**Survey of SNPs Associated with Total Number Born and Total Number Born Alive in Pig**

#### **Siroj Bakoev 1 , Lyubov Getmantseva 1, \*, Faridun Bakoev 1,2 , Maria Kolosova 1,3 , Valeria Gabova 2 , Anatoly Kolosov 1,3 and Olga Kostyunina 1**


Received: 12 April 2020; Accepted: 29 April 2020; Published: 30 April 2020

**Abstract:** Reproductive productivity depend on a complex set of characteristics. The number of piglets at birth (Total number born, Litter size, TNB) and the number of alive piglets at birth (Total number born alive, NBA) are the main indicators of the reproductive productivity of sows in pig breeding. Great hopes are pinned on GWAS (Genome-Wide Association Studies) to solve the problems associated with studying the genetic architecture of reproductive traits of pigs. This paper provides an overview of international studies on SNP (Single nucleotide polymorphism) associated with TNB and NBA in pigs presented in PigQTLdb as "Genome map association". Currently on the base of Genome map association results 306 SNPs associated with TNB (218 SNPs) and NBA (88 SNPs) have been identified and presented in the Pig QTLdb database. The results are based on research of pigs such as Large White, Yorkshire, Landrace, Berkshire, Duroc and Erhualian. The presented review shows that most SNPs found in chromosome areas where candidate genes or QTLs (Quantitative trait locus) have been identified. Further research in the given direction will allow to obtain new data that will become an impulse for creating breakthrough breeding technologies and increase the production efficiency in pig farming.

**Keywords:** pig; litter size; total number born alive; SNP (single nucleotide polymorphism); GWAS (Genome-Wide Association Studies)

#### **1. Introduction**

In the early 1990s the work on mapping the pig genome started and maps containing more than 1200 microsatellite markers appeared due to the development of the international project "PiGMap" and projects of the US Department of Agriculture and American agricultural institutes. These maps were used to identify the Quantitative Trait Loci (QTL) underlying the genetic architecture of pig productivity traits [1–3]. To date, an extended database has been created—Pig Quantitative Trait Locus Database (PigQTLdb), which presents 30170 QTLs for 688 pig trait of deferent classes [4].

The sows' reproductive potential is the basis for continuous and efficient production. In recent decades, the BLUP (Best linear unbiased prediction) method has made a significant contribution to improving reproduction rates [5,6]. However, low heritability coefficients of reproductive traits and their sex-limited phenotypic exhibition lead to developing new approaches revealing the biological nature of reproductive performance. Great hopes are pinned on GWAS (Genome-Wide Association Studies) to solve the problems associated with studying the genetic architecture of reproductive traits of pigs [7–10].

GWAS results can be represented by the information on detected associations with various genetic aberrations: chromosomal mutations (whole chromosomes or their fragments), large inserts or deletions (100–100,000 nucleotides), small inserts or deletions (1 to 100), single nucleotide polymorphisms (SNP) [11]. Each SNP is represented by at least two alleles: minor (rarer) and major. Genetic variations with minor allele frequency exceeding 0.01% are numbered and assigned to rs index [12].

Currently SNP BeadChip technology is more affordable for genome-wide research than sequencing is. SNP BeadChips have been developed to include high (HD), medium (MD), or low (LD) genome distribution of markers. Rigid structure is noted as a disadvantage, which allows us to analyze only what is already predetermined by the BeadChip design and in this connection the potentially important information can be omitted [13]. Markers do not have the same density across all chromosomes and not fully track structural genetic variations such as insertions and deletions [13]. However, despite these shortcomings, SNP BeadChip has recently gained great popularity in studies of the genetic architecture of quantitative traits of farm animals and pigs in particular [14–16].

The first SNP BeadChips with a resolution of about 60 thousand markers covering all autosomal and X chromosome genes PorcineSNP60 BeadChip v2 were presented by the American company Illumina (San Diego, CA, USA). In addition to SNP BeadChip with high density, LD SNP BeadChip with low density have been proposed to reduce genotyping costs. Commercial LD SNP BeadChip «GeneSeek/Neogen GPP-Porcine LD Illumina Bead Chip panel» were also developed by «GeneSeek/Neogen» (Lincoln, NE, USA). Besides this, the company introduced HD SNP BeadChips with a higher density (about 70 thousand markers). HD SNP BeadChips containing about 650 thousand markers and including all markers of the «Illumina PorcineSNP60 BeadChip v2» array, were produced by «Affymetrix» (Santa Clara, CA, USA). It should be noted that SNP BeadChip can be custom made including specific SNPs associated with given traits of productivity.

Reproductive productivity depend on a complex set of characteristics. The number of piglets at birth (Total number born, Litter size, TNB) and the number of alive piglets at birth (Total number born alive, NBA) are the main indicators of the reproductive productivity of sows in pig breeding [17–21]. These indicators reflect the level of all physiological processes associated with fertilization, intrauterine development of the fetus and a sow's labor, and are also quite easy to account for.

#### **2. SNPs Associated with Total Number Born Alive**

Currently on the base of Genome map association results 88 SNPs associated with NBA have been identified and presented in the Pig QTLdb database (Table 1). The results are based on a study of pigs such as Large White, Large White, Yorkshire, Landrace [7,14,22–26] Duroc [27] and Erhualian [28]. SNPs, associated with the NBA are represented in all Sus scrofa chromosomes (SSC) except SSCY.




**Table 1.** *Cont.*

*Genes***2020**, *11*, 491


**Table 1.** *Cont.*

*Genes***2020**, *11*, 491



*Genes***2020**, *11*, 491

Search for SNPs associated with NBA of Chinese Erhualian pigs was carried out by Ma et al. [28]. Sows with high and low estimated breeding values (EBVs) were selected for genotyping. According to the research results, 9 SNPs associated with Pig QTLdb with NBA were presented. The greatest effect was found for SNP rs81447100 (SSC13), which was additionally tested on Erhualian pigs (*n* = 313), Sutai (*n* = 173) and Yorkshire (*n* = 488). In all groups under study, a significant association between allelic variants of SNP rs81447100 (SSC13) and NBA was determined. However, allele A was desirable for the Erhualian pigs and allele G for the Sutai and Yorkshire pigs.

According to research work of Wu et al. [14], conducted on Landrace and Yorkshire pigs the database contains 15 SNPs localized on SSC1 and SSC8 and associated with NBA. 11 of these SNPs are located in the QTL regions annotated earlier, and 4 SNPs are presented for the first time. All these 4 SNPs (rs329624627, rs339929690, rs322202112 and rs330585697) are located on the SSC1. The most significant effect was established for SNP rs332924521 (SSC1).

Coster et al. [23] conducted associative studies on Large White pigs from two commercial lines of «Hypor» and «Topigs», and revealed 4 SNPs associated with NBA located on SSC7 (rs81397142 and rs81397215), SSC1 (rs81348779) and SSC2 (rs81356698).

Bergfelder-Drüing et al. [7] conducted research using Large White and Landrace pigs. Preliminary calculations based on graphs of multidimensional scaling showed the genetic distance between the breeds. For analysis pigs were divided into two clusters taking into account the breed and intra-breed clusters taking into account the breeding economy. As a result, 17 SNPs associated with NBAs were identified, 5 of these SNPs had a minor allele frequency less than 1%. The study of Bergfelder-Drüing et al. [7] was the first to show an association with NBA for SNP rs81430147 (SSC11). All other SNPs were found in chromosome regions where candidate genes or QTLs affecting pig reproductive traits have already been identified. It should be noted that different SNPs were established for each breed cluster, and no associative communications were established simultaneously in two breeds. 13 SNPs were identified for Large White sows, and 4 SNPs for Landrace sows. In sows Landrace SNPs are localized on SSC7, SSC9, SSC11 and SSC16. Large White on SSC3, SSC5, SSC9, SSC10, SSC11 and SS18. The most significant effects on NBA of Large White sows are found for 4 SNPs: rs81379421 (SSC3), rs81417393 (SSC9), rs81242222 (SSC11) and rs81469701 (SSC18).

Research carried out by Wang et al. [22] on Large White pigs identified 6 SNPs associated with NBA located on SSC2, SSC3, SSC13, SSC14 and SSC18. The most significant effect was determined for SNP rs334867206 (SSC3), according to which pigs of AA genotype had more NBA, compared with analogues of GG genotype.

Suwannasing et al. [26] investigated Large White and Landrace pigs and established 25 SNPs for the NBA. Of these, 11 SNPs located in SSC1 and SSCX were defined for Large White pigs, 14 SNPs for Landrace pigs in SSC2 and SSC6. It is remarkable that SNPs on SSCX showed significance only at the NBA for Large White sows.

In the course of research on commercial pigs, Li et al. [25] established 2 SNPs (rs342908929 (SSC6) and rs324003968 (SSC15)) associated with NBA. According to the results of Chen et al. [27] obtained from the sows of Duroc in the database presents 9 SNPs and all of them are localized on SSC6.

An et al. [24] studied the *IGFBP2* (SSC15) and *IGFBP3* (SSC18) genes in Berkshire pigs. Their results showed significant SNPs in these genes. This work also analyzed the expression levels of *IGFBP2* and *IGFBP3* mRNA in the endometrium in pigs of various genotypes. Homozygous GG pigs expressed higher levels of *IGFBP3* mRNA in the endometrium than pigs of other genotypes, and a positive correlation was observed between litter size traits and *IGFBP3* but not *IGFBP2* expression level. These results suggest that SNPs in the *IGFBP2* and the *IGFBP3* gene are useful biomarkers for the little traits of pigs. According to the results of this work 2 SNPs are included in PigQTLdb rs45435330 on SSC15 as Genome map association for NBA and TNB.

#### **3. SNPs Associated with Total Number Born**

In general, in Pig QTLdb for TNB showed 218 SNPs (as Genome map association), of which 155 SNPs were detected by He et al. [8] and Ma et al. [28] in Erhualians, 52 SNPs are defined in Large White, Yorkshire, Landrace, 1 SNP in Berkshire and 10 SNPs in Duroc (Table 2) [14,18,22–25,29–33]. SNPs associated with TNB are represented in all Sus Scrofa Chromosome except for SSCY.

For an associative study He et al. [8] selected Erhualan sows with high and low EBV values. According to the results of their work, the most significant SNPs were detected on SSC2 chromosomes (rs81367039), SSC7 (rs80891106), SSC8 (rs81399474), SSC12 (rs81434499), SSC14 (rs80938898). Among them SNPs on chromosomes SSC2, SSC7, and SSC12 were annotated for the first time. To study the effect of significant SNPs additional studies were conducted on a livestock of 331 Erhualan sows. According to the results of additional testing, significant differences in TNB were found only for SNP rs81399474 (SSC8). In the studies of Ma et al. [28] 8 SNPs were identified as associated with TNB in Erhualian sows and located on SSC1, SSC4, SSC7, SSC8, SSC10, SSC12, SSC13and SSC16.

Studies of Large White pigs conducted by Sell-Kubiak et al. [30] allowed the identification of 10 SNPs associated with the number of piglets at birth and located on SSC1, SSC5, SSC8, SSC11, SSC13 and SSC18. SNPs (rs80989787 and rs81289355) located on SSC11 were annotated for the first time in this paper. The most significant effect has been determined for SNP rs80989787 (SSC11). According to the results of associative studies conducted by Uimari et al. [18] on Finnish Landrace pigs 10 SNPs were identified. All established SNPs are located on SSC9.

The most significant effect was established for SNP rs81300575 (SSC9), which amounted to about 1 piglet between two homozygous genotypes. In the studies of Uimari et al. [18] it was also noted that in the past 15 years the frequency of the desired SNP rs81300575 (SSC9) genotype in the studied population has increased from 0.14 to 0.22.

Zhang et al. [29] conducted research on Duroc pigs and identified 10 SNPs associated with TNB. The most significant SNPs were rs80979042 and rs80825112 located on SSC14. In addition, the remaining potential SNPs were located on SSC5, SSC6, SSC12 and SSC17.

In the studies of Coster et al. [23], Wu et al. [14], Wang et al. [22] and Li et al. [25] the effects of SNPs on TNBs were investigated along with the search of SNPs associated with NBAs. So according to the results of Coster et al. [23] the database contains 16 SNPs defined on SSC1, SSC2, SSC7, SSC14 and SSC18. Wu et al. [14] established 5 SNPs associated with TNB located on SSC8 and SSC14. Wang et al. [22] identified 11 SNPs for TNB on SSC1-SSC5, SSC13 and SSC18, and Li et al. [25] 2 SNPs (rs342908929 (SSC6) and rs324003968 (SSC15)). Besides this, Pig QTLdb presents SNPs associated with TNB according to the studies of Onteru et al. [31]—SNP rs81452018 SSC15, Wang et al. [32]—SNP rs345476947 SSC6 and Liu et al. [33]—rs rs55618224 SSC6.









*Genes***2020**, *11*, 491


**Table 2.** *Cont.*

*Genes***2020**, *11*, 491






**Table 2.** *Cont.*

*Genes***2020**, *11*, 491






#### **4. Potential Candidate Genes for Litter Traits of Pig**

Accordingly, Pig QTLdb presents 306 SNPs associated with TNB (218 SNPs) and NBA (88 SNPs) in pigs of various breeds. We should note that 12 SNPs out of 306 SNPs provided in PigQTLdb as a Genome map association with TNB and NBA, are presented twice (add file 1). Perhaps this is due to the fact that the connection of SNPs with TNB and NBA was established in one study, for example, as for SNPs rs81348779 on SSC1 and rs81356698 on SSC2 [23]. These SNPs are localized in the intron of the *UBE3A* (SSC1) and *EIF3M* (SSC2) genes. The *UBE3A* gene encodes *ubiquitin protein ligase E3A*, which plays an essential role in the normal development and functioning of the nervous system, and helps regulate the balance between proteostasis synthesis and degradation in the joints between synapses. Human and mouse *UBE3A* is maternal imprinted [34]. However, there is no precise information regarding imprinting of pig *UBE3A*. According to the study performed by Wang et al. [35] *UBE3A* was not imprinted in the skeletal muscle of neonate pigs of Landrace boars and Laiwu sows cross. Further research will probably provide more information to clarify the effects of the *UBE3A* gene and its and its relation to pig fecundity. The *EIF3M* gene (the eukaryotic translation initiation factor 3 subunit M) is a complex translation initiation factor consisting of 13 subunits (*EIF3A-EIF3M*) and which is involved in mRNA modulation [36]. The *EIF3* complex is necessary for the key stages of protein synthesis initiation [37]. Previous studies have shown that *EIF3M* encodes a protein that is critical for mouse embryonic development [38].

In studies performed by Wang et al. [22] a relationship with TNB and NBA was defined for SNPs rs334867206 (intergenic\_variant) and rs319494663 (upstream gene variant *ssc-let-7a-2*) on SSC3 and rs81471172 (intron variant *HECW1*) on SSC18. *Sus scrofa let-7a-2 stem-loop (ssc-let-7a-2*) belongs to miRNAs, a class of small non-coding RNAs (~21 nt) that regulate the mRNAs translation on the post-transcriptional level, mainly by binding their targets with the three prime untranslated region (3'-UTR) [39]. A variety of studies have shown that miRNAs can play a potential regulatory role in porcine ovary, testis and spermatogenesis [40–42].

The *HECW1* gene, also known as *NEDD4-like ubiquitin protein ligase 1* (*NEDL1*), is expressed in human neuronal tissues and enhances p53-mediated apoptotic cell death [43]. Supposedly, it regulates the bone morphogenetic protein signaling pathway during embryonic development and bone remodeling [44]. In the work of Li et al. [25] associations with TNB and NBA are defined for SNP rs342908929, which is localized on SSC6 in the intron of the *ZFYVE9* gene (*zinc finger FYVE-type containing 9 domain*). The protein encoded by *ZFYVE9* is involved in the signaling pathway of *transforming growth factor-beta* (*TGFB*) and directly interacts with *SMAD2* and *SMAD3*, needed for normal follicular development and ovulation [45].

In the studies performed by Ma et al. [28] a relation with TNB and NBA was defined for SNPs rs80882306 (intergenic variant) on SSC7 and rs319494663 (intron variant PARD3) on SSC10. *PARD3* (*PAR-3*) is a *sca*ff*old-like PDZ-containing protein*. *PAR-3* forms a complex with *PAR-6* and *atypical protein kinase C* (*PAR-3-atypical protein kinase C-PAR-6 complex*) and is associated with the establishment of cell polarization [46–48]. McCole [49] argued that mutations in *PARD3* can also influence the recovery of wounds by weakening the response of the epithelial barrier to a damage or inflammation. Concerning its significant role in the regulation of various stages of ovarian development and the control of steroidogenesis in a ripening follicle An et al. [24] studied the *IGFBP-2* gene (SSC15) and defined SNP rs45435330 associated with TNB and NBA [50].

But more interesting are the variants repeated in two independent studies. Thus, rs80927364 (intron variant *DDAH1*) on SSC4, rs81434499 (intergenic variant) on SSC12, rs81447100 (intron variant *CLSTN2*) on SSC13 showed significant associations with TNB and NBA in the studies of He et al. [8] and Ma et al. [28] respectively.

The *DDAH1* gene (dimethylarginine dimethylaminohydrolase 1), along with other *DDAH* genes, is involved in the metabolic control of asymmetric dimethylarginine (ADMA), contributes to the maintenance of vascular homeostasis due to the expansion of blood vessels, suppression of inflammation and inhibiting vascular smooth muscle cells, adhesion and aggregation of platelets [51–53]. Data on

the metabolic control of ADMA by *DDAH* genes and their effect on endothelial cells were obtained in animal studies. Transgenic mice with overexpression of *DDAH1* showed a twofold decrease of ADMA in plasma, associated with a twofold increase of NOS activity in tissue [54]. Conversely, the *DDAH1* Knock Out Mice exhibited increased pulmonary endothelial permeability as a result of increased ADMA, which was prevented by over-expression of *DDAH1* and *DDAH2* in endothelial cells [55].

*CLSTN2* (calsintenin 2) is associated with obesity in mammals, especially in the process of increasing adipocytes in visceral tissues and in subcutaneous fat [56]. Santana et al. [57] defined *CLSTN2* as a candidate gene associated with ultrasound-derived measurements of the rib-eye area, backfat thickness and rumpfat thickness in Nellore cattle. Adipose tissue is an active endocrine organ and proteins involved in forming adipose tissue is increasingly attracting attention as mammalian reproduction markers [58,59].

We can also note SNPs rs81463092 (intergenic variant) and rs81289648 (intergenic variant) defined in the works of He et al. [8] and Ma et al. [28], localized on SSC16 and lying in close proximity to each other (68 kb). The table shows the intervals between neighboring SNPs (add file 1). Hence, we can note a number of closely related SNPs identified in various studies. For example, rs55618224 (3 prime UTR variant) and rs81388947 (intergenic variant) on SSC6 are defined by Liu et al. [33] and Chen et al. [27] and the interval between them is 140 kb. SNP rs336670754 (synonymous variant *ARID1A*) and SNP rs329711941 (intron variant *ZDHHC18*) are located at a distance of about 39 kb from each other, defined on SSC6 by Chen et al., (2019) and Zhang et al. [29] respectively.

On the whole, we can note that SNPs presented in PigQTLdb as associations with TNB and NBA are more localized in genes (intron variant 60%). Wherein that, 15% are intergenic variants and this can be considered as an evidence of significance for intergenic variants in the genetic architecture of reproductive traits (add file 2).

In conclusion, we wanted to estimate the gene-based protein-protein interactions obtained with the studied SNPs being localized. The total number of genes was 127, but only 40 genes formed some pairs or chains (Figure 1). It is interesting that some chains are composed of genes (SNPs) identified in only one study (add file 3). For example, *RBP7*, *LRRK1*, *UBE4B, TRPC5*, *LHFPL1* genes were identified by Suwannasing et al. [26]; *DLC1*, *SEMA3A*, *DPYSL3*, *NRPI* by He et al. [8] similarly. However, the other chains includes genes identified in various studies such as *GRIA4* [8], *CLSTN2* [8], *GRIP1* [28], *GABRA5* [22], *UBE3A* [22], *COPG2* [22], *ANKRD40* [8], *MYO10* [7], *SHISA9* [8].

The presented review shows that most SNPs annotated from genome-wide studies are found in chromosome regions where candidate genes or QTLs have already been identified. Future research will be aimed at annotating sequences and analyzing these data, which can contribute to better understanding the mechanisms of reproductive traits formation.

However, the main problem in summarizing the results is the design of the experiment (consideration of traits, features of the studied populations, etc.). Some of the factors affecting the result are not controlled by the researcher (such as the trait inheritance level, the genetic architecture of the population, linkage disequilibrium, etc.). On the other hand, control over the reference data design (the choice of markers, the model for evaluating effects, etc.) that affects the accuracy of the results is controlled by the researcher. In the analyzed studies the majority of genotyping works were done by using Illumina PorcineSNP60 BeadChip or GeneSeek PorcineSNP80 BeadChip. There are numerous statistical approaches to conducting GWAS. However, the mixed model is more preferable, being implemented in various software packages (GEMMA, ASREML, GenABEL, etc.). There is still no consensus on the best method. Many researchers emphasize that the defined associations and their significance depend on the methodologies and details of data analysis, and we need to develop statistical approaches in order to improve the accuracy of the obtained information [16,19].

‐ **Figure 1.** Protein-protein interactions obtained with the studied SNPs (STRING db).

#### **5. Conclusions**

The results of genome studies show the prospects of this approach to studying the genetic architecture of reproductive indicators in pigs. The material presented in this overview can be used in developing local test systems for a limited number of SNPs to estimate their effect on the own livestock. In addition, gene networks can be built on the basis of the presented SNPs to find potential candidate genes for reproductive signs. On the whole, these results help us to understand the genetic basis of pig reproductive traits and can be used in further studies. Further research in this direction will provide new data that will be a powerful impetus for creating breakthrough breeding technologies and improving the efficiency of breeding production in pig farming.

**Supplementary Materials:** The datasets produced and/or analyzed during the current study are available from the addition files. The following are available online at http://www.mdpi.com/2073-4425/11/5/491/s1, file 1: List of SNPs, file 2: Consequences of SNPs, file 3: List of genes in which SNP is defined.

**Author Contributions:** S.B., L.G., M.K.: Designed survey and wrote the paper, F.B., V.G., A.K., O.K.: Collected and analyzed works. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Russian Scientific Foundation (RSF) within Project No. 19-76-10012.

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

**Ethics:** This article is original and contains unpublished materials. The corresponding author confirms that all of the other authors have read and approved the manuscript and no ethical issues involved.

### **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* **AQP3 Facilitates Proliferation and Adipogenic Di**ff**erentiation of Porcine Intramuscular Adipocytes**

#### **Xiaoyu Wang, Jing Yang, Ying Yao, Xin'E Shi, Gongshe Yang and Xiao Li \***

Key Laboratory of Animal Genetics, Breeding and Reproduction of Shaanxi Province, College of Animal Science and Technology, Northwest A & F University, Yangling 712100, China; wangxy067@163.com (X.W.); yangjing@nwsuaf.edu.cn (J.Y.); yaoying@nwsuaf.edu.cn (Y.Y.); xineshi@163.com (X.S.); gsyang999@hotmail.com (G.Y.)

**\*** Correspondence: nice.lixiao@gmail.com; Tel.: +86-29-870-81531

Received: 19 March 2020; Accepted: 13 April 2020; Published: 22 April 2020

**Abstract:** The meat quality of animal products is closely related to the intramuscular fat content. Aquaglyceroporin (AQP) defines a class of water/glycerol channels that primarily facilitate the passive transport of glycerol and water across biological membranes. In this study, the AQP3 protein of the AQP family was mainly studied in the adipogenic function of intramuscular adipocytes in pigs. Here, we found that AQP3 was increased at both mRNA and protein levels upon adipogenic stimuli in porcine intramuscular adipocytes in vitro. Western blot results showed knockdown of AQP3 by siRNA significantly suppressed the expression of adipogenic genes (PPARγ, aP2, etc.), repressed Akt phosphorylation, as well as reducing lipid accumulation. Furthermore, deletion of AQP3 by siRNA significantly downregulated expression of cell cycle genes (cyclin D, E), and decreased the number of EdU-positive cells as well as cell viability. Collectively, our data indicate that AQP3 is of great importance in both adipogenic differentiation and proliferation in intramuscular adipocytes, providing a potential target for modulating fat infiltration in skeletal muscles.

**Keywords:** AQP3; pig; intramuscular fat; adipogenesis; proliferation

#### **1. Introduction**

Ectopic fat deposition in skeletal muscle has attracted increasing attention in recent decades. In humans, excessive fat accumulation in skeletal muscle always represents muscle weakness, myopathy, and metabolic diseases such as obesity, diabetes, coronary heart disease, etc. [1]. Conversely, in livestock animals (cattle, pig, sheep, etc.), distribution of intramuscular fat (IMF), referred to as marbling by customers, is closely related to meat quality, and a moderate increase in IMF benefits the taste and flavor of meat products [2]. The high content of intramuscular fat can increase the tenderness and flavor of pork [3]. Marbling develops either by an increase in adipocyte number, or adipocyte volume, or both. Thus, the molecular mechanisms underlying the proliferation and differentiation of intramuscular adipocytes deserve further study.

Aquaglyceroporin (AQP) refers to a subgroup of aquaporins that conduct glycerol, water, and other small polar solutes in response to osmotic gradients. Glycerol is a necessary constituent of triglyceride (TG) backbones, and glycerol uptake together with release across the plasma membrane are two key steps in triglyceride synthesis (lipogenesis) and hydrolysis (lipolysis) in adipose, liver, and other metabolic organs, thus AQPs have emerged as key players in adipose biology and the development of obesity [4]. In this context, AQP7, the first identified AQP in adipose tissue, has been emerging as an important player in whole-body metabolism and the progress of obesity and diabetes [5]. In recent studies, expression of AQP3, AQP9, AQP10, and AQP11 has been detected in cultured adipocytes and adipose tissues [6–8]. Studies have shown that AQP3 can promote the transport of cellular glycerol

and has affinity for glycerol [9]. AQP3 has been reported to regulate PPARα by adiponectin in hepatic stellate cells (HSC) [10]. The classical secretion factor, leptin, of adipocytes can improve systemic obesity and fatty liver in mice via AQP3 in ob/ob mice [11]. AQP3 mRNA expression in human adipose tissue was reported in another study [12]. These signs indicate that AQP3 may also have a regulatory role in IMF.

Our RNA-seq screening revealed that AQP3 [13], in addition to the well-known AQP7, was upregulated during adipogenic differentiation in porcine intramuscular preadipocytes (PIPAs), indicating a potential role of AQP3 in porcine IMF deposition. Work on the MDA-MB-231 breast cancer cell line has shown that knockdown of AQP3 modestly decreases water permeability (17%), but markedly decreases glycerol permeability (77%) [14], indicating AQP3 might be more permeable to glycerol. Currently, the most well-characterized role of AQP3 is the promotion of cancer metastasis, for AQP3 is abnormally escalated in various kinds of cancers [15] and knockdown of the AQP3 gene could significantly decrease cell proliferation, and increase cell death or apoptosis in cancer cells [14,16,17]. In addition, AQP3 deficiency can cause proliferation disorders and metabolic inhibition in gastric cancer cells [18]. More interestingly, another study showed that AQP3 in gastric cancer cells caused apoptosis in gastric cancer cells by downregulating cellular glycerol intake and inhibiting downstream adipogenesis [19]. These data encouraged us to explore the effects of AQP3 on adipogenic differentiation, lipid deposition, and proliferation, using porcine intramuscular preadipocytes as a model.

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

#### *2.1. Animal Care*

Piglets in our study were obtained from the experimental plot of Northwest A&F University (Yangling, China). Pigs were reared under standard light and temperature conditions and allowed food and water ad libitum. This project was approved by the Institutional Animal Care and Use Committee of Northwest A&F University.

#### *2.2. Cell Culture*

Porcine intramuscular preadipocytes were isolated from the longissimus dorsi muscle of 3-day-old piglets as previously described [20]. The specific method was as follows: LD (longissimus dorsi) muscles were quickly excised, rinsed twice in sterile pre-cooled phosphate-buffered saline (PBS), and then cut into 1 mm<sup>3</sup> sections. Muscle fragments were incubated in Dulbecco's Modified Eagle's Medium/F12 (DMEM/F12; Hyclone, Logan, UT, USA) containing 0.1% I type collagenase (270 U/mg; Gibco, Carlsbad, CA) for 1.5 h in a 37 ◦C water bath, with continuous shaking. The products were then sequentially passed through a 70 mesh (212 µm) and then a 200 mesh (75 µm) to obtain single cells. The cells were seeded in a dish with DMEM/F12 medium containing 10% fetal bovine serum (Gibco, Grand Island, NY, USA). After 2 h, we changed the medium to keep only adherent cells. In the proliferating stage, the primary preadipocytes were cultured in DMEM/F12 (Gibco, Grand Island, NY, USA) containing 10% FBS (Invitrogen, Carlsbad, CA, USA).

To induce adipogenic differentiation, when cells achieved 100% confluence, a mixture containing 10% FBS, 5 µg/mL (872 nM) insulin, 1 µM dexamethasone (DEX), and 0.5 mM isobutyl methylxanthine (IBMX; Sigma-Aldrich, St Louis, MO) was used to induce adipogenic differentiation. Two days later, a DMEM/F12 medium containing 10% fetal bovine serum (FBS) and 5 µg/mL (872 nM) insulin was changed to maintain differentiation.

#### *2.3. Transfection with siRNA*

Oligonucleotides of AQP3 siRNAs (forward: CCCUUAUCCUCGUGAUGUUTT, reverse: AACAUCACGAGGAUAAGGGTT) and NC (negative control, forward: UUCUCCGAACGUGUCACGUTT, reverse: ACGUGACACGUUCGGAGAATT) were obtained

from GenePharma (Shanghai, China). Transfection was performed with Lipofectamine® RNAiMAX Reagent (ThermoFisher, Waltham, MA, USA) when cells reached proper confluence (40–50% for proliferation test, 70–80% for differentiation test). Negative control and siRNA were added to Opti-MEM (Gibco) and mixed with Lipofectamine® RNAiMAX Reagent. The mixture was allowed to stand for 20 min before being added to the cell culture plate. The culture medium was replaced after 24 h. The final concentration of siRNA or negative control was 50 nM. Cells were changed into fresh growth medium 24 h post-transfection.

#### *2.4. RNA Isolation and RT-qPCR*

Total RNA was purified using Trizol (TaKaRa Bio, Inc., Dalian, China) and was subjected to reverse transcription using the PrimeScriptTM RT reagent Kit (TaKaRa Bio, Inc., Dalian, China). A tissue sample (0.5 g) from a 180-day-old pig was weighed, and high-throughput grinding of the tissue sample in Trizol infiltration extracted total RNA. Each experimental group was subjected to the reverse transcription reaction with 500 ng of RNA. cDNA was subjected to the Multicolor Real-Time PCR detection system (iQ5, Bio-Rad Laboratories, Inc., Hercules, CA, USA) with SYBR Premix Ex TaqTM II kit (TaKaRa Bio, Inc., Dalian, China). The procedure of PCR reaction was pre-denaturation for 5 min, followed by denaturation for 10 s, annealing for 30 s, and extension of 30 s for 35 cycles. Primers targeting AQP3 [21], PPARγ [22] (peroxisome proliferator activated receptor γ), FABP4 [23] (adipocyte fatty-acid binding protein 4), mGPAT [24] (mitochondrial glycerol-3-phosphate acyltransferase), Perilipin 1 [25], cyclin B [26], and β-actin [27] were picked out in previous reports. Primers for SCD (stearoyl-CoA desaturase), CD36, C/EBPα (CCAAT/enhancer binding protein α), ELOVL6 (elongase of long-chain fatty acids family), FASN (fatty acid synthetase), ACACA (acetyl-CoA carboxylase), DGAT2 (diacylglycerol O-acyltransferase 2), cyclin E, and cyclin D were designed online (https://www.ncbi.nlm.nih.gov/tools/primer-blast/) and synthesized by Sangon Biotech (Shanghai, China). Relative expression of each gene was calculated using the 2−∆∆ Ct method, using β-actin as the internal control. Sequences for all primers are shown in Table 1.


**Table 1.** Primer sequences for real-time qPCR.

#### *2.5. Western Blot*

Cells were scraped with RIPA (Radio Immunoprecipitation Assay) buffer (Beyotime, Shanghai, China) and lysates were subjected to SDS-PAGE and transferred to the PVDF (Polyvinylidene fluoride, Millipore, Burlington, MA, USA). Polyacrylamide gels were used to separate and mark proteins of different sizes. The proteins were then transferred to a PVDF membrane. Next, the membrane was soaked in 5% skim milk for 2 h and then incubated with primary antibodies overnight at 4 ◦C. After that, membranes were washed in Tris-buffered saline with Tween 20 and subsequently incubated with horseradish peroxidase-conjugated secondary antibodies. Finally, the stripes of target proteins were visualized by the enhanced chemiluminescent substrate (Millipore, MA) and observed using Gel Doc XR System (Bio-Rad). The densities of the brands were analyzed using Image Lab software (Bio-Rad). Target proteins were probed with primary antibodies (anti-AQP3, ab125219, Abcam, 1:1000; PPARγ, #2435, CST, 1:1000; FASN, sc-20140, Santa Cruz, 1:200; FABP4, sc-18661, Santa Cruz, 1:200; Akt, sc-8312, Santa Cruz, 1:200; p-Akt, sc-7985-R, Santa Cruz, 1:200; cyclin D, sc-753, Santa Cruz, 1:500; cyclin E, sc-247, Santa Cruz, 1:500; β-actin, sc-130656 Santa Cruz, 1:1000).

#### *2.6. Oil Red O Staining*

The well-differentiated cells were washed twice with PBS and fixed with 4% paraformaldehyde for 30 min, and then incubated with 1% filtered Oil Red O solution for 5 min. The stained lipid droplets in the cells were photographed (Nikon TE2000 microscope, Tokyo, Japan). For quantitative analysis, cellular Oil Red O was extracted by isopropanol and optical absorbance was detected at 510 nm.

#### *2.7. EdU Staining*

EdU assay was conducted with a Cell-LightTM EdU (5-ethynyl-2′ -deoxyuridine) Apollo®567 In Vitro Imaging Kit (RiboBio, Guangzhou, China) as per the manufacturer's instructions. Porcine intramuscular preadipocytes in growth medium were incubated with 50 mM EdU for 2 h, and then fixed by paraformaldehyde. Then cells were labeled with Apollo reaction solution, and the nuclei were stained with Hoechst 33,342 (Thermo Fisher Scientific, Waltham, MA, USA). Cells were visualized using a Nikon TE2000 microscope (Nikon, Tokyo, Japan), and the images were processed with Image J software by the National Institutes of Health (NIH).

#### *2.8. CCK-8 Assay*

Porcine intramuscular preadipocytes were seeded on a 96-well plate at a density of 1 × 10<sup>3</sup> per well. Some 24 h later, 10% CCK-8 solution (Vazyme, Nanjing, China) was added, and after 4 h incubation, the absorbance was measured at 490 nm.

#### *2.9. Statistical Analysis*

All experiments were carried out in triplicate and the results were analyzed by one-way analysis of variance (ANOVA) using SPSS 18 software (SPSS Inc., Chicago, IL, USA). *P* < 0.05 was set as statistical significance. Data were presented as mean ± standard error (SE).

#### **3. Results**

#### *3.1. AQP3 Is Upregulated during Adipogenesis*

In order to explore the expression pattern of AQP3 gene in pig adipose tissue and PIPAs we selected 180-day-old pig tissue to test the expression of AQP3. RT-qPCR results showed that AQP3 is highly expressed in adipose tissue (Figure 1A). In subcutaneous adipose tissue of pigs of different ages, the expression of AQP3 reached the highest at 30 days of age, and then began to decline (Figure 1B).

. **Figure 1.** AQP3 expression pattern in pigs in vivo by RT-qPCR. (**A**) AQP3 pig tissue expression profile. (**B**) AQP3 expression during pig growth by RT-qPCR. β-actin was used as internal reference gene. *n* = 3.

β

t γ In in vitro cell culture experiments, transcripts of AQP3 in PIPAs were rapidly increased upon adipogenic stimuli, reached a peak at 4 d post-differentiation, and then gradually decreased (Figure 2A). The expression of AQP3 proteins showed the same pattern (Figure 2B). For reference, the expression of PPARγ (Figure 2C), FASN (Figure 2D), and aP2 (Figure 2E) throughout the adipogenic process was profiled to represent the efficient differentiation of PIPAs in vitro. The data indicated a promising role of AQP3 in adipogenesis.

γ **Figure 2.** Expression pattern of porcine AQP3 in adipogenesis. The expression of AQP3 during adipogenesis was detected by RT-qPCR (**A**) and Western blot (**B**). The expression of PPARγ (**C**), FASN (**D**), and aP2 (**E**) was detected by RT-qPCR. β-actin was used as internal reference gene. *n* = 3.

β

γ

γ

γ

#### *3.2. Knockdown of AQP3 Blunts Adipogenesis*

In view of the rising trend of AQP3 in adipocyte differentiation, siRNAs were employed to explore the role of AQP3 on adipogenic differentiation. Three siRNAs were designed, and only siRNA-1 showed >70% knockdown efficiency 24 h post-transfection (Figure 3A), and still significantly reduced AQP3 expression 48 h, 4 d, and 8 d post-differentiation (Figure 3B). Thus, siRNA-1 was used in the following study. RT-qPCR results showed that AQP3 siRNA significantly inhibited the expression of adipogenic markers, such as PPARγ, aP2, ACACA, SCD, DGAT2, mGPAT, ELOVL6, and FASN 4d post-differentiation (Figure 3C), and the genes (except PPARγ) detected above were still significantly downregulated 8 d post-differentiation (Figure 3D). Western blot results presented the expression of PPARγ, aP2, and FASN at protein levels and phosphorylated Akt (Figure 3E), and the gray level analysis showed that these proteins were significantly decreased (Figure 3F). Oil Red O staining showed that AQP3 siRNA significantly repressed triglyceride accumulation in intramuscular adipocytes (Figure 3G,H). These data indicated that AQP3 was essential for adipogenesis and lipid accumulation in PIPAs.

**Figure 3.** *Cont*.

;

β

γ

β γ **Figure 3.** AQP3 silence repressed adipogenic differentiation in porcine intramuscular adipocytes. Cells were transfected with 3 candidate siRNAs targeting AQP3 when reaching 70–80% confluence, and only siRNA-1 could decrease AQP3 transcripts by 70% (**A**) 24 h post-transfection, and it also significantly repressed AQP3 expression 48 h, 4 d, and 8 d post-differentiation (**B**). Expression of adipogenic and lipogenic genes 4 d and 8 d post-differentiation was detected by RT-qPCR, using β-actin as reference gene (**C**). Expression of adipogenesis-related genes 8 d post-differentiation was detected by Western blot (**D**). Western blot images (**E**) and gray analysis statistics (**F**) of PPARγ, aP2, FASN, and Akt in PIPAs. Lipid accumulation was tested by Oil Red O staining (**G**) and quantified by isopropanol extraction (**H**). *n* = 3; \* *p* < 0.05, \*\* *p* < 0.01.

;

#### *3.3. AQP3 Deletion Inhibits Proliferation*

AQP3 siRNA could significantly downregulate AQP3 mRNA expression in the proliferating porcine intramuscular preadipocytes 24 h post-transfection (Figure 4A), indicating that this siRNA can be used to inhibit the expression of AQP3 during the proliferation phase. The results of RT-qPCR showed that siRNA repressed the expression of cyclin B, cyclin D, cyclin E, and CDK4 mRNAs 48 h post-transfection (Figure 4B). Consistently, cyclin B, cyclin D, and proliferating cell nuclear antigen (PCNA) were significantly repressed by AQP3 siRNA at the protein level too (Figure 4C,D). Additionally, transfection of AQP3 siRNA significantly reduced the ratio of EdU-positive cells (Figure 4E,F) and cell viability (Figure 4G). In summary, the above results show that during the proliferation phase, AQP3 exerts this effect to promote cell proliferation.

**Figure 4.** *Cont*.

′ ;

**Figure 4.** Knockdown of AQP3 inhibited the proliferation of porcine intramuscular adipocytes. Cells were transfected with AQP3 siRNA at 40–50% confluence, and the interference effect was over 70% 24 h post-transfection (**A**). Cell cycle genes were analyzed by RT-qPCR (**B**) and Western blot (**C**,**D**). EdU staining was captured (**E**) and EdU-positive cells were counted to monitor the proliferation of cells (**F**). CCK-8 was adopted to check cell viability (**G**). *n* = 3. CDK4, cyclin-dependent kinase 4; PCNA, proliferating cell nuclear antigen; DAPI, 4′ ,6-diamidino-2-phenylindole; \* *p* < 0.05, \*\* *p* < 0.01.

′ ;

#### **4. Discussion**

The relationship between AQP3 and adipose tissues or cells has been ignored for a long time, and its expression in adipose tissues or adipocytes remains ambiguous. In pioneering work, AQP3 mRNA was undetectable in adipose tissues of Meishan pigs by semi-quantitative RT-PCR [28]. Meanwhile, there were other studies that did not support the presence of AQP3 in mouse [29,30] or human adipose tissue [31]. However, mRNAs and proteins of AQP3 were later detected in human stroma vascular fraction of omental, subcutaneous adipocyte tissue, and also in freshly isolated adipocytes [6]. Additionally, the expression of AQP3 was confirmed in murine 3T3-L1 cell line [32,33]. The inconsistent reports might be due to the relatively lower expression levels of AQP3 in mature adipose tissues [6,12].

In the present study, RT-qPCR and Western blot uncovered the upregulation of AQP3 during the process of adipogenic differentiation, indicating that AQP3 may also play a regulatory role in intramuscular adipocytes. Furthermore, AQP3 knockdown by siRNA leads to reduced expression of adipogenic and lipogenic genes and defects of TG storage in porcine intramuscular preadipocytes. After AQP3 was silenced, the expression of the classic adipogenic factor PPARγ was significantly inhibited. PPARγ is an indispensable transcription factor for adipocyte differentiation [34], and a previous study has shown that AQP3 is the target of PPARγ in murine adipose cell line [35] and hepatic stellate cells [36]. The level of phosphorylation of AKT was also decreased, which indicates that the downstream signal of the classic insulin signaling pathway is weakened. At the same time, the expression levels of other adipogenic marker genes aP2, ACACA, etc. were also decreased, indicating that the silencing of AQP3 inhibits the adipogenic differentiation of porcine intramuscular preadipocytes from the overall gene expression level. A previous study has also shown that AQP3 is more permeable to glycerol compared with water [14], and AQP3 was increased in LPS-induced adipogenesis [32] and triglyceride sedimentation [37]. A combination of our and others' work supports a pivotal role of AQP3 in lipid accumulation in adipocytes.

Besides, AQP3 also has the potential to regulate cell proliferation. Accumulated studies have shown that AQPs are involved in tumor metastasis [15]. AQP3 and AQP5 can be used as new markers for breast cancer [38]. Another study showed that overexpression of AQP3 in mammalian cells can promote cell proliferation efficiency and cell cycle transition [39]. In our study, AQP3 siRNA weakened the proliferation ability of porcine intramuscular preadipocytes, which was reflected in the downregulation of genes such as cyclin B and cyclin D, and a decrease in the number of EDU<sup>+</sup> cells. This shows that AQP3 can not only regulate cell proliferation in cancer cells, but also in intramuscular preadipocytes. For IMF, the proliferation and differentiation of adipocytes are two very important fat deposition processes [40]. AQP3 shows regulatory capacity in both proliferation and differentiation, which further illustrates that AQP3 is a key gene for adipocytes.

In the AQPs family, AQP7 can promote the loss of glycerol from adipocytes and inhibit the accumulation of fat, while AQP9 is responsible for the uptake of glycerin [41]. Additionally, the weakening of AQP5 can inhibit the late adipogenesis of 3T3-L1 cells [42]. Compared with other AQP genes, the present study proves that AQP3 can regulate fat deposition in two ways, namely adipocyte proliferation and differentiation, revealing that AQP3 is likely to be a new key factor in regulating fat deposition, which supplements the regulation of adipocytes by AQPs.

#### **5. Conclusions**

Collectively, our work has identified AQP3 as a novel and essential modulator in proliferation, differentiation, and lipid accumulation in intramuscular adipocytes, providing a new theoretical basis for the regulation of skeletal muscle ectopic fat deposition. This conclusion may be due to the involvement of AQP3 in the transport of glycerol, but further research is needed.

**Author Contributions:** Wrote this paper and analyzed the data: X.W.; performed the cell culture and cell-based analysis: J.Y. and Y.Y.; gave critical suggestions about the experimental design and manuscript preparation: X.S. and G.Y.; designed the experiment and revised the manuscript: X.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by Major Projects for Genetically Modified Organisms Breeding (2016ZX08006003) and National Natural Science Foundation (31501925).

**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*
