructed for each cat using the TruSeq DNA PCR Free library preparation kit (Illumina). The Illumina HiSeq 2000 (Illumina) was used to generate sequence data.

Sequence reads were mapped to the latest feline genome assembly, Felis\_catus\_9.0, and processed as previously described [27]. Briefly, read mapping was conducted with Burrows-Wheeler Aligner (BWA) version 0.7.17 [32]. Duplicates were marked using Picard tool MarkDuplicates (http://broadinstitute. github.io/picard/). Potential insertions or deletions (indels) realignment was performed using the Genome Analysis Tool Kit (GATK version 3.8) [33] IndelRealigner. Variants were called using GATK HaplotypeCaller in gVCF mode [34]. VarSeq v2.0.2 (Golden Helix, Bozeman, MT, USA) was used to annotate variants with Ensembl 99 gene annotations and identify variants unique to the trio cats and absent from 192 unaffected unrelated domestic cats. Exonic variants were extracted from the dataset, including variants 21 bp flanking the exons to ensure inclusion of variants that may affect splice donor and accept sites. Candidate variants segregating across the trio were visualized using Integrative Genomics Viewer (IGV) [35].

#### *2.7. Variant Validation and Genotyping*

PCR and Sanger sequencing were performed to validate the 7 bp deletion in the candidate gene *GDF7* for cats that were submitted to WGS. The primer sequences were: forward primer: 5 -AGCGACATCATGAACTGGTG-3 , reverse primer: 5 -CCACGGAGCCCATGGACC-3 . PCR was performed using AccuPrime GC-Rich DNA Polymerase (Invitrogen, Carlsbad, CA, USA). PCR was performed following the manufacturer's instructions, with the annealing temperature of 61 ◦C and 35 cycles. PCR amplicon was purified using QIAquick Gel Extraction Kit (Qiagen), or using ExoSAP-IT PCR Product Cleanup Reagent (Thermo Fisher Scientific, Waltham, MA, USA). Sanger sequencing was conducted at the MU DNA Core Facility using an Applied Biosystems 3730xl DNA Analyzer (Applied Biosystems, Foster City, CA, USA) with BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems).

Fragment analysis was conducted for population screening. PCR conditions and reagents used were the same as above, except the forward primer was fluorescein amidite [FAM] labeled at the 5 end. Fragment analysis was conducted at the MU DNA Core Facility using an Applied Biosystems 3730xl DNA Analyzer (Applied Biosystems). The expected wildtype fragment size was 294 bp, while the mutant fragment size was expected as 287 bp. Amplicons were analyzed using STRand software [36].

#### **3. Results**

#### *3.1. Pedigree and Genotyping*

Using 18 STRs, the parentage for 69 of 129 cats was determined with a high likelihood using the COLONY software [24,25] (data not shown), producing a pedigree of 79 cats (Figure S1). For GWAS, 52 cats were selected using owner provided and pedigree information, including 26 cases, and 26 controls, in which 43 cats were included in the pedigree (Figure S1). Cat DNA samples were genotyped on Feline 63K SNP array (File S1). Selection criteria for genotyping focused on cats that were as unrelated as possible. Nine cats with call rates below 80% were removed, and 478 SNPs were removed with missingness rates > 20%. An additional 22,297 SNPs were also removed with minor allele frequencies < 0.05. After filtering, 20 cases and 23 controls remained with a genotyping rate of 0.977 across 40,263 SNPs. Furthermore, 372 SNPs were excluded, due to missing ≥10% of genotypes and Mendelian errors previously reported [22]. The GWAS was conducted with 39,891 SNPs.

#### *3.2. Association Studies*

Sib-TDT was conducted on the pedigree formed by the 20 cases and 23 controls. After permutation testing, no SNPs were significant; however, nine SNPs with the highest, the second-highest, or the third-highest association were localized to cat chromosome A3:123,055,238–128,667,138 on the Felis\_catus\_9.0, extending approximately 5.6 Mb (Table 1). The result of the sib-TDT analysis was presented as a Manhattan plot (Figure 2a). In the initial case-control association analysis, 65 SNPs had genome-wide significance and were located cat chromosome A3: 116,714,934–129,668,450, extending ~13.0 Mb and C1: 105,429,018–115,412,315, extending ~10.0 Mb (Table 1). However, the genomic inflation factor was 1.89; thus, the MDS plot (Figure S2) was used to reselect cases and controls for the analysis. A second case-control association analysis was performed with 14 cases and nine controls, and the genomic inflation factor was reduced to one. Seventeen SNPs showed genome-wide significance and were located cat chromosome A3: 119,105,247–129,372,537, encompassing ~10.3 Mb (Figure 2b, Table 1). This chromosome A3 region encompassed the entire region suggested by the sib-TDT, and was within the initial case-control association analysis.


**Table 1.** Single nucleotide polymorphism (SNP) associations for cats with heritable forebrain commissural malformation in the sib-transmission disequilibrium test (sib-TDT) and case-control association analyses.

*p*-values were presented with up to four decimal places. \* SNP IDs are based on an early cat genome assembly [26] † Positions based on current cat genome assembly [27].

**Figure 2.** Manhattan plot of the genome-wide association studies (GWAS) for heritable forebrain commissural malformation in cats. Cats (20 cases and 23 controls) were genotyped on the Illumina Infinium Feline 63K iSelect DNA Array (Illumina, San Diego, CA, USA) and used for GWAS. In both panels of (**a**) and (**b**), the upper plots exhibit the *Praw* value of the analysis, while the lower exhibits the *Pgenome* values after 100,000 permutations. Red horizontal lines indicate genome-wide significance (*Pgenome* = 0.05, −log10 = 1.3). (**a**) Sib-TDT analysis. Genome-wide significance was not achieved. (**b**) Case-control association analysis. Significant association is localized to chromosome A3 for 17 SNPs. The genomic inflation was 1.

#### *3.3. Haplotype Analysis*

The 6 Mb region, on chrA3: from approximately 123 to 129 Mb and encompassing the overlapped region identified in GWAS, was visually inspected for common haplotypes using Haploview. In affected cats, a large extended LD block encompassing approximately 4.3 Mb (A3: 123,082,369–127,348,216) was identified with a 95% frequency of the sequential haplotype. Considering that two cats had 82.7% and 91.4% genotyping rate, one cat had 98.8% and the others had 100% genotyping rate in this area, a few missing produced the remaining haplotypes (File S2). Short and discontinuous LD blocks are identified by Haploview in controls. There are various haplotype sequences and frequencies approximately within the 6 Mb regions in unaffected cats.

#### *3.4. Homozygosity Analysis*

Homozygosity mapping was performed on 20 cases and 23 controls. The homozygosity analysis identified the same location on chromosome A3 in 18 of 20 affected cats, excluding the same two cases that did not have sufficiently high genotyping rates, with A3: 125,601,560–127,684,693, spanning approximately 2.1 Mb, and no unaffected cats were homozygosity (Table S1). The region was identified by the two genome-wide association analyses (Table 1). Although other ROHs were identified, none were specific to cases or as extensive.

#### *3.5. Whole Genome Sequencing*

Cat genomes have been submitted to the NCBI short read archive under BioProject: PRJNA528515; Accessions PRJNA343385; SRX2654400 (Sire), SRX2654398 (dam) and SRX2654399 (offspring). Genome sequence analyses and variant calling for the 99 Lives project has been previously described [37]. Approximately 2.5 million variants were ascertained across 195 cats in the exonic portion of the dataset, which included 21 bp of exon flank sequence. No candidate genes were identified on cat chromosome A3 during the initial analysis, when considering the sire and offspring to be homozygous affected, and considering the dam as an obligate carrier for an alternative allele (Table 2). Only an intergenic variant (C1:106,990,675) and an intronic variant in *sperm antigen with calponin homology and coiled-coil domains* (*SPECC1*) (E1:9,973,078) met the segregation criteria. Using relaxed constraints, where affected cats were allowed to also be considered as carriers, four more variants were identified (C1:96,095,693, C1:96,839,645 and D2:33,368,378) with only one variant located within the critical region and also in a gene coding region (Table 2). This variant was a 7 bp deletion in the coding region of *GDF7* (c.221\_227delGCCGCGC [p.Arg74Profs\*17]) at the position A3:127002233 (ENSFCAT00000063603). The variant was identified as homozygous in the affected sire, heterozygous in the obligate carrier dam, heterozygous in the affected offspring, and absent from the other 192 domestic cats. Although each cat in the trio had an average of ~30× genome coverage, the sire had 18× coverage within the region, the dam had ~14× coverage with seven reads per allele, and the affected offspring had ~16× coverage, with only one of the reads representing the reference allele, likely misrepresenting the offspring as heterozygous, and visual inspection with IGV suggested the affected offspring was instead very likely homozygous for the variant (Figure 3). The affected cat was confirmed as a homozygote for the alternate allele by genotyping. The *GDF7* variant was predicted to cause a truncated protein with 89 amino acids, while the wildtype protein has 455 amino acids (Figure S3). Feline *GDF7* amino acid sequence is predicted to be 86.2%, 90.1%, 84.6%, 77.8% and 77.2% identical to human, horse, cow, rat and mouse, respectively (Figure S3). In addition, comparison of the *GDF7* locus between the Felis\_catus\_9.0 and Felis\_catus\_8.0 genome assemblies, revealed the region containing the *GDF7* candidate variant is absent from the Felis\_catus\_8.0 assembly, indicating the importance of the updated reference genome for trait discovery.


**Table 2.** Variants identified in 99 Lives whole genome sequence dataset considering segregation within the trio.

\* Only the dam should be heterozygous for the variant in the dataset and the sire and offspring homozygous for the variant. Sequence data was poor within the critical region on cat Chromosome A3, and the affected offspring was erroneously considered heterozygous by the Genome Analysis Tool Kit (GATK version 3.8). Four variants are identified when the offspring is considered heterozygous.

**Figure 3.** Depiction of the whole genome sequence reads using the Integrated Genome Viewer (IGV) of (**a**) affected sire, (**b**) carrier dam, and (**c**) affected offspring for the *GDF7* variant. Grey horizontal bars represent individual reads, while grey vertical bars at the top of each sub figure represent depth of coverage. Notice in affected individuals (**a**,**c**), coverage is close to zero in the deleted region, while, in the carrier sequencing (**b**), coverage is approximately 50%. Additionally, in affected individuals (**a**,**c**), reads with high numbers of mismatches are indicative of the misidentification of an indel, and tend to occur near the ends of reads.

#### *3.6. Variant Validation and Genotyping*

Sanger sequencing was performed to confirm the identified *GDF7* c.221\_227delGCCGCGC in affected and obligate carrier cats, including the cats in the WGS trio. The 7 bp deletion in *GDF7* was screened in 25 affected, 39 unaffected, and two cats with unknown phenotype in the extended pedigree using fragment analysis (Figure S4). Both unknown cats were homozygous for the variant allele. Overall, 13 of 14 suspected wildtype cats in the extended pedigree were concordant, and one cat genotyped as a heterozygote. Of 25 suspected carriers, 23 genotyped as heterozygote and two as wildtype normal. Of 22 suspected affected cats, 20 genotyped as homozygous for the variant, one as heterozygous and one as wildtype normal.

#### **4. Discussion**

Brain malformations are occasionally identified in veterinary practice. However, little is known about the genetic causes and interactions for brain malformation. Due to the health concerns associated with breed development, particularly in dog breeds [38,39], many breeders have become more vigilant to health-associated consequences of selection based on morphological phenotypes. Feline brain malformation syndrome seen in this extended family happened to be generated in the course of breeding selection for the ear morphological phenotype.

Most of the cat samples had been archived as frozen cadavers by the breeder, and later provided to the researchers. As a result of poor documentation of relationships and disease status, a pedigree was established by determining parentage using STRs, age, and gender of the cats and from interviews with the breeder. Ear phenotypes, which were used as a proxy for disease, were difficult to determine from frozen cadavers. Due to the significant inbreeding and backcrossing required to maintain the phenotype, 18 STRs were often insufficient to determine parentage. However, some known breedings were available from the university colony. Overall, an extended pedigree was developed, and was expected to be sufficient for GWAS and WGS investigations for the causal variant. Furthermore, a variant dataset from WGS of domestic cats, the 99 Lives Cat Genome Sequencing Initiative, which has revealed the causative variants for several cat diseases and traits in the last several years [31,40–46], was considered to facilitate the variant filtering to find the private variants.

In humans, HPE is the most common malformation of the prosencephalon, and its prevalence is approximate 1 in 10,000 births [47]. A common feature of HPE includes the incomplete separation of the anterior part of the forebrain or telencephalon. The previous study indicated this feline heritable brain malformation syndrome resembled a mild form of HPE [21]. Many genes have been reported to cause HPE in humans (reviewed in [48–50]). However, *GDF7*, also known as *bone morphogenetic protein 12* (*BMP12*), has not been reported to be associated with HPE in humans. Initially, *GDF7* activity was shown to be required for the specification of neuronal identity in the spinal cord [51]. *GDF7* mRNA is expressed within the roof plate, when commissural axons initiate to grow ventrally-directed. Furthermore, *GDF7*-null mutant mice show hydrocephalus, and they show considerable variation in the location of the dilated ventricle [51]. This evidence supports these findings that the frameshift mutation in *GDF7* causing the truncated protein is highly likely to be associated with this heritable brain malformation syndrome in cats. Transcriptomic and proteomic analyses would be essential to ascertain that this *GDF7* variant causes heritable forebrain commissural malformation in cats.

The variable severity of this syndrome in the cat pedigree was reported previously [21]. In humans, heterogeneity in familial HPE is also identified even if different individuals are carrying the same mutation [52–54]. The influence of environmental or teratogenic factors or modifier genes have been suggested for the spectrum (reviewed in [47,48,50]). Assuming no exposure to teratogen and relatively homogeneous living environment, the presence of modifier genes is suspected for the variable severity of the dilated ventricles and supratentorial cysts in cats presented here.

Bone morphogenetic proteins (BMPs) belong to the transforming growth factor-β (TGF-β) superfamily of proteins that are involved in many functions such as cell proliferation, differentiation, apoptosis, cell fate determination and morphogenesis [55]. The BMPs also play various roles in the

neural development [56]. Among them, *GDF7* also known as BMP12, plays an essential role in bone and cartilage formation as well [57]. Except for hydrocephalus seen in *GDF7*-null mutant mice [51], several phenotypes caused by *GDF7* deficient mice have been reported, including the subtle effect on Achilles tendon [58], increased endochondral bone growth [59], seminal vesicle defects and sterility [60], and smaller bone cross-sectional geometric parameters [61]. In addition, a variant in *GDF7* (rs3072) has been reported to increase risk for Barrett's esophagus and esophageal adenocarcinoma [62,63]. Although, to the authors' knowledge, there was no report about the involvement of *GDF7* in ear or skull morphology, there is a possibility that small rounded pinnae and/or domed craniums may be influenced by the *GDF7* variant, because *GDF7*, also known as BMP12, has been considered to play a negative role on chondrogenesis [59], and to be involved in the structural integrity of bone [61].

In conclusion, the combination of GWAS, homozygosity mapping and WGS identified a 7 bp deletion in *GDF7* (c.221\_227delGCCGCGC), which is the most likely variant causing feline forebrain commissural malformation, concurrent with ventriculomegaly and interhemispheric cysts in this domestic cat lineage, although the functional analysis has not been achieved to prove the deterministic mechanism. Furthermore, this study highlights the importance of *GDF7* in the neurodevelopmental course in cats, and brings new insight into neurodevelopmental biology. Cat breeders can now perform a genetic test to eradicate the *GDF7* mutation from the breeding population.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/11/6/672/s1. Table S1: Regions of homozygosity was unique to 18 cats with the inherited forebrain commissural malformation, and were absent in all the unaffected cats. Figure S1: Pedigree of cats segregating for an autosomal recessive forebrain commissural malformation. Relationships of 79 cats (27 nuclear families) provided by the breeder and confirmed with genetic testing of short tandem repeats when possible. Arrow indicates the proband. Circles indicate females, squares indicate males, and diamonds indicate unknown sex. Filled symbols represent cats with small rounded ears, which were suspected to have forebrain commissural malformation concurrent with ventriculomegaly and interhemispheric cysts. Half-filled represent obligate carriers. Symbols with question marks represent cats with unknown phenotype. A symbol with no fill indicates the cat is known to be completely unrelated and not expected to be a carrier. The cats genotyped on the DNA array and used for genome-wide association studies and homozygosity mapping are indicated by a "T" on the upper left of the symbol (The nine cats removed by quality control are not indicated). A black filled circle at the left bottom of symbol are individuals that were whole genome sequenced. Cats with a bar above the symbol were confirmed by magnetic resonance imaging. Cats with an open circle to the upper right had histology performed at necropsy. The cats' ID/name is indicated below the symbol. Size in basepairs of the genotypes for the 7 bp *GDF7* indel are indicated below each cat available. Figure S2: Multi-dimensional scaling plot and quantile-quantile plot of cases and controls for genome-association analyses. (a) Multi-dimensional scaling (MDS) plot of cats used for the initial case-control association analysis. The genomic inflation was 1.89. Therefore, cats clustered within the blue rectangular area were selected for the second case-control association analysis as visual inspection suggests less stratification between cases and controls. The genomic inflation factor was reduced to 1. (b,c) The quantile-quantile plots of cats used for the initial (b) and second (c) analyses demonstrate the observed versus expected–log(p) values. Figure S3: Protein sequence alignment of *GDF7* in cats (*Felis catus*) and other species. GDF7 protein sequences are aligned from wildtype cat (*Felis catus*), *GDF7* mutant cat, cow *(Bos Taurus*: NP\_001193030.1 [ARS-UCD1.2]), horse (*Equus caballus*: XP\_023475218.1 [EquCab3.0]), mouse (*Mus musculus*: NP\_001299805.1 [GRCm38.p4]), and rat (*Rattus norvegicus*: XP\_006239940.1 [Rnor\_6.0]). Identical amino acids to those of Felis catus sequence are represented as a dot (.). Deleted amino acids are represented as a dash (–). A 7 bp deletion causes a frameshift and changes the amino acid sequence from 74th position (highlighted in yellow), starting with an arginine to a proline change, which results in the truncated protein with a stop codon 17 amino acids downstream. Figure S4. Variant validation by Sanger sequencing and fragment analysis. (a) Sanger sequence of a wildtype and homozygous affected cat for the 7 bp *GDF7* variant (boxed region). (b) Fluorescence-based fragment analysis using an ABI 3730XL for the *GDF7* variant. Left—homozygous wildtype with 294 bp fragment, middle—heterozygous with 287 and 294 bp fragments, and right—affected with 287 bp fragment. LIZ standard (Applied Biosystems, Foster City, CA, USA) was used to size DNA fragments. File S1: Ped file for PLINK of cats genotyped using Illumina Infinium Feline 63K iSelect DNA Array. File S2: SNPs (*n* = 81) forming common haplotype for cats in the association studies.

**Author Contributions:** Conceptualization, L.A.L.; Methodology, L.A.L.; Software, R.M.B., Y.Y.; Validation, Y.Y.; Formal Analysis, E.K.C., Y.Y.; Investigation, E.K.C., Y.Y., R.M.B.; Resources, L.A.L., E.K.C.; Data Curation, L.A.L., R.M.B.; Writing—Original Draft Preparation, Y.Y., E.K.C.; Writing—Review & Editing, Y.Y., R.M.B., E.K.C., L.A.L.; Visualization, Y.Y., R.M.B., E.K.C.; Supervision, L.A.L.; Project Administration, L.A.L.; Funding Acquisition, L.A.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by in part by NIH Office of Research Infrastructure Programs (OD R24OD01092), Winn Feline Foundation (MT-13-010), the Cat Health Network (D14FE-552) and the MU Gilbreath McLorn Endowment for Comparative Medicine (L.A.L.). The authors thank the JSPS Overseas Challenge Program for Young Researchers (2017–2018) for sponsoring the visiting scholarship (Y.Y.) and the financial support from Mars, Inc. (R.M.B.).

**Acknowledgments:** We also thank Barbara Gandolfi and Thomas R. Juba for technical assistance and assistance with figures from Karen Clifford. 99 Lives Consortium (2019 cat analysis—99Lives195) Organizer: Leslie A. Lyons 1; Data analyst: Reuben M. Buckley 1; Each member of the 99 Lives Consortium (2019 cat analysis – 99Lives195) has provided at least one >15x coverage genome of the domestic cat or a wild felid to support the analyses of the dataset. Members: Reuben M. Buckley 1, Danielle Aberdein 2, Paulo C. Alves 3,4, Gregory S. Barsh 5,6, Rebecca R. Bellone 7, Tomas F. Bergström 8, Adam R. Boyko 9, Jeffrey A. Brockman 10, Margret L. Casal 11, Marta G. Castelhano 12, Ottmar Distl 13, Nicholas H. Dodman 14, N. Matthew Ellinwood 15, Jonathan E. Fogle 16, Oliver P. Forman 17, Dorian J. Garrick 2,15, Edward I. Ginns 18, Jens Häggström 19, Robert J. Harvey 20, Daisuke Hasegawa 21, Bianca Haase 22, Christopher R. Helps 23, Isabel Hernandez 24, Marjo K. Hytönen 25, Maria Kaukonen 25, Christopher B. Kaelin 5,6, Tomoki Kosho 26, Emilie Leclerc 27, Teri L. Lear 28, Tosso Leeb 29, Ronald H.L. Li 30, Hannes Lohi 25, Maria Longeri 31, Mark A. Magnuson 32, Richard Malik 33, Shrinivasrao P. Mane 34, John S. Munday 2, William J. Murphy 35, Niels C. Pedersen 36, Simon M. Peterson-Jones 37, Max F. Rothschild 15, Clare Rusbridge 38, Beth Shapiro 39, Joshua A. Stern 36, William F. Swanson 40, Karen A. Terio 41, Rory J. Todhunter 12, Wesley C. Warren 42, Elizabeth A. Wilcox 12, Julia H. Wildschutte 43, Yoshihiko Yu 21, Leslie A. Lyons 1.


**Conflicts of Interest:** Authors disclose no conflict of interest. The funding sponsors had no role in the design, execution, interpretation, or writing of the study. The authors may receive supportive funds from a genetic testing laboratory that would offer this variant as a commercialized test in the future.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

### **Mitochondrial PCK2 Missense Variant in Shetland Sheepdogs with Paroxysmal Exercise-Induced Dyskinesia (PED)**

**Jasmin Nessler 1,**†**, Petra Hug 2,**†**, Paul J. J. Mandigers 3, Peter A. J. Leegwater 3, Vidhya Jagannathan 2, Anibh M. Das 4, Marco Rosati 5, Kaspar Matiasek 5, Adrian C. Sewell 6, Marion Kornberg 7, Marina Ho**ff**mann 8, Petra Wolf 9, Andrea Fischer 10, Andrea Tipold <sup>1</sup> and Tosso Leeb 2,\***


Received: 9 June 2020; Accepted: 8 July 2020; Published: 9 July 2020

**Abstract:** Four female Shetland Sheepdogs with hypertonic paroxysmal dyskinesia, mainly triggered by exercise and stress, were investigated in a retrospective multi-center investigation aiming to characterize the clinical phenotype and its underlying molecular etiology. Three dogs were closely related and their pedigree suggested autosomal dominant inheritance. Laboratory diagnostic findings included mild lactic acidosis and lactaturia, mild intermittent serum creatine kinase (CK) elevation and hypoglycemia. Electrophysiological tests and magnetic resonance imaging of the brain were unremarkable. A muscle/nerve biopsy revealed a mild type II fiber predominant muscle atrophy. While treatment with phenobarbital, diazepam or levetiracetam did not alter the clinical course, treatment with a gluten-free, home-made fresh meat diet in three dogs or a tryptophan-rich, gluten-free, seafood-based diet, stress-reduction, and acetazolamide or zonisamide in the fourth dog correlated with a partial reduction in, or even a complete absence of, dystonic episodes. The genomes of two cases were sequenced and compared to 654 control genomes. The analysis revealed a case-specific missense variant, c.1658G>A or p.Arg553Gln, in the *PCK2* gene encoding the mitochondrial phosphoenolpyruvate carboxykinase 2. Sanger sequencing confirmed that all four cases carried the mutant allele in a heterozygous state. The mutant allele was not found in 117 Shetland Sheepdog controls and more than 500 additionally genotyped dogs from various other breeds. The p.Arg553Gln substitution affects a highly conserved residue in close proximity to the GTP-binding site of PCK2. Taken together, we describe

a new form of paroxysmal exercise-induced dyskinesia (PED) in dogs. The genetic findings suggest that PCK2:p.Arg553Gln should be further investigated as putative candidate causal variant.

**Keywords:** *Canis lupus familiaris*; whole genome sequencing; dog; mitochondrion; phosphoenolpyruvatecarboxykinase; inborn error of metabolism; precision medicine

#### **1. Introduction**

Paroxysmal movement disorders are a group of diverse neurological conditions characterized by the episodic occurrence of involuntary movements. In most cases with such disorders, patients have a normal interictal examination [1]. In human medicine, paroxysmal movement disorders are classified into paroxysmal dyskinesias (PxDs) and episodic ataxias (EAs). The PxDs are further subdivided into four related forms, paroxysmal kinesigenic dyskinesia (PKD), paroxysmal non-kinesigenic dyskinesia (PNKD), paroxysmal hypnogenic dyskinesia (PHD), and paroxysmal exercise-induced dyskinesia (PED) [1].

PED in humans is most frequently due to genetic variants in the *SLC2A1* gene encoding the GLUT1 transporter mediating glucose transfer across the blood–brain barrier [1–4]. Dominant and recessive forms of *SLC2A1* related PED have been described. Depending on the specific variant, the PED may occur isolated [2,3] or in combination with other phenotypes, such as epilepsy, delayed development and mental retardation [4]. In other human patients with isolated or syndromic PED forms, genetic variants in *GCH1* or *PARKN* have been described [5,6].

In veterinary medicine, several breed-specific episodic movement disorders characterized by spasticity have been reported [7]. Their etiopathophysiology is heterogenous in different breeds and the causal genetic variants are only partially known. The so-called Scottie Cramp in Scottish Terriers was already recognized 50 years ago and is characterized by generalized or hind limb spasticity. The molecular cause has not yet been reported in the scientific literature [8–10]. Related phenotypes also with unclear causative genetic defects were reported in Bichon Frisé [11,12], Border Terriers [13–15], Boxers [16], Chinooks [17], German Shorthair Pointers [18], Jack Russell Terriers [19] and Maltese dogs [20].

In Soft-Coated Wheaten Terriers, an autosomal recessive paroxysmal dyskinesia is caused by a variant in the *PIGN* gene encoding the phosphatidylinositol glycan anchor biosynthesis class N (OMIA 002084-9615) [21]. Episodic falling syndrome in Cavalier King Charles Spaniels is an autosomal recessive disorder caused by variants in the *BCAN* gene encoding the brain-specific extracellular matrix proteoglycan brevican (OMIA 001592-9615) [22,23].

In this manuscript, we describe the clinical and diagnostic findings, treatment and outcome of four Shetland Sheepdogs with a paroxysmal movement disorder classified as PED together with our efforts to elucidate the underlying causative genetic defect. The study was conducted as a retrospective multi-center investigation.

#### **2. Materials and Methods**

#### *2.1. Ethics Statement*

All animal experiments were performed according to local regulations. All dogs in this study were privately owned and examined with the consent of their owners. The "Cantonal Committee for Animal Experiments" approved the collection of blood samples (Canton of Bern; permit 75/16).

#### *2.2. Animal Selection*

This study included four PED affected Shetland Sheepdogs, one from Germany and three related cases from the Netherlands (Figure 1). In the Dutch family, one of the cases is the mother of the other two cases, who are full siblings. A half sibling to the two cases with similar clinical signs died before blood samples could be drawn. For the genetic analyses, we used 117 additional blood samples of Shetland Sheepdogs without any reports of neurological disease and 515 dogs of various other breeds, which had been donated to the Vetsuisse Biobank. Additional details on the samples are given in Tables S1 and S2.

**Figure 1.** Pedigrees of the four affected Shetland Sheepdogs. Squares represent males and circles represent females. Filled symbols indicate affected dogs and open symbols indicate non-affected dogs. Symbols with question marks indicate dogs of an unknown phenotype. The strike-through symbol represents a dog that died before the beginning of the investigation. According to the owner, this dog had similar clinical signs as the other affected Shetland Sheepdogs. The genotypes at the *PCK2*:c.1658G>A variant are given for dogs, from which samples were available (see Section 3.5).

#### *2.3. Clinical Examinations*

Clinical examinations were performed at the University of Veterinary Medicine Hannover, Foundation, Utrecht University, Department of Clinical Sciences of Companion Animals, AniCura Veterinary Clinic Trier GbR, and Veterinary Clinic Neandertal GbR. All examinations were performed after written informed owner´s consent according to the ethical guidelines of the University of Veterinary Medicine Hannover, Foundation and Utrecht University. A resident or diplomate of the European College of Veterinary Neurology performed the examinations in every dog.

Stress tests were performed to aggravate clinical signs by playing with the dogs for up to 30 min or by applying different external stressful stimuli (*n* = 3). Heart rate, rectal body temperature, blood glucose, lactate, creatinine kinase (CK) and electrolytes were measured before and after playing.

#### *2.4. Laboratory Examinations*

Blood examinations were performed immediately after blood sampling and included blood cell count (ADVIA 120 Hematology System, Siemens Healthcare GmbH, Erlangen, Germany), biochemistry (Cobas c 311 analyzer, Roche Deutschland Holding GmbH, Mannheim, Germany) and electrolytes (RAPIDLab 1260, Siemens Healthcare GmbH). Plasma was examined for total thyroxine content via electro-chemiluminescence immunoassay (ECLI, *n* = 3). Serum for insulin measurements was frozen at −4 ◦C immediately after sampling and was examined within 24 h via chemiluminescent Immunoassay (CLIA, *n* = 1, Biocontrol, Ingelheim am Rhein, Germany). Urine samples were taken via cystocentesis and immediately frozen at −20 ◦C until examination. Analysis of urinary organic acids (*n* = 3) was performed by gas chromatography and mass spectroscopy (Biocontrol, and Biochemical Genetics Laboratory, San Diego, CA, USA). For detailed case information, see Table S3.

#### *2.5. Cardiac Examinations*

Cardiac sonographywas performedin an awake statein all 4 cases [24]. Ambulatory electrocardiography (ECG) was performed for 24 hours in case 4 with a bipolar triaxial lead system via telemetric ECG on a holter (Televet 100 Version 4.2, Engel Engineering Service GmbH, Heusenstamm, Germany) [25].

#### *2.6. Muscle Examinations and Histopathology*

Electrodiagnostic examinations of the axial and abaxial muscles and peripheral nerves in all 4 cases were performed using a Vicking Quest electrodiagnostic device (Nicolet Viking Quest IV, Nicolet EBE GmbH, Kleinostheim Germany). Recordings of the compound muscle action potentials (CMAP) and measurement of the motor nerve conduction velocity (mNCV) and the amplitude of the CMAP as well as repetitive nerve stimulation were performed with 0.5, 2, 3, 10 and 30 Hz of the radial and peroneal nerves. Muscle and nerve biopsies were taken from the extensor carpi radialis and tibialis cranialis muscles of case 4 under general anesthesia according to Platt and Olby [26] and were sent to the Clinical and Comparative Neuropathology Laboratory of the Ludwig-Maximilians-Universität, Munich. Samples underwent routine cryohistological processing, including enzyme histochemistry for cytochrome oxidase and nicotinamide adenine dinucleotide tetrazolium reductase, myofiber typing and special stains for the detection of polysaccharides (periodic acid Schiff) and lipids (oil red O), mitochondria and protein aggregates (Engel´s modified Gomori stain). Further samples of both muscles were subjected to transmission electron microscopy following glutaraldehyde fixation, embedding in epoxy resin, ultrasectioning and contrasting with lead citrate and uranyl acetate.

#### *2.7. Additional Diagnostic Examinations*

Electroencephalography (EEG, NicoletOne nEEG, Nicolet) of case 4 was obtained in an awake state with a montage according to Brauer et al. [27].

Low field magnetic resonance imaging (MRI) of the brain of case 4 was performed by the referring veterinarian. T2weighted (T2w), T1weighted (T1w) and fluid attenuation inversion recovery (FLAIR) sequences were available for review.

Cerebrospinal fluid (CSF) was sampled from case 4 in general anesthesia from the cisterna magna and was immediately examined for protein (Cobas c 311 analyzer) and cell content [26].

#### *2.8. Autoantibodies*

Serum and CSF of case 4 were screened for known and novel nervous system autoantibodies with cell-based assays (GAD65, NMDAR, GABABR, AMPARI, AMPAR2, DPPX, LGl1, CASPR2, GlyR, mGlu5) and immunofluorescence test (IFT) on mice hippocampi (Epilepsiezentrum Bethel, Bielefeld, Germany).

#### *2.9. Fibroblast Culture*

Dermal fibroblasts from case 4 were cultured to further examine mitochondrial function according to [28]. A skin biopsy as a starting material was taken from case 4 under local anesthesia (Lidocain 2 mL subcutaneously with 2 cm of spatial distance to the biopsy site; Lidocard 2% Mini-Plasco; B. Braun Melsungen AG, Melsungen, Germany).

#### *2.10. Tryptophan Content of Therapeutic Diet*

Tryptophan contents of a conventional gluten-containing diet (Markus Mühle®, Langenhahn, Germany), a conventional gluten-free diet (Wildkind®, Das Futterhaus-Franchise GmbH & Co. KG, Elmshorn, Germany) and a gluten- and grain-free, seafood-based diet (Purizon Fisch®, Matina GmbH, München, Germany) fed to case 4 were measured (Routine Laboratory, University of Rostock).

#### *2.11. Whole Genome Sequencing of Two A*ff*ected Shetland Sheepdogs*

Genomic DNA was isolated from the EDTA blood of affected dogs and healthy controls with the Maxwell RSC Whole Blood Kit using a Maxwell RSC instrument (Promega, Dübendorf, Switzerland). Illumina TruSeq PCR-free DNA libraries with 350 bp insert size of one affected Shetland Sheepdog from each of the two families were prepared (cases 2 and 4). We collected 331 and 321 million 2 × 150 bp paired-end reads on a NovaSeq 6000 instrument (37.4× and 38.0× coverage). Mapping and alignment were performed as described [29]. The sequence data were deposited under the study accession

PRJEB16012 and the sample accessions SAMEA104091573 and SAMEA4867921 at the European Nucleotide Archive.

#### *2.12. Variant Calling*

Variant calling was performed as described [29]. To predict the functional effects of the called variants, the SnpEFF [30] software together with NCBI annotation release 105 for CanFam 3.1 was used. For variant filtering, we used 654 control genomes, which were publicly available. The control genomes were derived from 648 dogs of genetically diverse breeds and 8 wolves (Table S2).

#### *2.13. Gene Analysis*

We used the dog CanFam 3.1 reference genome assembly for all analyses. Numbering within the canine *PCK2* gene corresponds to the NCBI RefSeq accessions XM\_537379.6 (mRNA) and XP\_537379.2 (protein).

#### *2.14. Sanger Sequencing*

To genotype the *PCK2*:c.1658G>A variant, a 468 bp PCR product was amplified from genomic DNA using the AmpliTaqGold360Mastermix (Thermo Fisher Scientific, Waltham, MA, USA) together with primers 5 -GCT ACA ACT TTG GGC GCT AC-3 (Primer F) and 5 - ATG AGG GGT AGG AAG GGA TG-3 (Primer R). After treatment with exonuclease I and alkaline phosphatase, amplicons were sequenced on an ABI 3730 DNA Analyzer (Thermo Fisher Scientific, Waltham, MA, USA). Sanger sequences were analyzed using the Sequencher 5.1 software (GeneCodes, Ann Arbor, MI, USA).

#### **3. Results**

#### *3.1. Clinical Examinations and Family History*

Four female Shetland Sheepdogs (age 2–6 years) were presented due to progressive dyskinetic episodes. Three of the four dogs were closely related, suggesting an inherited disorder. Although not conclusive, the pedigrees were compatible with an autosomal dominant mode of inheritance (Figure 1).

The episodes were characterized by generalized ataxia with hypermetria and muscular hypertonia of all limbs, dystonia, normal to mildly reduced mentation, and a mild tremor. In the more severe episodes, the dogs were no longer ambulatory. No signs of autonomic dysfunction were visible (normal size of pupils, no salivation, no defecation or urination, no signs of increased gastro-intestinal motility) (Video S1). The episodes varied from minutes to hours and could start while at rest, or during activity. In case 4, they were triggered by excitement or stress, like playing or after being startled by noise, according to the owner. Episodes in this dog were more intense and of longer duration after physical exercise. In cases 1–3, hot weather seemed to aggravate the clinical signs.

General and neurological examination was normal in all dogs, except for mild generalized muscle atrophy in case 4. Stress tests led to intermittent stiff gait in cases 1 and 2 but could not provoke a dystonic episode in case 4. The findings were consistent with a movement disorder or paroxysmal dyskinesia. Encephalopathy or neuromuscular disorders were also considered. All clinical examination results are summarized in Table S3.

#### *3.2. Laboratory Examinations*

Interictal blood cell count, biochemistry, and electrolytes were mostly within the reference range in all four affected dogs. Mildly increased creatine kinase activity (CK) was seen in cases 2 and 3 (*n* = 2/4, 221 U/l and 350 U/l, respectively, reference: <220 U/l). Total thyroxine (tT4) was normal in all tested dogs (*n* = 3/3, Table S3).

In case 4, fasted blood glucose was at the lower boundary of the reference range (72 mg/dL, reference: 70–110 mg/dL) and lactate at the upper reference limit (21.7 mg/dL, reference: 4.5–22.5 mg/dL). In venous blood, the base excess (BE) was −8.5 mmol/L (reference: −4 to 4 mmol/L) with a HCO3 −

of 13 mmol/L (reference: 20–30 mmol/L), decreased pCO2 (19.7 mm Hg, reference: 35–55 mm Hg) with normal pH (7.44, reference: 7.3–7.45). These findings are consistent with a metabolic acidosis with respiratory compensation. Blood examination after the stress test showed increased CK values in cases 1 and 2 (*n* = 2/3), and mildly decreased blood glucose (69 mg/dL) and the deterioration of BE (−10 mmol/L) in case 4 (*n* = 1/3). The parallel insulin measurement in case 4 was normal (6.8 μU/mL, reference: 5–25 μU/mL; glucose-insulin-ratio 10.14, reference: <30).

Urinary organic acid analysis an showed increased excretion of lactate (*n* =3/3) and 2-hydroxybutyrate (*n* = 1/3) or 3-hydroxybutyrate (*n* = 2/3) compared to normal control (Figure 2). A full overview of all clinical and diagnostic findings in the four affected dogs is given in Table S3.

**Figure 2.** Metabolic screening of urinary organic acids in case 4. Gas chromatographic examination of a 3-year-old paroxysmal exercise-induced dyskinesia (PED)-affected dog's urine showed an increased excretion of lactate (red circle) and 2-hydroxybutyrate (blue circle) compared to a healthy control dog.

#### *3.3. Additonal Examinations*

Cardiac sonography and ECG were normal between episodes. ECG immediately before an episode revealed normal sinus rhythm with mild tachycardia (120–180 bpm). When a dyskinetic episode started, ECG was overlapped by muscle artefacts. ECG did not reveal cardiac pathology (Figure S1).

EEG in case 4 while the dog was awake showed mostly muscle artefacts and otherwise predominantly normal low voltage beta-rhythm.

Low-field MRI of the brain was performed in case 4 by the referring veterinarian and was available for review. No pathological abnormality was visible.

A CSF tap was performed in case 4 and was unremarkable: no cells were present, protein content was 11.8 mg/dL (reference <25 mg/dL), and glucose was 69 mg/dL (reference 42–77 mg/dL).

Electromyography (EMG) showed normal insertional discharge in all examined muscles without pathological spontaneous activity. Nerve-conduction studies showed normal nerve conduction velocities with normal CMAP amplitudes in all four dogs.

A search for autoantibodies in case 4 did not reveal any pathological findings.

#### *3.4. Histopathology*

Histopathological muscle changes in case 4 were sparse and featured a diffuse type II fiber predominant muscle atrophy. There was no evidence of mitochondrial changes and/or substrate accumulation on histology, enzyme histochemistry, or electron microscopy.

#### *3.5. Clinical Management*

Treatment with oral phenobarbital up to 3 mg/kg twice daily (cases 1–4) and levetiracetam 60 mg/kg three times daily (case 4) did not decrease the frequency of episodes according to the owners. Diazepam 2 mg/kg rectally did not seem to decrease the length of the episodes (case 4). Acetazolamid 50 mg/kg three times daily initially was reported to improve the clinical signs in case 4, but only for 6 months when the frequency started to increase again. Changing to zonisamide 10 mg/kg three times daily seemed to have an impact on episode frequency in the long term.

Supplementation with L-carnitine and multivitamins did not improve clinical signs in any of the four affected dogs.

A number of various commercial diets and gluten-free diets were tried but appeared to be unsuccessful in three dogs (case 1–3). However, in case 4, the feeding of a gluten-free, home-made fresh meat diet improved clinical signs. In case 4, the frequency of episodes improved with a commercially available gluten-free diet. Interruption of this diet led to the recurrence of an increased frequency of the episodes. A further change in diet to a seafood-based, gluten- and grain-free diet markedly improved the clinical signs again, even better than before. The last diet had the highest tryptophan content (Table 1).



Additional supplementation with tryptophan 50 mg/kg twice daily in case 4 combined with the prevention of stress and exhausting exercise further decreased clinical signs according to the owner. A summary of treatments in all four cases is given in Table S3.

#### *3.6. Outcome*

We were able to monitor all four affected dogs over a period of 5 to 12 years and the dogs remained stable (Table S3). Two dogs were event-free, one dog suffered from only one to two events per year. Exercise- and stress-management in combination with zonisamide, a seafood-based, gluten-free diet and tryptophan helped the fourth dog to stabilize on a low-level frequency of episodes. The dog continued to display one short episode every 2–3 days, during which she was still ambulatory (Figure 3). Six years after diagnosis, case 3 died of an acute renal failure, case 1 died of old age at 15 years of age. Cases 2 and 4 were still alive at the time of writing (Table S3).

**Figure 3.** Frequency of dyskinetic episodes in case 4. The estimated frequency of episodes per month (black line), contemporaneous medications and diets are displayed with green and blue bars, respectively. A combination of seafood-based, gluten-free diet with supplementation of tryptophan and treatment with acetazolamide or zonisamide seemed to reduce frequency of episodes.

#### *3.7. Genetic Analysis*

We sequenced the genomes of cases 2 and 4 at 37.4× and 38.0× coverage and called SNVs and short indels with respect to the CanFam 3.1 reference genome assembly. We then searched for shared heterozygous and homozygous variants in the genome sequence of the two affected dogs that were not present in 654 control genomes. This analysis yielded 1066 variants that were exclusively shared by the two cases and not present in any of the control genomes (Table S4). None of the shared homozygous variants were predicted to have a protein-changing effect, but 10 heterozygous variants shared by the two cases were predicted to be protein changing (Table 2).



We then prioritized the 10 private protein-changing variants according to the functional knowledge on the altered genes. We considered the *PCK2* gene encoding the mitochondrial phosphoenolpyruvate carboxykinase 2 the most likely candidate gene for the observed clinical phenotype (Table 3).


**Table 3.** Heterozygous protein-changing variants shared by the 2 PED cases and absent in 654 controls.

The top candidate variant on the genomic level was Chr8:4,107,413G>A. The corresponding variant designations on the cDNA and protein level are XM\_537379.6:c.1658G>A or XP\_537379.2:p.(Arg553Gln), respectively. We confirmed the variant by Sanger sequencing. (Figure 4A). The variant is predicted to alter a positively charged arginine close to the GTP-binding site of PCK2 into a neutral glutamine without major changes to the threedimensional structure of the enzyme. The wildtype arginine is strictly conserved across animals from *C. elegans* up to mammals (Figure 4B).

**Figure 4.** Details of the PCK2:c.1658G>A variant. (**A**) Representative Sanger sequencing electropherograms of two dogs with the different genotypes are shown. (**B**) Evolutionary conservation of the arginine residue at position 553 of the PCK2 protein. A multiple species alignment illustrates that this residue is strictly conserved across animals from worms to mammals. Amino acids 548-551 form part of the GTP-binding site of PCK2 and are underlined [31].

We then genotyped all four affected dogs, 117 Shetland Sheepdog controls without any movement disorder or seizures and 515 control dogs from 71 genetically diverse dog breeds for the PCK2:c.1658G>A variant and found a perfect genotype-phenotype association. All four cases carried one copy of the mutant allele, while all control dogs were homozygous for the reference allele (Table 4; Table S1). The segregation of the genotypes in the available family was compatible with an autosomal dominant inheritance (Figure 1). An attempt to establish a fibroblast culture from case 4 to perform analyses on mitochondrial function was unsuccessful. While fibroblasts from a healthy control dog grew as expected, fibroblasts from case 4 did not grow appropriately to allow biochemical studies.

**Table 4.** Association of the genotypes at *PCK2*:c.1658G>A with paroxysmal dyskinesia.


<sup>1</sup> Independent from the 654 controls of the variant discovery.

#### **4. Discussion**

The present article describes the clinical and diagnostic findings of an inherited PED in four Shetland Sheepdogs with a suspected deficiency in the mitochondrial phosphoenolpyruvate carboxykinase 2 (PCK2). There are two isoforms of phosphoenolpyruvate carboxykinase, a cytosolic isoform, encoded by the *PCK1* gene, and a mitochondrial isoform, encoded by the *PCK2* gene. The tissue specificity of these isoforms has been described [32]. PCK1 is hormonally regulated, with insulin switching the enzyme off, whereas PCK2 does not seem to be regulated by hormones, but rather by mitochondrial GTP levels [32–35].

Our genetic analysis revealed a *PCK2* missense variant in the affected dogs, which was predicted to change an evolutionarily conserved amino acid located close to the GTP-binding site of PCK2 [31]. The mutant allele was exclusively found in a heterozygous state in the four studied PED cases and absent in more than one thousand control dogs. As PCK2 is a monomeric enzyme, we consider it unlikely that the p.Arg553Gln substitution will have a dominant negative effect. We speculate that haploinsufficiency during periods of high energy demand may be causing the phenotype in the affected dogs. We have to caution that our genetic analysis was strictly based on the assumption of a shared causative genetic variant between the two sequenced cases. If the phenotypes in the studied dogs are due to different genetic and/or environmental causes, the detected PCK2 variant might be functionally neutral. We also have to caution that our bioinformatics analysis considered

only small genetic variants and would not have detected any large structural variants involving more than ~25 consecutive nucleotides.

We did not succeed in proving the functional relevance of the genetic variant by enzyme activity assay as fibroblasts from a PED-affected dog did not proliferate appropriately, whereas a parallel culture of control fibroblasts from a healthy dog grew sufficiently. We have observed this phenomenon before in some human diseased fibroblasts (personal observation, A.M.D.).

Few human patients with primary phosphoenolpyruvate carboxykinase deficiency have been reported in the literature, mainly with isolated variants in PCK1 [36–38]. These patients suffer from liver dysfunction, sometimes leading to liver failure, hypoglycemia, lactic acidosis and sometimes complex symptoms [36–38]. To the best of our knowledge, no PCK2 variants have been described in human patients. Some reports in the older literature claim a deficiency of mitochondrial phosphoenolpyruvate carboxykinase in human patients [39,40]; however, later on, these diagnoses were withdrawn or revised. One publication reports a patient with a complex phenotype suffering from both cytosolic and mitochondrial phosphoenolpyruvate carboxykinase deficiency [41].

The current human genome and exome data of the gnomAD browser do not indicate any intolerance of heterozygous *PCK2* loss of function variants [42,43]. Interestingly, the variant found in the affected Shetland Sheepdogs, p.Arg553Gln, also represents a rare variant in humans. The gnomAD data lists 27 heterozygotes for the 553Gln-allele, which has a frequency of 9.55 <sup>×</sup> <sup>10</sup>−<sup>5</sup> in the dataset [43]. It is unknown whether these persons have any clinical phenotype.

In the four affected dogs, paroxysmal hypertonic dyskinetic episodes often being triggered by stress, excitement or hot weather were the prominent clinical signs. Transient hypoglycemia, lactaturia, ketonuria and subsequent metabolic acidosis were noted. A muscle biopsy showed mild type II fiber predominant muscle atrophy, which is common in metabolic disease [44]. No cardiac abnormalities, structural changes of brain parenchyma, or signs of hepatopathy were seen in our canine patients. Similar clinical signs were described in a Shetland Sheepdog in 1992, but at this time the etiopathology remained obscure [45].

We suggest that decreased PCK2 activity may have led to impaired gluconeogenesis and energy metabolism in the affected Shetland Sheepdogs. The most severely affected dog (case 4) showed borderline low glucose. We suggest that clinical signs are therefore most pronounced in times of stress or exercise, when the body has an increased demand for energy. The resulting shortage of energy might first affect those organs with a high energetic turnover, such as muscles and the brain. This might result in the paroxysmal dyskinetic events seen in the Shetland Sheepdogs similar to glucose transporter GLUT1 deficiency where gait abnormalities are observed [46]. Phosphoenolpyruvate carboxykinase is expressed in astrocytes [47] and impaired gluconeogenesis may have a similar effect. GTP levels in neuronal cells may be altered, which may affect the production of tetrahydrobiopterin [48] and thus synthesis of neurotransmitters and NO.

As not all dogs had measurable episodes of hypoglycemia, another albeit highly speculative pathomechanism should be considered: PCK2 acts as a sensor of the citric acid cycle ('Krebs cycle') flux by removing oxaloacetate [34]. A reduced PCK2 activity could result in the accumulation of oxaloacetate hampering the citric acid cycle flux. PCK2 is the only isoform providing phosphoenolpyruvate carboxykinase activity in pancreas and possibly other tissues, linking the production of mitochondrial GTP to anaplerotic phosphoenolpyruvate cycling [35]. PCK2 also has pyruvate kinase activity, which theoretically could enhance mitochondrial pyruvate formation and transformation into oxaloacetate when energy levels in the mitochondrion are low, thus contributing to the citric acid cycle, a situation suggested to be stress-related [49]. We have to caution that our two mechanistic hypotheses so far are not supported by any experimental data. As heterozygous PCK2 loss-of-function variants have not yet been identified as causative for a corresponding phenotype in human patients, functional validation will be of the utmost importance to further evaluate the hypothetical link of canine PCK2 deficiency and PED.

In human medicine, a ketogenic diet with a low glycemic index is recommended for diseases that affect energy metabolism to avoid high insulin peaks [50,51]. In humans, ketogenic diets classically include high amounts of fat and low amounts of carbohydrates, which forces the body to change to a ketogenic metabolism, to provide ketone bodies as an alternative energy source for the brain and muscles [52,53]. In contrast, dogs do not tend to change into a ketogenic metabolism as easily as humans [54]. Additionally, a diet with a high amount of fat may cause pancreatitis in dogs [55]. Therefore, it is common practice in veterinary medicine to use a gluten- and grain-free diet as a more compatible form of a diet with a low glycemic index. This was also successfully applied in the presented Shetland Sheepdogs, whose clinical signs improved after being fed a gluten-free, high-protein diet. In one dog, a seafood-based diet showed the best results. This diet was rich in tryptophan, which is an essential amino acid involved in the production of serotonin [56]. Increased serotonin levels in the brain reduce stress [57]—an important trigger of dyskinetic episodes observed in this dog. Tryptophan may enhance de novo synthesis of NAD [58], an important metabolic regulator for energy metabolism. However, these mechanisms are highly speculative as we did not measure tryptophan nor neurotransmitter levels. Additionally, in one dog, antiepileptic drugs of the class of carbonic anhydrase inhibitors (acetazolamide or zonisamide) were used and correlated with improved clinical signs. Successful treatment with acetazolamide has been described in Soft-coated Wheaten Terriers and Golden Retrievers with paroxysmal dyskinesia [59,60]. The exact mechanism of action is not clear yet, but it is thought that acetazolamide supports ion transport across the blood–brain barrier, modifying the intracellular pH and, therefore, the transmembrane potential, which lowers the excitability of neurons [51].

#### **5. Conclusions**

We describe a new PED with presumed autosomal dominant inheritance in Shetland Sheepdogs. Stress management, a specific diet and pharmacological therapy resulted in the partial or complete suppression of hypertonic dyskinetic episodes and enabled a good quality of life. The genetic analysis suggested that the *PCK2*:p.Arg553Gln missense variant should be considered and further evaluated as potential candidate causal variant for this phenotype. This study provides an interesting potential link between exercise-induced hypoglycemia, mitochondrial energy metabolism and paroxysmal dyskinesia that warrants further investigation.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/11/7/774/s1, Figure S1: Long-term electrocardiography (ECG) of case 4 before a dyskinetic episode and at the beginning of an episode, Table S1: PCK2:c.1658G>A genotypes of 636 dogs from 69 different dog breeds; Table S2: Accession numbers of 648 dog and 8 wolf genome sequences; Table S3: Summary of clinical and diagnostic findings in the 4 affected dogs. Table S4: List of heterozygous private variants that were shared in the two sequenced cases and absent from 654 control genomes. Video S1: 3 year old PED-affected female Shetland Sheepdog experiencing a hypertonic dyskinetic episode triggered by excitement while playing at the beach.

**Author Contributions:** Conceptualization, J.N. and T.L.; investigation, J.N., P.H., P.J.J.M., V.J., K.M., M.R., M.K., P.W., A.F., A.T.; resources, J.N., P.J.J.M., P.A.J.L.; data curation, V.J.; writing—original draft preparation, J.N., P.H., A.M.D., M.R., K.M., M.K., T.L.; writing—review and editing, J.N., P.H., P.J.J.M., P.A.J.L., V.J., A.M.D., M.R., K.M., A.C.S., M.K., M.H., P.W., A.F., A.T., T.L.; supervision, T.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** We would like to acknowledge the owners of the dogs for donating samples and data including precise recording of clinical signs and video recording. We thank Nathalie Besuchet Schmutz, Catia Coito, Marion Ernst and Daniela Steiner for expert technical assistance, the Next Generation Sequencing Platform of the University of Bern for performing the high-throughput sequencing experiments, and the Interfaculty Bioinformatics Unit of the University of Bern for providing high performance computing infrastructure. We thank the Dog Biomedical Variant Database Consortium (Gus Aguirre, Catherine André, Danika Bannasch, Doreen Becker, Brian Davis, Cord Drögemüller, Kari Ekenstedt, Kiterie Faller, Oliver Forman, Steve Friedenberg, Eva Furrow, Urs Giger, Christophe Hitte, Marjo Hytönen, Vidhya Jagannathan, Tosso Leeb, Frode Lingaas, Hannes Lohi, Cathryn Mellersh, Jim Mickelson, Leonardo Murgiano, Anita Oberbauer, Sheila Schmutz, Jeffrey Schoenebeck, Kim Summers, Frank van Steenbeek, Claire Wade) for sharing whole genome sequencing data from control dogs. We also acknowledge all canine researchers who deposited dog whole genome sequencing data into public databases.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **A Missense Variant in** *ALDH5A1* **Associated with Canine Succinic Semialdehyde Dehydrogenase Deficiency (SSADHD) in the Saluki Dog**

**Karen M. Vernau 1,\*, Eduard Struys 2, Anna Letko 3, Kevin D. Woolard 4, Miriam Aguilar 5, Emily A. Brown 5, Derek D. Cissell 1, Peter J. Dickinson 1, G. Diane Shelton 6, Michael R. Broome 7, K. Michael Gibson 8, Phillip L. Pearl 9, Florian König 10, Thomas J. Van Winkle 11, Dennis O'Brien 12, B. Roos 2, Kaspar Matiasek 13, Vidhya Jagannathan 3, Cord Drögemüller 3, Tamer A. Mansour 5,14, C. Titus Brown <sup>5</sup> and Danika L. Bannasch 5,\***


Received: 7 August 2020; Accepted: 27 August 2020; Published: 2 September 2020

**Abstract:** Dogs provide highly valuable models of human disease due to the similarity in phenotype presentation and the ease of genetic analysis. Seven Saluki puppies were investigated for neurological abnormalities including seizures and altered behavior. Magnetic resonance imaging showed a diffuse, marked reduction in cerebral cortical thickness, and symmetrical T2 hyperintensity in specific brain regions. Cerebral cortical atrophy with vacuolation (status spongiosus) was noted on necropsy. Genome-wide association study of 7 affected and 28 normal Salukis revealed a genomewide significantly associated region on CFA 35. Whole-genome sequencing of three confirmed cases from three different litters revealed a homozygous missense variant within the aldehyde dehydrogenase 5 family member A1 (*ALDH5A1*) gene (XM\_014110599.2: c.866G>A; XP\_013966074.2: p.(Gly288Asp). *ALDH5A1* encodes a succinic semialdehyde dehydrogenase (SSADH) enzyme critical in the gamma-aminobutyric acid neurotransmitter (GABA) metabolic pathway. Metabolic screening of affected dogs showed markedly elevated gamma-hydroxybutyric acid in serum, cerebrospinal fluid

(CSF) and brain, and elevated succinate semialdehyde in urine, CSF and brain. SSADH activity in the brain of affected dogs was low. Affected Saluki dogs had striking similarities to SSADH deficiency in humans although hydroxybutyric aciduria was absent in affected dogs. *ALDH5A1*-related SSADH deficiency in Salukis provides a unique translational large animal model for the development of novel therapeutic strategies.

**Keywords:** inborn error of metabolism; encephalopathy; SSADHD; *ALDH5A1*; GABA; 4-hydroxybutyric acid; succinic semialdehyde; encephalopathy; whole-genome sequencing; precision medicine; GWAS; inherited

#### **1. Introduction**

Inborn errors of metabolism (IEMs) are a group of diseases caused by an enzymatic deficiency in a metabolic pathway, most commonly caused by a genetic mutation. While individually these diseases are rare, as a group they are relatively common, with more than 500 IEM diseases reported in people [1]; in animals, they are becoming increasingly recognized [2–6]. In diseases caused by an IEM, clinical signs are due to the reduced or lack of production of a biochemical product, or accumulation of an abnormal amount of substrate or substrates produced by alternative metabolic pathways, secondary to the enzymatic deficiency. The diagnosis of an IEM may be a challenging, as clinical signs can be vague and non-specific, and targeted diagnostic testing is required [7]. IEMs are often recognized in young people and animals, and many have neurological manifestations [8,9].

Seizures are a common neurological sign in dogs [10]. Disorders causing seizures arise either extracranially (reactive seizures), or intracranially [11] Epilepsy is a brain disease characterized by a lasting predisposition to generate seizures, which is classified in dogs as structural epilepsy or idiopathic epilepsy (OMIA 000344-9615) [11]. Causes of structural epilepsy include inflammation (e.g., granulomatous meningoencephalitis), neoplasia, nutritional alterations (e.g., thiamine deficiency), infection, anomalous entities (e.g., hydrocephalus), inborn errors of metabolism and trauma [11]. Dogs with idiopathic epilepsy (IE) are typically 6 months to 6 years of age and usually have normal physical and neurological examinations between seizures [12]. Dogs younger than 6 months or older than 6 years of age usually have reactive seizures or structural epilepsy, rather than idiopathic epilepsy [12].

A seizure disorder reported in Salukis is called central nervous system status spongiosus in Saluki dogs (SSSD). There are only brief reports of this disease in the literature [13,14]. One affected 8-month-old male puppy from a litter of 9 was reported with a 5 month history of seizures and behavioral changes. The sire and dam were full siblings. All nine puppies and the sire were euthanized; pathological changes were noted in the affected puppy and in two clinically normal littermates, and the rest of the puppies and the sire were pathologically normal. Pathological changes in the clinically affected puppy included widespread bilaterally symmetrical status spongiosis of the cerebrum, brainstem and cerebellum at the grey–white matter junction, which extended into both the grey and white matter. There were also lesions in the thalamus, optic nerve and internal capsule but no lesions were noted in the spinal cord [13].

Recognized causes of early-onset symmetrical brain lesions include metabolic, nutritional and toxin-induced diseases [15]. In Saluki dogs with SSSD, the clinical signs and lesions on MRI and pathology appear to be breed specific, identical in distribution and type, and diagnosed in multiple dogs over a long period of time (1987 [13] to 2020). Clinical signs developed while puppies were still with the breeders, making toxicity a less likely cause; pathology differed from previously reported nutritional [16] or toxic [17–19] central nervous system problems, and thus a genetic cause was considered most likely. Although a metabolic disorder was not identified by routine diagnostic testing in affected Salukis, an underlying genetic abnormality causing a metabolic problem was most likely based on the age of onset of clinical signs.

The purpose of this study was to define the phenotype of Salukis with SSSD and to determine the underlying genetic cause in this breed. Comprehensive evaluations including MRI and necropsy, as well as metabolic and enzyme activity testing, were performed on urine, serum, cerebrospinal fluid and brain tissue from four affected puppies from two litters from the USA and a litter with three affected puppies from Germany. All seven affected dogs were used for a genome-wide association study (GWAS) followed by whole-genome sequence analysis of three affected puppies, which identified a private homozygous missense variant in the canine *ALDH5A1* gene.

#### **2. Materials and Methods**

#### *2.1. A*ff*ected Dogs*

From 2005 to 2015, seven Saluki dogs affected with SSSD from the USA (4) and Germany (3) had DNA collected. Four dogs were examined—three dogs at the William R. Pritchard Veterinary Medical Teaching Hospital at the University of California Davis (UCD) and one dog was evaluated at Fachtierarzt fur Kleintiere, in Wiesbaden, Germany. All 4 dogs were presented by their breeders for examination. A fifth Saluki dog affected with SSSD from the USA had a necropsy completed at UCD Two additional German dogs were not evaluated clinically beyond the breeder's description of the clinical signs.

#### *2.2. Control Dogs*

#### 2.2.1. MRI Evaluation

Four unaffected Saluki dogs related to the affected USA Saluki dogs were examined and had magnetic resonance imaging of their brain and completed at Advanced Veterinary Medicine Imaging in Los Angeles, California.

#### 2.2.2. Targeted Metabolic Testing

Archived urine = 4, serum = 4, cerebrospinal fluid (3) and brain tissue (4) from 15 different non-Saluki dogs unaffected by SSSD were utilized as control samples.

#### *2.3. A*ff*ected Saluki Dogs*

Blood work (complete blood count, and serum biochemical profile) was performed by the referring veterinarians in two dogs (dogs 3 and 4). Cerebrospinal fluid sample (CSF) was collected in two dogs (dogs 1 and 5). One CSF sample was routinely analyzed in Germany and the other sample was collected at UCD and frozen at −80 degrees for further analysis. Two dogs had quantitative urine organic acid testing completed at the University of California San Diego Biochemical Genetics Laboratory (dogs 1 and 2). Urine was shipped to the lab by the breeder and was analyzed by the lab 18 days later. Four dogs had complete necropsies completed at UCD, and one had a necropsy at Ludwig Maximilians Universität München in Germany. The owners consented to the necropsy and processing of postmortem samples. Following necropsy, the brain was immediately immersed in 10% neutral buffered formalin followed by standard paraffin embedding. Selected regions were sectioned at 5 μm slice thickness and stained with hematoxylin-eosin and luxol fast blue-cresyl violet.

#### MRI and Histopathology

Six Saluki dogs underwent MRI of the brain—two affected Saluki dogs underwent magnetic resonance imaging (MRI) of the brain at UC Davis, and four unaffected Saluki dogs had imaging of the brain at Advanced Veterinary Medicine Imaging in Los Angeles, California. Both locations used a 1.5 T MRI system (GE Signa, GE Healthcare, Waukesha, WI, USA), with paired 5" general purpose radiofrequency coils. Sagittal T1-weighted (T1W) and T2-weighted (T2W) images, transverse T1W, T2W, fluid attenuating inversion recovery (FLAIR), and T2\*-weighted (T2\*W) images, and dorsal T2W images were acquired of the brain. Sagittal, transverse, and dorsal T1W images were repeated after intravenous administration of 0.1 mmol/kg gadopentetate dimeglumine (Magnevist, Bayer, Whippany, NJ, USA).

#### *2.4. Sample Collection and DNA Extraction*

Blood samples, pedigree, and phenotype information were collected from 7 affected dogs and 18 close relatives of affected dogs including 4 parents (Figure 1). Additional samples were collected from 48 healthy Saluki dogs. Healthy dogs of other breeds (*n* = 228) were used that were part of a DNA repository at UC Davis. DNA was extracted from EDTA whole blood samples using Gentra Puregene DNA purification extraction kit (Qiagen, Valencia, CA, USA). Collection of canine samples was approved by the University of California, Davis Animal Care and Use Committee (protocol #18561) and the Cantonal Committee for Animal Experiments (Canton of Bern; permit 75/16).

**Figure 1.** Pedigree of seven succinic semialdehyde dehydrogenase deficiency (SSADHD)-affected Saluki dogs. Females are depicted by circles and males by squares. Black fill indicates affected puppies. Numbers indicate 25 dogs from which samples were available. The blue contour indicates animals that were genotyped on SNP array, and the red contour the three affected dogs selected for the WGS. Note that the two litters on the left were seen in the USA and the third litter on the right in Germany. A common male ancestor illustrates the genealogical relatedness.

#### *2.5. Genome-Wide Association Scan*

SNP genotypingwas performed using the Illumina Canine HD 174,000 SNP array (Illumina, San Diego, CA, USA) for 7 affected cases and 28 neurologically normal adult Saluki controls. Genome-wide association analysis was performed using Plink [20]. SNPs were pruned from analysis if the minor allele frequency was <5% and the call rate <90%. Chi-square association analysis, Bonferroni adjustments, and genomic inflation calculations were performed in Plink. Figure 5 was made in R using ggplot2 [21,22].

#### *2.6. Whole-Genome Sequence*

Whole-genome sequencing (WGS) was performed on the two affected Salukis from the USA and compared to 98 controls dogs from various breeds as reported, and coverage was 6.4× for 1052 and 5.3× for 5813 SRA: SRR5311685 and SRR5311664 (study: PRJNA377155) [23]. Segregation of variants was performed using 2 cases compared to 98 controls within the homozygous interval identified by visual inspection of genotype calls from the array data. Variant Effect Predictor (VEP), as employed in Ensembl using Refseq annotation and additional EST/CCDs, was used to predict the effect of segregating variants [24]. Polyphen2 [25] and SIFT [26] were used to evaluate the severity of the missense variants.

In one German Saluki (dog 5), whole-genome sequencing using genomic DNA isolated from the blood sample of the affected dog was performed as described previously [27]. Data corresponding to approximately 15× coverage of the genome was collected on an Illumina HiSeq2000 instrument (2 × 100 bp). Read mapping and variant calling were carried out as previously described [28], with respect to the CanFam3.1 genome reference assembly and the NCBI annotation release 105. Variant filtering was performed against 581 dog and 8 wolf genomes which were publicly available [28]. WGS data of the affected dog was made available under study accession PRJEB16012 at the European Nucleotide Archive (www.ebi.ac.uk/ena; sample accession SAMEA4504825).

Annotationswithin the canine*ALDH5A1* gene refer to the NCBImRNA accession no. XM\_014110599.2 and the protein accession no. XP\_013966074.2. Annotations within the *GPLD1* gene refer to the NCBI mRNA accession no. XM\_005640079.3 and the protein accession number XP\_005640136.1. Annotations for the putative *PTCHD3* gene refer to the mRNA accession no. ENSCAFT00000039442.3 and the protein accession no. ENSCAFP00000035301.

#### *2.7. Genotyping*

The variant in *ALDH5A1* disrupted a Sau96I restriction enzyme site, allowing rapid genotyping by PCR-RFLP analysis. PCR primers were designed using primer 3 [29] to amplify an 872 bp product, which upon digestion with Sau96I produced 700 and 150 bp fragments for the variant allele and 550 and 150 bp fragments for the wild-type allele. All PCR was carried out using Qiagen HotStart DNA polymerase kit (Qiagen, Valencia, CA, USA) at an annealing temperature of 58 degrees using the following PCR primers: F\_TCCCGAGTTAGGGGTTCTTT, R\_TCACGTTTTCCTGATTTCACC. The same primers were used to verify the mutation by Sanger sequencing on an Applied Biosystems 3500 Genetic Analyzer using the Big Dye Terminator Sequencing Kit (Life Technologies, Burlington, ON, Canada).

#### *2.8. RT-PCR*

RT-PCR was performed for liver cDNA from a case and control. *RPS5* was included as a housekeeping gene control [27]. Primers were designed using Primer3 (SAL\_2F: TTGTATTTGACAGC GCCAAC, SAL\_2R: CAAGGCCAGATTGCTTCAC) except for RPS5, in which the primers were as recommended [30]. Each reaction included 13.9 μL of water, 2 μL of 10× buffer with MgCl2, 1 μL of dNTP, 1 μL of each forward and reverse primers (20 μM), 1 μL of HotStarTaq DNA Polymerase (Qiagen, Valenica, CA, USA), and 1 μL of cDNA made from 1000 ng of RNA. Amplified products were visualized on a 2% agarose gel.

#### *2.9. Targeted Metabolic Testing*

#### *Gamma-hydroxybutyrate (GHB) and succinate semialdehyde (SSA):*

GHB and SSA in fluids and brain tissue were quantified by isotope dilution mass spectrometric methodology, as previously described [31,32].

#### *The 4,5-dihydroxyhexanoic acid (DHHA):*

Analysis of DHHA in fluids and brain was comparable to that for GHB as previously described [31], with some modifications. 2H3-DHHA was used as the internal standard and the samples were extracted a single time with ethylacetate. For quantitation, positive chemical ionization was employed.

#### *Succinic semialdehyde dehydrogenase (SSADH) activity:*

SSADH activity was quantified fluorometrically in brain tissue samples using the NADH/NAD couple and SSA as substrate, as previously described [33].

#### **3. Results**

#### *3.1. A*ff*ected Dogs*

Four dogs from two litters from the USA were closely related, and the third litter from Germany was distantly related to the other two litters. There were four females and three males affected (Figure 1).

#### *3.2. Clinical Phenotype*

The breeders of affected Saluki puppies noted that puppies were first abnormal between six and ten weeks of age. Historical clinical signs included seizures, abnormal behavior such as episodes of vocalization (Video S1, SupplementaryMaterials), and difficulty being aroused from sleep. Four puppies (dogs 1 and 3–5) were evaluated by a board-certified veterinary neurologist (Table 1). No abnormalities were noted on physical examination. On neurological examination, puppies had mild generalized ataxia with thoracic limb hypermetria (two puppies) (Video S2, Supplementary Materials), absent menace reflex in both eyes and delayed proprioceptive positioning in all four limbs, consistent with a multifocal disease process. Two dogs (dogs 3 and 4) had a normal CBC and serum biochemical profile completed by their referring veterinarian, and two dogs had normal quantitative urine organic acid analysis (Biochemical Genetics Laboratory, University of California San Diego, San Diego, CA, USA). One dog (dog 5) had a normal cisternal cerebrospinal fluid analysis. Five dogs were treated for seizures with oral phenobarbital or levetiracetam. Although clinical signs did not progress, all affected dogs were euthanized as puppies at the request of the breeders when they were still in their care. Puppies were euthanized between three and nine months of age for quality of life concerns, primarily due to the recurrent episodes of vocalization. Five dogs (dogs 1–5) had necropsies completed.

#### *3.3. MRI and Histopathology*

Two affected dogs (dogs 1 and 2) and four other related but unaffected dogs had an MRI of the brain. All four unaffected dogs had unremarkable MR images. Both affected dogs exhibited prominent sulci (Figure 2A,C–E) compared to normal dogs (Figure 2F), consistent with diffuse cortical atrophy. Bilateral, symmetrical, T2 and FLAIR hyperintensity was present in the diencephalon, deep cerebellar nuclei (Figure 2A–C), midbrain (Figure 2D), and multiple basal nuclei (Figure 2E). Multifocal, symmetrical T2 and FLAIR hyperintensity was also present in the deep cortical laminae of the grey matter throughout the cerebral cortex (Figure 2A–F).

No hypointense lesions or signal voids were observed associated within the brain parenchyma on T2\*W images.

Histopathologically, there was severe bilaterally symmetric spongiform change, worse within the mesencephalon (Figure 3), brainstem, and deep cerebellar nuclei, but also severe in the thalamic nuclei and deep cortical grey matter. The corpus striatum was less affected but exhibited similar lesions most notably in the entopeduncular nuclei and putamen. Neurons exhibited single to multiple, clear, well-demarcated vacuoles that compressed and displaced the nucleus (Figure 4). There was marked proliferation of enlarged astrocytes associated with the spongiform change, and both neurons and astrocytes appear affected. The grey matter was more severely affected, particularly at the grey–white matter junction. The spinal cord was not affected.



**Figure 2.** MRI abnormalities in SSADHD-affected Saluki Dogs. Transverse T2-weighted (**A**,**C**–**F**) and FLAIR (**B**) MR images at the level of the cerebellum (**A**–**C**), midbrain (**D**,**F**) and corpus striatum (**E**) demonstrating symmetrical involvement of predominantly grey matter structures. Images from dog 1 (**A**,**B**,**D**,**E**), dog 2 (**C**) and an unaffected littermate (**F**). Consistent bilateral symmetrical T2 hyperintensity of the deep cerebellar nuclei ((**A**–**C**); white arrowheads) is the most prominent finding. Similar bilaterally symmetrical hyperintensity is seen involving the tectum and dorsal tegmentum ((**D**); black arrowhead) and endopeduncular (medial) and lentiform nuclei ((**E**); white arrowheads) but not present in unaffected dog images (**F**). Sulci are prominent (**D**) compared to an unaffected age matched control (**F**), consistent with atrophy of cortical grey matter. Hyperintensity of deep cortical grey mater laminae is evident on T2-weighted and FLAIR images at all levels (white arrows) is not present on unaffected dog MR images (**F**).

#### *3.4. Genetic Analysis*

Both sexes are affected and in-depth pedigree analysis revealed the presence of a common male ancestor connecting the American and European families (Figure 1). As all parents of affected offspring show no clinical signs, it could be speculated that the observed disease phenotype follows autosomal monogenic recessive inheritance. Genome-wide association was performed using DNA samples from the seven affected dogs (Figure 1) and 28 phenotypically normal Saluki controls. After quality control, there were 108579 SNPs available for association. A single genome-wide significant association signal based on a pBonferroni (0.006) on CFA 35 (chr35: g23,654,869; praw 5.27 <sup>×</sup> <sup>10</sup><sup>−</sup>8) was identified (Figure 5). Furthermore, a 2.683 Mb region of homozygosity was identified in the seven affected dogs on CFA 35: 21,925,974–24,608,949 bp (CanFam3.1).

In order to identify a causative variant, initially paired-end whole-genome sequences of 2 affected puppies from two American litters (1052, 5813: Figure 1) and 98 unaffected controls from various breeds were investigated in the associated interval. There were 35,982 single-nucleotide variants (SNVs) and 16,832 insertion/deletion (indel) variants identified within the critical interval defined by homozygosity: CFA 35: 21,925,974–24,608,949 bp (CanFam3.1). There were 259 SNVs and 41 indels that segregated with the phenotype in the 100 animals. There were three coding variants identified: a synonymous variant (g.22,506,956G>A) in the *GPLD1* gene, and two protein-changing missense

variants, g.22,572,768G>A in the *ALDH5A1* gene, and g.23,908,560T>C in the putative *PTCHD3* gene. The synonymous variant in *GPLD1* was not investigated further.

The two missense variants in *ALDH5A1* (XM\_014110599.2: c.866G>A; XP\_013966074.2: p.(Gly288Asp)) and *PTCHD3* (ENSCAFT00000039442.3: c.1247T>C; ENSCAFP00000035301: p.(Iso416Thr)) were evaluated to identify whether the substitutions were potentially deleterious. There is a gap in the canine genome assembly that likely contains at least one additional exon of the *ALDH5A1* gene. Aligning the predicted canine ALDH5A1 protein sequence with the human protein sequence places the canine missense variant at amino acid 381 in human (NP\_001071.1). Both PolyPhen2 [25] (probably damaging—1.0) and SIFT [26] (deleterious—0.0) predicted this amino acid substitution to be deleterious. It occurs in a well -conserved portion of the protein (Figure 6B). The missense variant in *PTCHD3* is not predicted to affect the protein based on PolyPhen2 [25] (benign-0.436) and SIFT [26] (tolerated-0.05). In addition, based on the known functions of these two proteins and the independent findings presented below, the *ALDH5A1* variant was the only one pursued further.

**Figure 3.** Histopathology of mesencephalon and brainstem from dog 1. (**A**) Bilaterally, the mesencephalic nuclei of cranial nerve V (\*), the red nuclei (\*\*), and the substantia nigra (\*\*\*) exhibit decreased staining intensity (H&E). (**B**) Bilaterally, the deep cerebellar nuclei (\*), the dorsal nuclei of the trapezoid body (\*\*), and the reticular formations (\*\*\*) exhibit decreased staining intensity (H&E). (**C**) Higher magnification of (**A**), inset. The substantia nigra shows prominent vacuolation of affected neurons. (**D**) Higher magnification of (**B**), inset. The interposital nucleus shows reactive astrocytes (arrow), some of which also contain prominent cytoplasmic vacuolation (arrowhead).

**Figure 4.** Histopathology of forebrain from dog 1. (**A**) Within the frontal cortex, the grey matter is predominantly affected by spongiotic change, with gliosis. The spongiosis within the forebrain is most severe in the deep laminar cortex. (**C**) The caudate nucleus is also affected at the grey–white matter junction. Affected neurons are characterized by enlarged, vacuolated cytoplasm with a peripheralized nucleus (arrowhead). Luxol fast blue (LFB) staining highlights the deep cortical nature of the vacuolation (**E**) within the cerebrum. (**F**) Vacuolation is discrete and often displaces cellular nuclei LFB staining.

(**B**)

(**D**)

**Figure 5.** GWAS for SSADHD-affected Saluki dogs. (**a**) Manhattan plot showing –log10 of the raw p-values for each genotyped SNP by chromosome (*x*-axis). Genomic inflation was 1.25. Line denotes genome-wide significance based on Bonferroni-corrected p-values. (**b**) Q–Q plot of samples used in GWAS showing the −log10 of the expected versus the observed p-values. The SNPs on CFA35 are shown in light grey.

**Figure 6.** SSADHD-associated *ALDH5A1* missense variant in Saluki dogs. (**A**) The electropherograms from a normal dog (top panel), a heterozygous dog (middle panel) and a dog homozygous for the variant in *ALDH5A1* indicated by an arrow. (**B**) The amino acid alignment around the missense variant in ALDH5A1 (XP\_013966074.2: p.(Gly288Asp)). Yellow boxes indicate 100% conservation across the species listed to the left and blue boxes indicate 75% conservation. The variant amino acid residue is boxed and the variant allele detected in affected Salukis is shown in red. (**C**) The PCR-RFLP genotyping assay for the *ALDH5A1* missense variant is shown. After PCR amplification, the products were digested with Sau96I. L is the DNA ladder, WT stands for wild type (542 bp), C for carrier and M for mutant (702 bp).

Independently, the genome of an affected puppy (dog 5) from the German litter was sequenced. No private variants were found in the *GPLD1* and *PTCHD3* genes. Furthermore, only one proteinchanging missense variant (*ALDH5A1*: XM\_014110599.2: c.866G>A) remained after filtering for homozygous private variants in the region of interest on chromosome 35 against the 589 control genomes from the Dog Biomedical Variant Database Consortium (DBVDC) variant catalog [27].

The *ALDH5A1* variant was confirmed by Sanger sequencing of genomic DNA (Figure 6A). In order to genotype the *ALDH5A1* missense variant, a PCR-RFLP genotyping assay was used (Figure 6C).

Genotyping of the *ALDH5A1* missense variant was performed in the seven affected dogs used for the GWAS, and the four available parents (Figure 1). The variant was homozygous in all cases and heterozygous in the parents. Siblings and other relatives, as well as unrelated Saluki dogs, were genotyped; 13 were heterozygous carriers and 48 were homozygous wild type. The segregation of the *ALDH5A* variant fits perfectly with the assumed monogenic recessive Mendelian inheritance within the studied

family. In-depth pedigree analysis revealed the presence of a common male ancestor connecting the American and European families (Figure 1). To experimentally determine whether the *ALDH5A* variant was a common canine variant, 228 dogs from various other breeds were genotyped and all were found to have the wild-type allele, which was also confirmed by the absence of the variant in 581 dogs from 125 breeds and eight wolves of the DBVDC cohort [27].

The presence of the variant in cDNA of an affected dog was verified by Sanger sequencing of RT-PCR product from liver of an affected Saluki compared to a control unaffected dog. There was no obvious difference in expression level between the case and the control. Quantitative evaluation was not possible since only one affected dog sample was available. Our results have been integrated in the Online Mendelian Inheritance in Animals (OMIA) database (https://omia.org/OMIA002250/9615/).

#### *3.5. Targeted Metabolic Testing*

Targeted quantitative organic acids were analyzed on urine, serum, CSF, and brain tissue (Table 2) in affected (*n* = 1 to 4) and control dogs (*n* = 2 to 4). Compared to control dogs, there were marked elevations in urine succinate semialdehyde (SSA) and urine 4,5-dihydroxyhexanoic acid (DHHA) but levels of gamma hydroxybutyrate (GHB) in the urine were normal. Serum GHB and serum DHHA from affected dogs were markedly elevated compared to controls. Serum SSA could not be measured in either affected or control dogs. In cisternal cerebrospinal fluid (CSF) and brain, SSA, GHB and DHHA were markedly elevated in the affected dog compared to controls, with the CSF GHB having the highest elevation (by a factor of at least 4800). Activity of succinate semialdehyde dehydrogenase (SSADH) was absent or markedly reduced to 0.18% of normal in the affected dogs compared to control dogs.


**Table 2.** Specific quantitative organic acids in urine, serum, CSF, and brain tissue in affected and control dogs (nd = not

 done).

#### **4. Discussion**

A pathogenic variant in the canine *ALDH5A1* gene associated with recessive SSADH deficiency, formerly known as status spongiosus in Saluki dogs (SSSD) [13]. Seven Saluki puppies from two continents had an onset of multifocal cranial neurological signs at 10 weeks of age or younger. Blood work was normal. On MRI of the brain, lesions were similar in all dogs, with bilateral and symmetrical lesions of the cerebrum, brainstem and cerebellum, predominantly affecting grey matter structures. Because this disorder occurred in purebred Saluki puppies from different environments with a consistent clinical phenotype and a normal extracranial work up, structural epilepsy from an inborn error of metabolism was considered the most likely etiology. Using genome-wide association, followed by whole-genome sequencing, a missense mutation in the *ALDH5A1* gene was identified as the presumed cause of status spongiosus in Saluki dogs, previously reported in Saluki puppies [13]. The mutation segregated completely in family members as an apparently fully penetrant monogenic recessive disorder.

The *ALDH5A1* gene encodes the mitochondrial enzyme succinic semialdehyde dehydrogenase (NAD+) (SSADH), which is involved in the catabolism of the inhibitory neurotransmitter gammaaminobutyric acid (GABA) (Figure 7). GABA is the major inhibitory neurotransmitter in the central nervous system(CNS), where it is utilized in up to 30% of cerebral synapses [34]; it is also found in non-nervous-system tissue. GABA is synthesized from L-glutamate via glutamate decarboxylase (GAD). The first step in GABA metabolism is by GABA-transaminase (GABA-T) to form succinic semialdehyde (SSA). SSA is then oxidized by the mitochondrial protein succinic semialdehyde dehydrogenase (NAD+) (SSADH) to succinic acid, which then enters the tricarboxylic acid (TCA) cycle for energy generation (Figure 7). In people, recognized disorders of GABA synthesis are GAD deficiency, and recognized disorders of GABA degradation are GABA transaminase deficiency and succinic semialdehyde dehydrogenase (NAD+) (SSADH) deficiency.

The SSADH protein is expressed in the mammalian brain, as well as liver, pituitary, heart, ovary and kidney [35]. In people, the *ALDH5A1* gene is located on chromosome 6p22, and is 10 exons long extending over 38 kb [36]. Succinic semialdehyde dehydrogenase (NAD+) is an enzyme which is member of the aldehyde dehydrogenase family of proteins. In people with SSADH enzyme deficiency, SSA is not catabolized to succinic acid, and thus excess levels of SSA build up in tissues and fluids (Figure 7). Excess succinic semialdehyde is converted to 4-hydroxybutyric acid (GHB) by succinic semialdehyde reductase. Excess SSA may also interact with an intermediate in the pyruvate dehydrogenase complex to form 4,5-dihydroxyhexanoic acid (DHHA). People with a deficiency of SSADH have elevations of SSA, DHHA, and GHB in body fluids [35]. The activity of SSADH is reduced in people, and thus levels of SSA rise, with associated high levels of GHB and DHHA (Figure 7) [37]. The hallmark of SSADH deficiency in people is persistent and elevated levels of the GHB in urine, plasma and CSF [37]. The diagnosis is confirmed by molecular genetics by sequencing the *ALDH5A1* gene for pathogenic variants. There is no effective therapy [37].

In Saluki dogs with SSADH deficiency, levels of SSA and DHHA are elevated in urine, serum, CSF and brain, and GHB is elevated in serum, CSF and brain. Unlike in people, where GHB is elevated in urine, the level of GHB in urine in Saluki dogs with SSADH deficiency is normal. Since the activity of succinate semialdehyde dehydrogenase (NAD+) (SSADH) was absent or markedly reduced, along with elevated levels of SSA, DHHA and GHB, we believe that the previously described central nervous system status spongiosus in Saluki dogs should be more appropriately termed succinic semialdehyde dehydrogenase deficiency (SSADHD).

In people, SSADH deficiency is a rare autosomal recessive neurological disorder caused by a mutation in the *ALDH5A1* gene, reported in 1981 (OMIM 271980) [38]. There are 44 unique mutations in the *ALDH5A1* gene, which occur in exons 1–10; there are no other mutations in genes other than *ALDH5A1* associated with SSADH deficiency in people [36]. The clinical features in people include developmental delay, hypotonia, intellectual disability, ataxia, seizures, hyperkinetic behavior, aggression and sleep disturbances. Approximately 50% of patients have seizures, 45% have neuropsychiatric problems such as sleep disturbances, and many patients also have behavioral abnormalities [34], which all worsen with age [39]. The encephalopathy is considered non-progressive and has wide phenotypic heterogeneity from mild to severe. Symptoms are first noted at a mean of 11 months (range 0–44 months of age), with a mean age at diagnosis of 6.6 years (range of 0 to 25 years) [40]. On MR imaging in people with SSADHD, MR images may be normal, or there may be hyperintensities on T2-weighted imaging in the globus pallidus, cerebellar dentate nuclei and brainstem [41]. A small percentage of people are reported with cerebral white matter hyperintensity on MR imaging as well [41]. There is one single case report of the pathology of SSADH deficiency in a young adult, where there was discoloration of the globus pallidi, congestion of the leptomeninges and scar tissue in the cerebral cortex [42].

**Figure 7.** GABA catabolism pathway. In Saluki dogs with SSADH deficiency, levels of SSA and DHHA are elevated in urine, serum, CSF and brain, and GHB is elevated in serum, CSF and brain (red arrows) as in people with SSADH deficiency. Unlike in people, where GHB is elevated in urine (red arrow), the level of GHB in urine (red \*) in Saluki dogs with SSADH deficiency is normal.

The pathophysiology of SSADH deficiency in people is complex. The disease is thought to be caused primarily by the elevation of GHB in the brain, particularly during neurodevelopment or due to the imbalance of neurotransmitters [43] but also potentially from oxidative stress in the brain [44]. GHB is a neuromodulator with a wide array of pharmacological effects [45]. It was initially produced as an injectable anesthetic agent but is now longer utilized for this purpose due to adverse effects but is prescribed to treat cataplexy in narcolepsy/cataplexy, opiate dependency and alcoholism, and is a drug of abuse, where its street names include Grievous Bodily Harm, Liquid Ecstasy and Soap [46]. GHB may cause anxiolytic, hypnotic and euphoric effects as well as short-term memory loss and CNS depression, causing sedation [44,47]; intoxication may cause bradycardia, myoclonus and

seizures, hypoventilation, coma and death from respiratory depression [33]. GHB is a monocarboxylate that is primarily cleared from the plasma via metabolism in a dose-dependent fashion through the Kreb's cycle. Renal clearance of GHB is minor and non-linear, due to the carrier-mediated saturable renal reabsorption of GHB through the proximal tubules via sodium-dependent and pH-dependent monocarboxylate transporters. Methods to increase the renal excretion of GHB have been investigated in animals as part of a treatment plan for people with GHB intoxication, such as with the intravenous administration of L-Lactate which increases the renal excretion of GHB [48,49].

As there are no effective specific treatments for SSADH deficiency in people, treatment is currently aimed at managing the clinical signs of seizures and neurobehavioral disturbances [34]. Broad-spectrum anticonvulsants are generally utilized, avoiding valproate which inhibits SSADH which may worsen GHB accumulation and clinical signs [50]. However, there are many therapeutic options under investigation which include pharmacological (e.g., targeting neurotransmitter receptors), enzyme-replacement therapy, gene therapy and treatment with pharmacological chaperones [51,52].

In animals, SSADH has been produced in knockout mice; clinical signs are progressive ataxia, failure to thrive, seizure and death at a young age [53,54], and thus they are utilized as a model for the severe and poor survival phenotype in humans. There are no reports of the MR imaging features of SSADH-knockout mice [51] and, on pathology, there are no reported abnormalities on routine hematoxylin and eosin staining, but detailed neuropathological examinations have not been performed [55]. There is a single case report of a dog with suspected SSADH deficiency, which had a progressive encephalopathy with profound and persistent lactic acidosis, elevated urine GHB and a 30% reduced activity of SSADH measured in cultured lymphoblasts compared to normal dogs. Intracranial MR imaging of the brain was not performed; on histopathology of the brain, there was a spongiform change in the cerebral cortex [56].

Saluki dogs have a more severe phenotype of SSADHD than people, but with very similar clinical signs of seizures, abnormal behavior, abnormalities of sleep and multifocal brain disease on neurological examination. Like people [42], affected Salukis have multifocal abnormalities in the brain on MR imaging, where affected dogs have identical bilaterally symmetrical T2-weighted hyperintensities in the same anatomical areas as people such as the basal nuclei, deep cerebellar nuclei, and brain stem. In people with SSADH deficiency, these abnormalities have considerable consistency, and are almost universal. However, in some people, there are reports of non-specific hyperintensities in subcortical white matter and the substantia nigra in the brainstem, as well as cerebellar atrophy [57,58]. These abnormalities are not considered specific for SSADHD, but considered to be imaging characteristics of cytotoxic edema, secondary to oxidative stress from the underlying SSADHD [43]. Differently to what is noted in people, dogs have hyperintensity of the deep cortical laminae of the grey matter of the cerebral cortex and atrophy of the cerebral cortex. On histopathology in dogs with SSADH deficiency, there is bilaterally symmetric multifocal spongiform change in the brainstem, deep cerebellar nuclei, but also severe in the thalamic nuclei and deep cortical grey matter. The brain lesions in affected Salukis on MRI (two dogs) and pathology (three dogs (same two dogs that had MRI plus one other) were identical. There is only one case report of the pathology of SSADHD in a person where the histopathology was not described and thus the comparative pathology between dogs and people is not possible [42].

Unlike in people with SSADHD, where GHB is elevated in the urine and therefore is an excellent biomarker, GHB is not elevated in the urine of affected dogs. GHB is, however, elevated in the serum, CSF and brain tissue of affected dogs. Since GHB is extensively reabsorbed from the proximal tubules in the kidney [59], it is possible that species differences in handling GHB resulted in extensive reabsorption of GHB and in a normal level of GHB in the affected dog's urine. It is also plausible that since urine GHB was evaluated in only two affected dogs, that evaluation of additional SSADHD-affected dogs may have yielded elevated levels of urine GHB. There may also have been loss of GHB in the samples during shipping to the lab or storage, as GHB is a volatile compound [40] and a reduction in GHB levels is reported with storage, which varied between 10% loss at just 3 days and in excess of 20% after

4 weeks of storage [60]. If the activity of the D-2-hydroxyglutarate dehydrogenase in liver (non-cofactor enzyme converting GHB to SSA and ketoglutarate to D-2-hydroxyglutarate) was very active in the liver of Salukis, this could be a plausible explanation for why GHB was not elevated in the urine; however, D-2-HG was not measured in affected or control dogs. Otherwise, aside from serum SSA that we were not able to measure in dogs with SSADHD, dogs are similar to human patients in regards to elevated body fluid levels of SSA, DHHA and GHB, and zero to low levels of brain SSA activity.

#### **5. Conclusions**

Saluki dogs with SSADHD have a disease phenotype resembling SSADHD in people, although it appears to be more severe clinically. On the other hand, it appears to be less severe than the phenotype described in in knockout mice, which is lethal [58]. Furthermore, GHB may not be an acceptable biomarker for the disease in dogs; alternatively, urine SSA or GHB in serum may be more appropriate biomarkers. Compared to mice models of human disease, dog models have naturally occurring disease, are more similar to humans in regards to size, and have more longevity than mice. Dogs are proven and valuable models of human disease, particularly in the field of lysosomal storage diseases [6]. This first *ALDH5A1*-related large animal model for SSADHD may provide an opportunity for evaluation of potential therapeutics for this rare orphan disease in people. Dogs may be a more appropriate disease model than the murine SSADH model, as dogs appear to have a more similar disease phenotype and similar MR imaging features to people. The identification of the pathogenic *ALDH5A1* variant will allow the screening of carriers to avoid producing further affected puppies and thereby contribute to maintaining breed health.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/11/9/1033/s1. Video S1: A Saluki puppy (dog number 5) showing an episode of abnormal vocalization, Video S2: A Saluki puppy (dog number 1) with bilateral thoracic limb hypermetria.

**Author Contributions:** Conceptualization, K.M.V., D.L.B., E.S., P.L.P., K.M.G., and C.D.; methodology, D.L.B. and E.S.; software, D.L.B.; validation, D.L.B., E.S., and Z.Z.; formal analysis, D.L.B., E.S., B.R., and A.L.; investigation, K.M.V., D.L.B., D.D.C., E.A.B., M.A., K.D.W., G.D.S., F.G., M.R.B., K.M., T.A.M., M.R.B., A.L., V.J., D.L.B., T.J.V.W., and F.K.; resources, D.L.B., E.S., and C.D.; data curation, D.L.B. and E.S.; writing—original draft preparation, K.M.V. and D.L.B.; writing—review and editing, D.L.B., P.J.D., C.D., K.D.W., E.S., P.L.P., K.M.G., G.D.S., F.K.,T.A.M., D.O., and A.L.; visualization, K.M.V., D.L.B., and A.L; supervision, D.L.B., C.T.B., and C.D.; project administration, K.M.V.; funding acquisition, D.L.B. and K.M.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Center of Companion Animal Health, School of Veterinary Medicine, University of California Davis, grant numbers 2015-11F and 2016-18F and the Maxine Adler Endowed Chair Funds.

**Acknowledgments:** The authors acknowledge the support of Sharron Kinney, Lorrie Boldrick, Ember, Encore, Khrome, Kara, Yamal, Yari, Yanam and Kelly Kohen for this study. We acknowledge the Next-Generation Sequencing Platform and the Interfaculty Bioinformatics Unit of the University of Bern for performing the whole-genome sequencing experiments and for providing the computational infrastructure.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article LAMB3* **Missense Variant in Australian Shepherd Dogs with Junctional Epidermolysis Bullosa**

**Sarah Kiener 1,2,**†**, Aurore Laprais 3,**†**, Elizabeth A. Mauldin 4, Vidhya Jagannathan 1,2, Thierry Olivry 5,\* and Tosso Leeb 1,2,\***


<sup>4</sup> School of Veterinary Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA; emauldin@vet.upenn.edu


Received: 10 August 2020; Accepted: 3 September 2020; Published: 7 September 2020

**Abstract:** In a highly inbred Australian Shepherd litter, three of the five puppies developed widespread ulcers of the skin, footpads, and oral mucosa within the first weeks of life. Histopathological examinations demonstrated clefting of the epidermis from the underlying dermis within or just below the basement membrane, which led to a tentative diagnosis of junctional epidermolysis bullosa (JEB) with autosomal recessive inheritance. Endoscopy in one affected dog also demonstrated separation between the epithelium and underlying tissue in the gastrointestinal tract. As a result of the severity of the clinical signs, all three dogs had to be euthanized. We sequenced the genome of one affected puppy and compared the data to 73 control genomes. A search for private variants in 37 known candidate genes for skin fragility phenotypes revealed a single protein-changing variant, *LAMB3*:c.1174T>C, or p.Cys392Arg. The variant was predicted to change a conserved cysteine in the laminin β3 subunit of the heterotrimeric laminin-322, which mediates the binding of the epidermal basement membrane to the underlying dermis. Loss-of-function variants in the human *LAMB3* gene lead to recessive forms of JEB. We confirmed the expected co-segregation of the genotypes in the Australian Shepherd family. The mutant allele was homozygous in two genotyped cases and heterozygous in three non-affected close relatives. It was not found in 242 other controls from the Australian Shepherd breed, nor in more than 600 other controls. These data suggest that *LAMB3*:c.1174T>C represents the causative variant. To the best of our knowledge, this study represents the first report of a *LAMB3*-related JEB in domestic animals.

**Keywords:** dog; *Canis lupus familiaris*; whole genome sequence; wgs; dermatology; genodermatosis; skin; laminin; precision medicine

#### **1. Introduction**

When a human or animal, usually at or soon after birth, develops erosions and epithelial sloughing on the mucosae, areas of friction, and extremities, a genetic disorder of skin fragility is to be considered. A consensus reclassification of skin fragility disorders was published recently, which separates those that affect the basement membrane itself or the basal keratinocytes (i.e., hereditary epidermolysis bullosa (EB) variants) from others, in which the separation occurs more superficially in the epidermis [1]. In this reclassification, four main categories of inherited "classical" EB are proposed, which reflect the differences in the level of cleavage in the basement membrane zone [1]. Also included in this reclassification are four new categories of epidermal disorders of skin fragility associated with 20 possibly mutated genes, namely: peeling skin disorders, erosive skin fragility disorders, keratinopathic ichthyoses, and pachyonychia congenita [1]. Finally, a single syndromic connected tissue disorder with (dermal) skin fragility associated with *PLOD3* variants and a lysyl hydroxylase-3 deficiency was also included in this group of diseases [1]. All of the known 37 candidate genes for these human diseases are summarized in Table 1.


**Table 1.** Consensus reclassification of epidermolysis bullosa and other disorders with epidermal fragility and their known functional candidate genes, as of 2020 [1].

<sup>1</sup> AD—autosomal dominant; AR—autosomal recessive.

In domestic dogs, only two other epidermal disorders of skin fragility have been reported, namely: epidermolytic ichthyosis associated with a *KRT10* variant in Norfolk terriers [2], and ectodermal dysplasia/skin fragility syndrome with a *PKP1* variant in Chesapeake Bay Retrievers [3] (Table S1). In contrast, cases of hereditary EB have been recognized for decades, and the causative genetic variants have now been characterized in three canine, one feline, two equine, two ovine, and five bovine EB variants (Table S1).

In dogs, there is at least one example for each of the three main subtypes of classical EB in which the genetic defect has been reported, namley: a *PLEC* variant in the EB simplex of Eurasier dogs in the USA [4]; a *LAMA3* variant in the junctional EB of German Shorthaired Pointers in France [5]; and *COL7A1* variants in the dystrophic EB (mild) in Golden Retrievers, also in France [6], or severe in Central Asian Shepherds [7].

Laminin-332, a rod-like heterotrimer composed of the laminin α3, β3, and γ2 chains, is a critical component of hemidesmosomes, adhesion complexes that attach the basal epidermal keratinocytes to the underlying dermal connective tissue [8–10]. The prominent role of laminin-332 for skin integrity stems from its ability to link two important molecules—one in the epidermis and the other in the dermis. Via its carboxy-terminus, laminin α3 binds to the external domains of the integrin α6β4 that protrude from the basal keratinocytes. At the other end of the laminin trimer, the amino-terminal domains of the laminin β3 and γ2 chains bind to the NC1 amino-terminus of the superficial dermal collagen type VII [11].

Genetic variants in the *LAMA3*, *LAMB3,* and *LAMC2* genes that encode the laminin α3, β3, and γ2 chains are causative for the intermediate and severe forms of junctional EB (JEB), not only in humans [1], but also in animals (Table S1). Variants in any one of these genes can lead to a similar phenotype, as the abnormal expression or function of either of the three individual laminin chains is expected to impair the assembly or the function of the entire laminin-332. A good example of this phenomenon is the near identical phenotype exhibited by American Saddlebred horses with severe JEB associated with a *LAMA3* variant [12], and that found in Belgian, Breton, Comtois, and Italian draft horses caused by a *LAMC2* variant [13–15].

While JEB subsets associated with *LAMB3* variants are common in humans [16–18], they have not yet been reported in animals. So far, an abnormal epidermal expression of laminin β3—without investigation of the underlying molecular genetics—has only been shown in a single cat exhibiting a phenotype of mild EB [19].

Herein, we report a missense variant in *LAMB3,* which we believe is causative of a JEB phenotype with intermediate severity in a litter of Australian Shepherds in Ontario, Canada. Of clinical interest is the demonstration, for the first time or so it seems, of intestinal epithelial sloughing in a case of animal JEB.

#### **2. Materials and Methods**

#### *2.1. Ethics Statement*

The affected Australian Shepherds in this study were privately owned, and skin and biopsy samples were collected with the consent of their owners. The collection of all other blood samples was approved by the "Cantonal Committee for Animal Experiments" (Canton of Bern; permits 75/16 and 71/19).

#### *2.2. Animal Selection*

This study included 247 Australian Shepherds. Genomic DNA was either isolated from EDTA blood samples with the Maxwell RSC Whole Blood Kit, or from formalin-fixed paraffin-embedded (FFPE) tissue samples with the Maxwell RSC DNA FFPE Kit using a Maxwell RSC instrument (Promega, Dübendorf, Switzerland).

#### *2.3. Histopathological Examinations*

Skin punch biopsies (8 mm) were obtained under general anesthesia. The samples were fixed in 10% neutral buffered formalin and routinely processed, including staining with hematoxylin and eosin.

#### *2.4. Whole Genome Sequencing*

An Illumina TruSeq PCR-free DNA library with ~400 bp insert size of a JEB affected Australian Shepherd was prepared. We collected 175 million 2 × 150 bp paired-end reads or 18.9× coverage on a NovaSeq 6000 instrument. The reads were mapped to the dog reference genome assembly CanFam3.1 and were aligned as previously described [20]. The sequence data were submitted to

the European Nucleotide Archive, with study accession number PRJEB16012 and sample accession number SAMEA6862980.

#### *2.5. Variant Calling*

Variant calling was performed as previously described [20]. To predict the functional effects of the called variants, SnpEff software [21], together with NCBI annotation release 105 for the CanFam 3.1 genome reference assembly, was used. For variant filtering, we used 73 control genomes (Table S2).

#### *2.6. Gene Analysis*

We used the dog reference genome assembly CanFam3.1 and NCBI annotation release 105. Numbering within the canine *LAMB3* gene corresponds to the NCBI RefSeq accession numbers XM\_014115071.2 (mRNA) and XP\_013970546.1 (protein). For a multiple species comparison of the *LAMB3* amino acid sequences, we used the following accessions: NP\_000219.2 (*Homo sapiens*), NP\_001075065.1 (*Bos taurus*), XP\_023496552.1 (*Equus caballus*), NP\_001264857.1 (*Mus musculus*), NP\_001094311.1 (*Rattus norvegicus*), XP\_425827.3 (*Gallus gallus*), XP\_002933550.2 (*Xenopus tropicalis*), and XP\_700808.6 (*Danio rerio*).

#### *2.7. Sanger Sequencing*

To confirm the candidate variant *LAMB3*:c.1174T>C, and to genotype all of the dogs in this study, Sanger Sequencing was used. A 403 bp PCR product was amplified from genomic DNA using AmpliTaqGold360Mastermix (Thermo Fisher Scientific, Waltham, MA, USA) and primers 5 -TCT TGT GCC AAG CAC TGT TC-3 (Primer F) and 5 -GGC ATA GGT GAG TCC CGT AA-3 (Primer R). A smaller PCR product of 153 bp size was amplified for FFPE-derived DNA with primers 5 -GGT GGC TGC TTT TCT GTC TC-3 (Primer F) and 5 -GGT GAG TCC CGT AAA TCC TG-3 (Primer R). After treatment with shrimp alkaline phosphatase and exonuclease I, PCR amplicons were sequenced on an ABI 3730 DNA Analyzer (Thermo Fisher Scientific). Sanger sequences were analyzed using the Sequencher 5.1 software (GeneCodes, Ann Arbor, MI, USA).

#### **3. Results**

#### *3.1. Family Anamnesis, Clinical Examinations, and Histopathology*

Three Australian Shepherd puppies with severe skin lesions were identified in a highly inbred litter resulting from a father–daughter mating. The litter consisted of three affected and two non-affected puppies that were born out of normal parents. The pedigree relationships were suggestive for a monogenic autosomal recessive inherited disease (Figure 1).

**Figure 1.** Pedigree of the investigated Australian Shepherd family. Squares represent males and circles represent females. The three affected puppies are indicated by the filled symbols. Note that the father of the litter was simultaneously the maternal grandfather. A close inbreeding loop greatly increases the risk for recessive hereditary defects. Genotypes at the *LAMB3*:c.1174T>C variant are indicated for all animals, from which a DNA sample is available (see Section 3.2).

At the time of their first presentation to the breeder's veterinarian for vaccination at 7 weeks of age, the three affected puppies were noted to have ulcers in the mouth, inner pinnae, and abdomen. The puppies reportedly also had marked lymph node enlargement. The average weight of the affected puppies was half that of their unaffected siblings.

At 17 weeks of age, one of the affected dogs, a blue merle with copper intact female was presented to the dermatologist for evaluation of severe ulceration of both the oral cavity and haired skin. Ulcers were located on the gingival and buccal mucosa, tongue, and hard and soft palates (Figure 2a). The concave pinnae, bilaterally, were also ulcerated, oozing, and covered with exudate (Figure 2b), but the otoscopic examination only revealed mild erythema in the ear canal. Several footpads, either digital or central, were also ulcerated (Figure 2c), and four claws were missing or misshapen. Erosions and ulcers were covered by thick crusts on the elbows, hocks, and the tip of the tail. The vulva and anus were grossly normal.

**Figure 2.** Clinical and histopathological phenotype. (**a**) Severe coalescing ulcers on the gingiva and hard and soft palate, (**b**) concave pinna (**c**) and footpads. Biopsy samples collected from the (**d**) oral cavity and (**e**) duodenum revealed widespread separation of the epithelium from the underlying connective tissue (asterisks).

Thoracic auscultation, abdominal, and lymph node palpation were all unremarkable. Blood was collected for a complete blood count and a serum chemistry panel, and the most relevant changes were a mild regenerative anemia (hemoglobin: 129 (reference range: 134–207 g/L); reticulocytes: 118 (10–110 k/μL)) and hypoproteinemia (total proteins: 46 (55–75 g/L); albumin: 23 (27–39 g/L); globulins: 23 (24–40 g/L)). To determine if these abnormal changes were due to digestive ulcers, an upper gastrointestinal endoscopy was performed under general anesthesia, two weeks after the original admission to the specialty clinic. The esophagus appeared normal, and the stomach and duodenum were hyperrhemic but did not show a visible loss of epithelium; endoscopic biopsies were nevertheless collected from the stomach and duodenum. During this general anesthesia, punch skin biopsies were collected from the concave pinnae, footpads, and oral cavity (hard palate, buccal mucosa, and tongue).

Microscopically, the skin and mucosal biopsy samples all exhibited limited-to-widespread epidermal detachment (Figure 2d), and ulcers were covered with serocellular crusts; inflammation was sparse in non-ulcerated areas. In some sections (as in Figure 2d), the basement membrane could be discerned at the base of the clefts, thus suggesting the diagnosis of JEB. The endoscopic biopsies from the stomach (pyloric and nonpyloric areas) and duodenum all showed mild-to-moderate inflammation with lymphocytes, plasma cells, and eosinophils, with a detachment of the epithelium from the underlying lamina propria (Figure 2e). Because of the severity of the lesions, the dog was euthanized at 7.5 months of age.

The medical records of the two other affected puppies were also reviewed. A blue merle with copper intact male puppy was noted to have ulcers on the tongue, gingiva, soft palate, pharynx, tonsils, and larynx. Skin ulcers were found on the concave pinnae and pressure points of one elbow, one hock, and both stifles. Because of the worsening lesions, this puppy was euthanized at 16 weeks of age, with biopsy samples of the tongue, soft palate ear, and footpad collected post-mortem. As for the samples obtained from the littermate described above, microscopic lesions consisted of subepidermal vesicles leading to dermo-epidermal separation, ulceration, and granulation tissue.

The third affected puppy, a blue merle female, had been euthanized at 5 months of age because of severe gingival, labial, oropharyngeal, and esophageal ulceration. The dog had crusts on the chin, ulcers and crusts on the concave pinnae and footpads, and exudate at the base of multiple claws; samples for histopathology were not collected.

Finally, both the sire and dam, as well as the two healthy siblings, were examined by veterinarians, and they were deemed to be free of skin lesions.

#### *3.2. Genetic Analysis*

In order to characterize the underlying causative genetic variant, we sequenced the genome of one affected dog at 18.9× coverage and searched for homozygous variants in the 37 genes known to cause human skin fragility (Table 1), which were exclusively present in the affected dog and absent from the genomes of 73 other dogs (Table 2, Tables S2 and S3).

**Table 2.** Results of variant filtering in the affected Australian Shepherd dog against 73 control genomes. Only homozygous variants are reported.


This analysis identified a single homozygous private protein-changing variant in *LAMB3*, a known candidate gene for JEB in humans [1]. The variant can be designated as Chr7:8,286,613A>G (CanFam3.1 assembly). It is a missense variant, XM\_014115071.2:c.1174T>C, predicted to change a highly conserved cysteine residue in the third EGF-like domain of laminin β3, XP\_013970546.1:p.(Cys392Arg).

We confirmed the presence of the *LAMB3* missense variant by Sanger sequencing (Figure 3). The mutant allele showed the expected co-segregation with JEB in the available family. The two available DNA samples from the JEB affected puppies carried the mutant allele in a homozygous state, while their parents were heterozygous, as expected for obligate carriers (Figure 1).

We determined the genotypes at *LAMB3*:c.1174T>C in a cohort comprising 247 Australian Shepherd dogs, including the index family. The mutant *LAMB3* allele was not detected in the homozygous state in any of the 245 non-affected Australian Shepherd dogs or 663 dogs from other breeds. Three of these dogs, all members of the index family, carried the mutant *LAMB3* allele in a heterozygous state (Table 3).

**Figure 3.** Details of the *LAMB3*:c.1174T>C, p.Cys392Arg variant. (**a**) Representative electropherograms of three dogs with different genotypes are shown. The variable position is indicated by an arrow, and the amino acid translations are shown. (**b**) Domain organization of the 1172 amino acid laminin β3 precursor [8]. The N-terminus consists of a globular domain (LN), followed by six laminin EGF-like (LE) domains. These N-terminal domains are located in the basement membrane and may be involved in binding to collagen VII. The C-terminal half of laminin β3 participates in two coiled-coil domains that mediate trimerization with the α3 and γ2 chains in the laminin-332 heterotrimer. The small Lβ domain mediates the binding of agrin. (**c**) Multiple-species alignment of the beginning of the LE3 domain harboring the p.Cys392Arg variant. The variant affects a highly conserved cysteine residue that forms a disulfide bridge with Cys-379 [22]. Note that all six cysteine residues in this region contribute to disulfide bonds, and are strictly conserved across vertebrates.

**Table 3.** Genotype-phenotype association of the *LAMB3*:c.1174T>C variant with JEB.


<sup>1</sup> These genotypes were derived from 590 genome sequences reported in the literature [20], and the 73 control genomes used in this study.

#### **4. Discussion**

In the affected Australian Shepherds described in this study, the age of lesion onset, as well as the presence of ulceration in the oral cavity and pressure points on the limbs with a loss of claws, all suggested the clinical diagnosis of a skin fragility disorder, of which EB is the most representative disease group in domestic animals and humans (Table S1). Because of the resembling phenotypes, clinical signs cannot alone reliably permit differentiation between the three main subtypes of animal EB (simplex, junctional, and dystrophic). For a more precise diagnosis, the specific location of the dermo-epidermal separation must be determined, for example, with a periodic acid Schiff (PAS) stain to visualize the glycoproteins in the basement membrane lamina densa [7], single or double antigen immunomapping [23], or transmission electron microscopy [4]. In this case, the routine histopathology enabled the visualization of the basement membrane delineating the contour of dermal imprints of the epidermal ridges, thus establishing that clefting occurred in a supra-lamina densa manner; this confirmed the diagnosis of JEB.

There is only one other occurrence of JEB in the canine species [5,23,24]. In the early 1990s, JEB was first discovered in German Shorthaired Pointers in the French Alps. The clinical signs were indistinguishable from those present in the Australian Shepherds described herein. Both the Pointer and Australian Shepherd puppies exhibited the first clinical signs weeks after, and not at, birth. In both cases, lesions consisted of ulcerative skin lesions affecting the inner (medial and concave) pinnae, footpads, and at pressure points of the extremities [23,24]. Shedding of the claws was also reported [24]. Of interest is that dental enamel abnormalities, a common finding in human JEB [1], were not recognized in either the German Shorthaired Pointers or the Australian Shepherds described in this study. A unique finding seen in one of the three Australian Shepherd puppies was the endoscopic observation of duodenal hyperrhemia, which was found on the histopathology to be associated with an extensive detachment of the digestive epithelium from its underlying connective tissue. Unfortunately, as ulceration of the duodenum was not seen during endoscopy, we cannot rule out that the digestive epithelial detachment seen on the histopathology might have been artifactual. Nevertheless, the *LAMA3, LAMB3,* and *LAMC2* genes, which encode the three laminin-332 chains, are all expressed in the small intestine [25]. As a result, based on our hypothesis, that the *LAMB3* missense variant affects the adhesive function of the laminin-332, it is conceivable that any trauma to the small intestine during the endoscopic biopsy process could result in a forced epithelial separation from the lamina propria, a phenomenon that normally does not happen to that extent in healthy individuals. To our knowledge, such a lesion has never been reported in a case of animal EB, and these are findings seen more often in the severe generalized than intermediate variants of human JEB; they are typically not found in localized JEB [26].

In this study, we identified a homozygous missense variant, *LAMB3*:p.Cys392Arg, as a candidate causative variant for a new JEB in Australian Shepherd dogs. *LAMB3* encodes the laminin β3 chain, which, together with the α3 and γ2 chains, forms the heterotrimer laminin-332. Laminin β3 has two coiled-coil domains for the heterotrimer formation with the α3 and γ2 chains at its C-terminal end. The N-terminus consists of a globular domain (LN) and six laminin-type epidermal growth factor-like (LE) repeats [8,22]. The LE domains have conserved disulfide bonds, which may be important for the tertiary structure of these domains [27,28]. The LN and LE domains form a short arm in the cross-shaped laminin-332 heterotrimer, and mediate binding to type VII collagen in hemidesmosomes, which are necessary for the stable association between the epithelium and the stroma underneath [8,11].

The p.Cys392Arg variant changes one of the highly conserved cysteine residues in the third LE domain, which prevents the formation of the disulfide bond between Cys-392 and Cys-379. We hypothesize that this may lead to a change in the tertiary structure of laminin β3, and impair the binding of laminin-322 to collagen type VII in hemidesmosomes. Further experiments at the protein level are required in order to confirm this putative pathomechanism.

With this description, we now have two variants of canine intermediate JEB due to variants in related genes (*LAMA3* in German Shorthaired Pointers and *LAMB3* in Australian Shepherds) encoding the laminin α3 and β3 chains that assemble with the γ2 chain to form the laminin-332 heterotrimer. In both of these breeds, the variants are predicted to result in some residual protein function (Australian Shepherds), or in the secretion of some normal laminin-332 trimers (German Shorthaired Pointers) [5], which may explain the similar absence of lesions at birth and the intermediate clinical phenotype.

In humans, the specific variants and their consequences at the mRNA and protein levels contribute to the spectrum of severity encountered in different subtypes of EB [10]. Severe forms of JEB are associated with nonsense, frameshift, or out-of-frame splicing variants that result in nonfunctional or complete loss of the protein. Intermediate JEB occurs when a laminin chain is mutated, but the LM-332 heterotrimer can still form, which is often the case for missense variants [16]. Missense variants affecting cysteine residues in the LE domains, *LAMB3*:p.Cys355Arg and p.Cys433Trp, have been reported in human patients with intermediate JEB [16–19]. The clinical phenotype observed in the investigated dogs homozygous for p.Cys392Arg can also be classified as JEB of intermediate severity, and corresponds well to the human spectrum of genotype–phenotype correlations.

#### **5. Conclusions**

We characterized a new recessive form of JEB in Australian Shepherd dogs. A precision medicine approach identified a missense variant in the *LAMB3* gene, c.1174T>C or p.Cys392Arg as likely candidate causative variant. Our data enable genetic testing to avoid the unintentional breeding of further affected dogs and provide the first spontaneous large animal model for JEB due to altered laminin β3.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/11/9/1055/s1: Table S1, genetic variants causing epidermolysis bullosa or skin fragility disorders in animals. Table S2, accession numbers of 74 dog genome sequences. Table S3, private variants in a JEB affected Australian Shepherd dog.

**Author Contributions:** Conceptualization, T.O. and T.L.; data curation, V.J.; investigation, S.K., A.L., E.A.M., T.O., and T.L.; writing (original draft), S.K., A.L., T.O., and T.L.; writing (review and editing), S.K., A.L., E.A.M., V.J., T.O., and T.L. supervision, T.O. and T.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** We thank Sophie Saati for collection of the endoscopic samples and Andrew Lowe for the skin biopsies. The authors are grateful to all dog owners and referring veterinarians who donated samples and shared health and pedigree data of their dogs. We thank Nathalie Besuchet Schmutz, Catia Coito, Marion Ernst, and Daniela Steiner for their expert technical assistance; the Next Generation Sequencing Platform of the University of Bern for performing the high-throughput sequencing experiments; and the Interfaculty Bioinformatics Unit of the University of Bern for providing high performance computing infrastructure. We thank the Dog Biomedical Variant Database Consortium (Gus Aguirre, Catherine André, Danika Bannasch, Doreen Becker, Brian Davis, Cord Drögemüller, Kari Ekenstedt, Kiterie Faller, Oliver Forman, Steve Friedenberg, Eva Furrow, Urs Giger, Christophe Hitte, Marjo Hytönen, Vidhya Jagannathan, Tosso Leeb, Frode Lingaas, Hannes Lohi, Cathryn Mellersh, Jim Mickelson, Leonardo Murgiano, Anita Oberbauer, Sheila Schmutz, Jeffrey Schoenebeck, Kim Summers, Frank van Steenbeek, and Claire Wade) for sharing the whole genome sequencing data from the control dogs. We also acknowledge all canine researchers who deposited dog whole genome sequencing data into public databases.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
