**Current Evidence on the E**ffi**cacy of Gluten-Free Diets in Multiple Sclerosis, Psoriasis, Type 1 Diabetes and Autoimmune Thyroid Diseases**

**Moschoula Passali 1,2, Knud Josefsen 3, Jette Lautrup Frederiksen 1,2,\* and Julie Christine Antvorskov <sup>3</sup>**


Received: 15 July 2020; Accepted: 30 July 2020; Published: 1 August 2020

**Abstract:** In this review, we summarize the clinical data addressing a potential role for gluten in multiple sclerosis (MS), psoriasis, type 1 diabetes (T1D) and autoimmune thyroid diseases (ATDs). Furthermore, data on the prevalence of celiac disease (CD) and gluten-related antibodies in the above patient groups are presented. Adequately powered and properly controlled intervention trials investigating the effects of a gluten-free diet (GFD) in non-celiac patients with MS, psoriasis, T1D or ATDs are lacking. Only one clinical trial has studied the effects of a GFD among patients with MS. The trial found significant results, but it is subject to major methodological limitations. A few publications have found beneficial effects of a GFD in a subgroup of patients with psoriasis that were seropositive for anti-gliadin or deamidated gliadin antibodies, but no effects were seen among seronegative patients. Studies on the role of gluten in T1D are contradictive, however, it seems likely that a GFD may contribute to normalizing metabolic control without affecting levels of islet autoantibodies. Lastly, the effects of a GFD in non-celiac patients with ATDs have not been studied yet, but some publications report that thyroid-related antibodies respond to a GFD in patients with concomitant CD and ATDs. Overall, there is currently not enough evidence to recommend a GFD to non-celiac patients with MS, psoriasis, ATDs or T1D.

**Keywords:** gluten; gluten-free diet; gliadin; autoimmunity; neurology; multiple sclerosis; psoriasis; autoimmune thyroid disease; type 1 diabetes; celiac disease

#### **1. Introduction**

Wheat is a major component of Western diets, however, abstaining from gluten is becoming a popular trend [1]. Adhering to a lifelong gluten-free diet (GFD) is the current treatment for celiac disease (CD)—an immune-mediated small intestinal enteropathy triggered by the ingestion of gluten [2]. It has been hypothesized that gluten may contribute to deteriorating the course of immune-mediated disorders [3–5]. According to a U.S. national survey, a GFD was the most common special diet to be used by patients with psoriasis [6]. Similarly, an American dietary survey found that 5.6% of the surveyed patients with multiple sclerosis (MS) reported adhering to a GFD [7], whereas in an Australian survey, a GFD was adopted by 16.4% of the included patients with MS [8]. Type 1 diabetes (T1D) and autoimmune thyroid diseases (ATDs) affect the endocrine system. The contribution of dietary factors to the pathogenesis of autoimmune endocrine disorders is currently an active research

area. This review summarizes the currently available clinical data on a potential involvement of gluten in MS, psoriasis, T1D and ATDs.

#### **2. Gluten**

Gluten proteins have long been of interest to the food industry due to their high impact on the baking quality of wheat flours [9,10]. From a chemical perspective, gluten has been defined as the proteinaceous mass that remains when wheat dough is washed with water and consists primarily of the prolamin and glutelin fractions of the storage proteins of wheat [11,12]. The terms prolamin and glutelin originate from the classification of grain proteins into four fractions according to their solubility properties (Osborne fractions) [13]. Prolamins are insoluble in water but soluble in alcohol, whereas glutelins are insoluble in both water and alcohol [14]. The terms gliadin and glutenin account for the prolamin and glutelin fractions of wheat, whereas the terms secalin, hordein and avenin describe the prolamin fraction of rye, barley and oats, respectively [14]. Likewise, the glutelin fractions of rye and barley are commonly described as secalinin and hordenin, however, similar terminology does not apply for oat glutelins [12]. Codex Alimentarius has defined gluten as "a protein fraction from wheat, rye, barley, oats or their crossbred varieties and derivatives thereof, to which some persons are intolerant and that is insoluble in water and 0.5M NaCl" [15]. As a result, gluten is nowadays considered to be a common term for the prolamin and glutelin fractions of wheat, rye, barley and, in some cases, oats.

Gluten proteins contain repetitive sequence sections that are rich in the amino acids proline and glutamine [14,16]. Such sections cannot be fully degraded by the human gastrointestinal enzymes [10,13], resulting in the presence of relatively long gluten peptides in the small intestine. In patients with CD, such gluten peptides trigger an inflammatory reaction, however, their presence in the small intestine of most healthy individuals is believed to be rather unproblematic. In vitro studies using caco-2 cell lines [17,18] as well as ex vivo studies on human biopsy explants from both CD patients and healthy controls (HCs) [18,19], suggest that exposure to gliadin disrupts the integrity of the intestinal epithelium. The effect of gliadin on intestinal permeability is believed to be mediated through the secretion of the protein zonulin [20]. Zonulin has been identified as prehaptoglobin-2 [21] and serum zonulin is often used as a marker of intestinal permeability. Levels of zonulin have been found to be elevated in autoimmune diseases [22–25], however, widely used ELISA kits cross-react with proteins, such as properdin and complement C3 [26,27], which shows why caution should be practiced when interpreting data on this topic.

#### **3. Multiple Sclerosis**

Multiple sclerosis (MS) is an autoimmune, yet incurable, disease of the central nervous system [28] and one of the leading causes of disability among young adults. A recent publication reports that 31% (10/32) of websites providing MS-specific dietary advice recommend patients with MS to abstain from "grains (gluten)" [29]. Despite a high interest in the use of dietary modifications to ameliorate the course of the disease [30], MS-specific, evidence-based dietary guidelines have not been developed yet.

#### *3.1. Gluten-Free Interventions in Multiple Sclerosis*

The effects of a GFD among patients with MS have only been investigated by a single open label, non-randomized, controlled trial. Thirty-six patients, who followed a GFD for a median of 4.5 years (mean 5.3 ± 1.6), were compared with 36 patients who followed a regular diet [31]. At the end of the study, the group on the GFD had significantly lower disability measured by the expanded disability status scale (EDSS) (1.5 ± 1.4 vs. 2.1 ± 1.5, *p* = 0.001, baseline EDSS was 1.7 for both groups) and significantly lower activity on magnetic resonance imaging (MRI) (28% vs. 67%, *p* = 0.001) compared to the group on a regular diet [31]. There was no effect on annual relapse rate. Unfortunately, this study was subject to important limitations. Group allocation was performed by instructing all 72 patients to follow a GFD for the first six months of the study, whereafter non-compliant patients were asked to

resume a regular gluten-containing diet. In addition, eight study participants were diagnosed with CD and they all remained in the GFD group, further supporting the idea that the method used for group allocation was suboptimal. Apart from above described methodological issues, major inconsistencies, including the use of the word "randomised" in the title of the study, reduce our capacity to trust this publication [31].

Eliminating gluten from the diet is also part of the "The Wahls Protocol", a multimodal lifestyle intervention including, among others, adherence to a modified paleolithic diet. Clinical studies have illustrated that "The Wahls Protocol" can contribute to improving primarily self-reported outcomes, such as mood, fatigue and quality of life among patients with relapsing remitting MS [32] and progressive MS [33–35]. The risk of placebo and/or nocebo effects should not be neglected when evaluating the results of lifestyle interventions. Furthermore, due to the multimodal nature of the interventions, it is not possible to quantify the effects of eliminating gluten from the diet. The lack of disease-specific endpoints is a major limitation of these studies. Nevertheless, these publications highlight that lifestyle modifications can contribute to improving the quality of life of patients with MS. This is of utmost importance for patients with the progressive forms of MS, as highly effective treatments for these patients are still lacking [36].

#### *3.2. Prevalence of Celiac Disease and Gluten-Related Serology in Multiple Sclerosis*

Several publications have reported the prevalence of gluten-related antibodies among patients with MS. Among six studies estimating the prevalence of seropositivity for anti-gliadin (AGA) immunoglobulins (Ig) in patients with MS [37–42], only one study found a significantly higher prevalence of IgG-AGA among patients with MS (7/98) compared to HCs (2/140) (*p* = 0.03) [42]. However, when investigating whether patients with MS have elevated mean values of IgA-AGA or IgG-AGA compared to HCs, the results are highly contradictive [38,42–44]. We can therefore not exclude that patients with MS may have slightly elevated AGA titers compared to HCs, however, this is still far from sufficiently different for diagnostic use.

Data from twelve studies [37–48] estimating the prevalence of seropositivity for IgA tissue transglutaminase (tTG) in patients with MS do not support an increased prevalence of CD among patients with MS, whereas a single study found higher mean values of IgA-tTG and IgG-tTG among 30 patients with MS compared to 25 HCs [49]. So far, only one publication supports an association between CD and MS by reporting the prevalence of CD to be 11% in a cohort of 72 patients with relapsing remitting MS and 32% (23/126) among their first-degree relatives [50]. According to the last-mentioned study, the diagnosis of MS was made at a younger age among celiac (35 +/− 7 years) compared to non-celiac (44 +/− 10 years) patients (*p* < 0.05) [50]. For an overview of studies measuring other gluten- and celiac-related antibodies among patients with MS, the reader is referred to Thomsen et al. (2019) [51].

The most powerful studies investigating a potential association between CD and MS are two Danish population-based studies [52,53] and a Swedish case–control study including 14,371 CD patients and 70,096 reference individuals [54], however, none of them found any association. The first Danish study investigated the comorbidity of 31 autoimmune diseases and calculated an odds ratio (OR) of 1.0 for CD and MS [52]. The second Danish study investigated the prevalence of autoimmune comorbidities among patients with CD and failed to find an increased prevalence of MS among patients with CD. According to the Swedish case–control study, the presence of CD did not increase the risk of subsequent MS diagnosis (hazard ratio (HR) = 0.9; 95% confidence interval (CI) = (0.3–2.3)) [54]. Lastly, two French studies estimated the prevalence of MS among patients with CD to be 0.11% (1/924) [55] and 0.14% (1/741) [56]. This is similar to the crude estimate of MS in the French population, which is 0.15% [57].

#### **4. Psoriasis**

Psoriasis is a chronic autoimmune skin disease characterized by the development of erythematous scaly lesions. Psoriasis vulgaris, also known as plaque psoriasis, is the most common type of psoriasis, but several other types of psoriasis also exist [58]. The Medical Board of the National Psoriasis Foundation conducted a systematic review in 2018 with the aim of developing nutritional recommendations for patients with psoriasis or psoriatic arthritis [59]. The board states "We weakly recommend a gluten-free diet only in patients who test positive for serologic markers of gluten sensitivity" [59]. The popularity of the GFD among patients with psoriasis is highlighted in a U.S. study from 2017, in which 38% of the responding patients (*n* = 1206) reported avoiding gluten and 53.4% (247/459) of them reported to have experienced an improvement or clearance of their disease as a result of the GFD [6].

#### *4.1. Intake of Gluten and Risk of Psoriasis*

Using data from the Nurses´ Health Study II, a publication examined whether higher intakes of gluten were associated with a higher risk of future psoriasis, psoriatic arthritis and atopic dermatitis [60]. When comparing the highest and lowest gluten intake quintiles, the multivariate HRs were 1.15 (95%CI = (0.98–1.36)), 1.12 (95%CI = (0.78–1.62)) and 0.91 (95%CI = (0.66–1.25)) for psoriasis, psoriatic arthritis and atopic dermatitis, respectively. No dose–response relationship was observed, but the fact that the effect of a strictly GFD was not investigated is a minor limitation of this study.

#### *4.2. Gluten-Free Interventions in Psoriasis*

The potential role of gluten in psoriasis has been addressed in several publications from Michaëlsson and his colleagues. In 2000, they published a study illustrating clinical improvement in 73% (22/30) of patients who adhered to a GFD for three months (reduction of psoriasis area and severity index (PASI) score from 5.5 ± 4.5 to 3.6 ± 3.0 (*p* = 0.001)) [61]. All patients were positive for IgA-AGA or IgG-AGA and no clinical improvement was observed among six seronegative patients who also adhered to a GFD. The study was originally designed as a cross-over trial and, after three months on a GFD, the participants had to reintroduce gluten to their diet for three months. However, the last part of the study was discontinued as 60% (18/30) of the AGA-positive patients, but none of the seronegative patients, required increased treatment due to a worsening of their skin lesions after the reintroduction of dietary gluten [61]. Immunohistochemical analyses of skin biopsies from 19 of the above seropositive patients were later published in a separate publication that revealed a reduction in Ki67 positive cells in the involved dermis after the GFD [62]. Moreover, a higher expression of tTG was found in involved, compared to uninvolved, dermis (5.06 ± 3.80% vs. 0.67 ± 0.54%, *n* = 13, *p* = 0.0002) and the GFD resulted in a drop in tTG expression in the dermis by 50% [62].

Similarly, in 2007 Michaëlsson et al. presented results from 16 cases of palmoplantar pustulosis, who adhered to a GFD [63]. AGA-seropositive patients who strictly adhered to the GFD (*n* = 9) experienced great improvements or even the clearance of their lesions. Improvements were only seen in two out of four patients with lower compliance to the GFD and none of the seronegative patients (*n* = 3) [63].

According to a more recent publication by Kolchak et al. in 2018, a one-year gluten-free intervention resulted in a 56% and 36% improvement in the PASI score in patients with very high (>30 U/ml, *n* = 5) and high (11.5–30.0 U/ml, *n* = 8) levels of IgA against gliadin peptides (not clear whether native or deamidated gliadin), respectively [64]. As no other antibodies were measured and biopsies were not performed, it is not known whether some of the included patients suffered from CD. The effects of a GFD in patients with concomitant psoriasis and CD have been explored in an Italian multicenter study [65]. At a three-month follow-up, patients (*n* = 9) experienced major improvements in their PASI scores (two by at least 50%, five by at least 75%, total clearance in one patient and one drop-out). A single patient had worsened by the six-month follow-up, whereas most patients maintained their

clinical improvement (*n* = 5) and two patients further improved [65]. Overall, evidence suggests that psoriasis patients with gluten-related antibodies may benefit from a GFD, however, larger trials are still lacking.

#### *4.3. Gluten-Related Serology in Psoriasis*

According to a meta-analysis from 2014, both the prevalence of seropositivity for IgA-AGA, as well as mean values of IgA-AGA, are higher among patients with psoriasis compared to HCs [66]. To illustrate the heterogeneity among studies, an overview of identified case–control publications is provided in Table 1. Additional studies have described the prevalence of AGA seropositivity among patients with psoriasis [67,68] and palmoplantar pustulosis [63] without including a relevant control group. Moreover, most studies [69–72] have found significantly higher concentrations of IgA-AGA among patients with psoriasis compared to HCs, whereas one study [73] did not. Regarding IgG-AGA, only one [71] out of four studies [70,72,73] found higher concentrations in patients with psoriasis compared to HCs. Using a different approach, no difference was found in the proliferative response of peripheral blood mononuclear cells from patients with psoriasis (*n* = 37) and HCs (*n* = 37) after stimulation with wheat peptides, however, the five highest responses against peptide p62-75 were observed among patients with psoriasis [74].

**Table 1.** Case–control studies estimating the prevalence of IgA-AGA and IgG-AGA in patients with psoriasis and HCs. Results are presented as "number of seropositive individuals"/"number of individuals tested". HCs: healthy controls, Ig: immunoglobulin, AGA: antigliadin antibody, NA: not available, NS: not significant.


<sup>1</sup> Possibly measures of antibodies against deamidated gliadin peptide.

To investigate whether gluten-related antibodies correlate with disease activity in psoriasis, a study screened 130 patients for IgG-AGA, IgA-AGA and IgA-tTG and identified 21 patients (16.2%) who were positive for at least one of the antibodies [80]. Psoralen and ultraviolet A (PUVA) phototherapy (57% vs. 30%, *p* = 0.03) and systemic therapy (48% vs. 22%, *p* = 0.04) was currently given or had previously been given to a higher percentage of seropositive compared to seronegative patients. There was no difference for ultraviolet B (UVB) phototherapy and the presence of arthritis or arthralgia [80]. A similar but smaller study (*n* = 41) found a significant relationship between seropositivity for IgA-AGA and disease duration (*p* < 0.001), however, being seropositive was not related to PASI scores [69]. In contrast, in a study with 120 patients with psoriasis (eight seropositive for IgA-AGA/five seropositive for IgG-AGA), severe disease at the reported time or past treatment for high disease severity was not associated with seropositivity for AGA [73].

According to a publication from 2020 that tested for antibodies against an array of 75 antigens, IgG4 antigliadin antibodies were the only antibodies to be elevated in the sera of 12 patients with severe psoriasis (PASI > 30) [81]. IgG4 antigliadin antibodies were not present in sera from 12 HCs. Later validation, using a cohort with 73 psoriasis patients and 75 HCs, resulted in an area under the curve of 0.98 (*p* < 0.001) in the receiver operating characteristic (ROC) analysis, suggesting that IgG4 antigliadin antibodies could potentially function as a diagnostic biomarker for psoriasis. For a subgroup of patients with the highest levels of anti-gliadin IgG4, there was a significant correlation between antibody levels and PASI scores (r = 0.65, *p* < 0.001) [81].

With regard to the prevalence of IgA-tTG antibodies in psoriasis, we have identified five cross-sectional cohort studies [63,68,80,82,83] and ten case–control studies [65,69,71,75,76,84–88] (previously mentioned in Table 2). In addition, a study of 67 patients with psoriasis and 85 HCs found significantly elevated mean values of IgA-tTG in patients with psoriasis (0.943 ± 1.131 vs. 0.852 ± 0.576, *p* < 0.05) [71]. Many case–control studies fail to reveal a significant difference between groups, however, it must be stressed that the majority of studies on the topic are underpowered, considering the fact that the global seroprevalence of CD has been estimated to be 1.4% [89].

**Table 2.** Case–control studies estimating the prevalence of IgA-tTG in patients with psoriasis and HCs. Results are presented as "number of seropositive individuals"/"number of individuals tested". HCs: healthy controls, IgA-tTG: class A immunoglobulins against tissue transglutaminase, NA: not available, NS: not significant.


#### *4.4. Comorbidity between Celiac Disease and Psoriasis*

Results from 18 publications included in a systematic review and meta-analysis from 2019 are summarized below [90]. Out of two studies investigating the incidence of CD among patients with psoriasis, only one found a statistically significant increased risk (HR = 1.9, 95%CI = (1.6–2.2) [91] and HR = 1.20, 95%CI = (0.91–1.59)) [92]. Similarly, two studies estimated the incidence of psoriasis among patients with CD, but in this case, both studies found significant results (HR = 1.72, 95%CI = (1.54–1.92) [93] and HR = 1.9, 95%CI = (1.5–2.3) [91]). With regard to the prevalence of CD in patients with psoriasis or psoriatic arthritis, significantly increased ORs were found in five [65,92,94–96] out of nine [86,87,97,98] studies (meta-analysis: OR = 2.16, 95%CI = (1.74–2.69) [90]). Likewise, the prevalence of psoriasis among patients with CD was found to be increased in four [53,93,99,100] out of eight [101–104] publications (meta-analysis: OR = 1.8, 95%CI = (1.36–2.38) [90]). The two-way meta-analysis concluded that clinicians should be aware of the significant association between CD and psoriasis [90].

#### **5. Type 1 Diabetes**

T1D is a chronic, autoimmune disease characterized by the destruction of the insulin-producing beta cells in the pancreas. T1D is often diagnosed in childhood and results in a lifelong need for exogenous insulin. Several animal studies support the potential involvement of gluten in the pathogenesis of T1D [105] and have previously been summarized by Antvorskov et al. [105] and Haupt-Jorgensen et al. [106].

#### *5.1. Exposure to Gluten during Early Life and Risk of Type 1 Diabetes*

Recent mother and child cohort studies suggest that exposure to gluten during early life may affect the risk of developing T1D [107,108], whereas earlier studies among predisposed individuals did not reveal such an association [109,110]. According to a Danish study from 2018 [107], offspring from mothers with the highest intake of gluten had a twofold higher risk of developing T1D compared to offspring from mothers with the lowest intake of gluten during pregnancy (adjusted HR = 2.00, 95%CI = (1.02–4.00)). A dose–response relationship was demonstrated, however, only the difference between the groups with the highest and lowest intakes of gluten reached statistical significance. These results were not replicated in a similar Norwegian study from 2020 [108] but, this time, the intake of gluten by the offspring themselves was associated with a higher risk of T1D (adjusted HR = 1.46, 95%CI = (1.06–2.01), *p* = 0.02).

Similarly, publications have explored whether and how infant feeding patterns could affect the risk of T1D [105]. Data from the Diabetes Autoimmunity Study in the Young (DAISY) support that introduction of cereals between the age of 4-6 months leads to the lowest risk of islet autoimmunity (<4 months: HR = 4.32, 95%CI = (2.0–9.35), >6 months: HR = 5.36, 95%CI = (2.08–13.8)) [111]. Likewise, the late (≥7 months) introduction of gluten-containing porridge has been found to be a risk factor for the development of β-cell autoantibodies [112], whereas Ludvigsson [113] did not find an association between the time of the introduction of gluten and levels of islet autoantibodies. Results from the BABYDIAB study support the idea that the introduction of gluten-containing foods at or before the age of three months increases the risk of islet autoimmunity (HR = 5.2, 95%CI = (1.7–15.5), *p* = 0.003), however, the late (>6 months) introduction of gluten-containing foods was not associated with increased risk of islet autoimmunity [114]. Lastly, the BABYDIET study—a pilot study in which 150 infants at high risk of T1D where randomized to either control (6 months) or late (12 months) introduction of gluten—did not find any difference in islet autoimmunity at three years in the per protocol analysis (compliance = 70%) [115].

#### *5.2. Gluten-Free Interventions in Type 1 Diabetes*

Two studies have investigated whether a GFD could have a protective effect among children with a high risk of developing T1D. In the first study, 17 first-degree relatives of T1D patients with at least two β-cell autoantibodies were included in a cross-over trial consisting of six months on a GFD followed by six months on a gluten-containing diet [116]. Glucose tolerance tests revealed an improved acute insulin response after the GFD (*p* = 0.004) and a non-significant deterioration after the reintroduction of dietary gluten (*p* = 0.07). The results were similar for insulin sensitivity measured by the homeostasis model of insulin resistance (HOMA-IR), however, this time, a non-significant increase after the GFD was followed by a significant decrease (*p* < 0.005) after the gluten-containing diet. An effect on the levels of autoantibodies was neither observed in this study [116] nor in a similar study with a longer gluten-free intervention of 12 months (*n* = 7) [117]. A five-year follow-up to the latter study suggests that the 12 months on a GFD did not affect the risk of progressing to T1D [117].

In 2012, a case report suggested that a GFD introduced 2–3 weeks after the diagnosis of T1D may have prolonged remission in a five-year-old boy without CD [118]. Both his HbA1c and his fasting blood glucose stabilized without insulin therapy and, twenty months after diagnosis, he still remained without the need for daily insulin therapy [118]. This case led to the performance of a Danish pilot study evaluating the effects of a one-year gluten-free intervention among 15 children with newly diagnosed T1D [119]. Compared to two previous reference cohorts, the children on the GFD had a 21% lower HbA1c and a higher prevalence of partial remission (insulin dose-adjusted A1c (IDAA1c) ≤ 9), but no difference was seen in stimulated C peptide [119].

Prolonged partial remission in response to a GFD was also illustrated in a study that was methodologically stronger due to the inclusion of a control group which remained on a standard gluten-containing diet (*n* = 19) during the study time [120]. The trial was not randomized and 20 out of 26 children completed the one-year gluten-free intervention with satisfactory compliance. The GFD was introduced within a median of 38 days from the onset of T1D. At follow-up, the children adhering to the GFD had a lower IDAA1c (by 1.37; *p* = 0.01), a lower mean HbA1c (by 0.7% (7.8 mmol/mol); *p* = 0.02) and there was a tendency towards a lower insulin dose (by 0.15 U/kg/day; *p* = 0.07) compared to the control group [120]. Last, but not least, several studies have investigated whether a GFD affects metabolic control in individuals with both CD and T1D [121–142] (presented in Table A1 in Appendix A).

#### *5.3. Prevalence of Celiac Disease and Gluten-Related Serology in Type 1 Diabetes*

A systematic review and meta-analysis published in 2014 calculated the prevalence of biopsy-confirmed CD among patients with T1D to be 6.0% (95%CI = (5.0–6.9%)) [143]. Similarly, a systematic review and meta-analysis from 2019 reports the weighted prevalence of CD and any gluten-related antibodies among patients with T1D to be 4.7% (95%CI = (4.0–5.5)) and 10.2% (95%CI = (8.4–12.7)), respectively [144]. Among gluten-related antibodies, the highest weighted prevalence was estimated for IgG-AGA as 12.7% (95%CI = (6.1–21.0)) [144]. Equally relevant, a Swedish population cohort study has estimated the HR of subsequent T1D before the age of 20 years to be 2.4 (95%CI = (1.9–3.0), *p* < 0.001) among patients with CD [145].

Although the association between CD and T1D is well supported, the heterogeneity among studies is large. A better understanding of the factors that contribute to this variation may therefore be relevant. With regard to measurements of gluten-related antibodies, technical differences in the analytical assays being used may hinder direct comparisons among publications highlighting the importance of including a healthy control group in all studies. This is especially relevant for AGA due to their lower specificity for CD and the fact that biological factors may contribute to their variation within healthy populations.

The meta-analysis by Elfström et al. [143] revealed that CD was less frequent in adults (2.7%, 95%CI = (2.1–3.3%)) compared to children (6.2%, 95%CI = (6.1–6.3%)) with T1D (*p* < 0.001). Tiberti et al. [146] on the contrary, found a significantly higher prevalence of gluten-related antibodies among patients with a high (> 15 years) compared to a low (5–15 years) duration of T1D. Similarly, Nederstigt et al. [144] reported that the prevalence of IgA-AGA increased with the duration of T1D, whereas endomysium antibodies decreased with age. Interestingly, IgA-tTG-seropositive patients with T1D have been found to have lower titers of IgG-tTG and deamidated gliadin peptide antibodies compared to CD patients without T1D [147]. In addition, longitudinal studies suggest that AGA titers can fluctuate over time [148] but also that diabetes-related antibodies may respond to a GFD in cases with CD [149]. Lastly, data from Salardi et al. [150] support that the prevalence of CD significantly increased among Italian patients with T1D after 1994, however, this might also reflect an increase in the prevalence of CD in the general population.

#### **6. Autoimmune Thyroid Diseases**

ATDs affect 2–5% of the population with a female predominance. The most common ATDs are Hashimoto's thyroiditis (HT) and Graves' disease, which lead to hypothyroidism and hyperthyroidism, respectively [114].

#### *6.1. Gluten-Free Interventions in Autoimmune Thyroid Diseases*

Few studies have investigated whether a GFD can contribute to ameliorating thyroid-related pathology among patients with concomitant CD, but we have not been able to identify publications exploring the effects of a GFD in ATDs in the absence of CD or celiac-related antibodies.

A controlled trial has investigated the effects of six months on a GFD (*n* = 16) compared to no dietary intervention (*n* = 18) among drug-naive women with HT [151]. The GFD resulted in a drop in the levels of thyroid peroxidase (TPO) and thyroglobulin antibodies, an increase in 25-hydroxyvitamin D and an improvement in the structure parameter inference approach (SPINA)-GT index, which correlated with the changes in antibody titers. No effect was seen on levels of thyrotropin and free triiodothyronine. The study population included patients that were seropositive for IgA-tTG, however, no intestinal biopsies were performed and patients with symptomatic CD were excluded [151].

An Italian multicenter study evaluating the thyroid function of 128 patients with newly diagnosed CD before and one year after the introduction of a GFD reports that, in some patients, a GFD can reverse thyroid abnormalities [152]. Valentino et al. [153] also noted an improvement in symptoms related to hypothyroidism and thyroxine dosage among three ATD patients with concomitant CD that followed a GFD for six months. However, levels of thyroglobulin and TPO antibodies only changed for one patient, who had an additional follow-up at 18 months [153].

On the contrary, Mainardi et al. [154] report that a GFD did not seem to influence thyroid function and antibodies among two cases of concomitant CD and ATD. Likewise, a more recent study found no effect of one year on a GFD on levels of TPO antibodies that were present among 10 (37%) patients with newly diagnosed CD [155]. On the contrary, thyroid volume significantly decreased compared to a group of patients without CD, indicating that thyroiditis was continually progressing even after the establishment of a GFD [155]. It is possible that a longer study time is necessary to reveal an effect of a GFD, as TPO antibodies were only present among 76.9% (10/13), 46.1% (6/13) and 15.3% (2/13) of CD patients with ATD at, respectively, 6-, 12- and 24-month follow-ups on a GFD [149].

Interestingly, a study found that patients with concomitant CD and HT (*n* = 14) needed an almost 50% higher dose of levothyroxine to reach target thyroid-stimulating hormone (TSH) values when compared to patients with HT alone (*n* = 68) [156]. The authors suggest that this could potentially be explained by reduced absorption of levothyroxine in cases of untreated CD, as an increased need for levothyroxine was prevented by the introduction of a GFD (*n* = 21). However, reduced absorption capacity cannot explain why patients with concomitant HT and CD had significantly higher TSH (5.7 vs. 7.26, *p* = 0.0099) and significantly lower free T4 (1.12 vs. 0.01, *p* < 0.0001) compared to patients with isolated HT before the initiation of levothyroxine treatment [156]. In accordance with the above, Zubarik et al. [157] reported that patients requiring high doses of levothyroxine to maintain an euthyroid state were more likely to have CD, but this was not confirmed by Sharma et al. [158].

#### *6.2. Gluten-Related Serology in Autoimmune Thyroid Diseases*

Identified studies measuring levels of IgA-AGA and IgG-AGA in ATDs are summarized in Table 3. Furthermore, a study measuring the presence of IgG antibodies against 125 foods found no difference in IgG positivity for wheat or gliadin between 74 patients with HT and 245 HCs [180], but IgG positivity for barley was significantly higher among patients with HT compared to HCs (93.2% vs. 71.0%, *<sup>p</sup>* <sup>=</sup> 8.4 <sup>×</sup> <sup>10</sup><sup>−</sup>5) [180]. Moreover, a study supporting the previously discussed association between CD and an increased need for levothyroxine found that patients treated with high dosage of levothyroxine (125–200 μg/day) had significantly higher levels of IgA-AGA compared to patients receiving low levels of levothyroxine (50–100 μg/day) (medians: 19.69 vs. 13.00, *p* = 0.033) [162].


An interesting study found that the prevalence of chronic thyroiditis or seropositivity for TPO antibodies was higher among 16 patients with T1D that were seropositive for AGA compared to 37 AGA-seronegative T1D patients (38% vs. 2.7%, *p* = 0.005 for chronic thyroiditis and 69% vs. 27%, *p* = 0.01 for TPO seropositivity) [181]. This is further supported by a study reporting that the prevalence of tTG (*p* = 0.023) and glutamic acid decarboxylase (GAD) (*p* < 0.00001) antibodies increased with increasing titers of TPO antibodies [164]. A correlation between TPO and IgA-tTG antibodies has also been illustrated in a study suggesting that IgA-tTG may contribute to thyroid dysfunction by binding to thyroid tissue [182]. Similarly, IgG-tTG and IgA-AGA have been found to be predictors of TPO and thyroglobulin antibodies, respectively (IgG-tTG/ TPO: β = 0.12, 95%CI = (0.03–0.21), *p* = 0.008, IgA-AGA/thyroglobulin: β = −0.10, 95%CI = (−0.19–−0.002), *p* = 0.045) [168]. The association between ATDs, T1D and CD has been confirmed by additional publications [178,183], including a population-based cohort study concluding that CD is a risk factor for later development of ATDs among patients with T1D [184].

#### *6.3. Comorbidity between Celiac Disease and Autoimmune Thyroid Diseases*

A systematic review and meta-analysis of 27 studies calculated the median prevalence of CD in ATDs to be 3.2%, however, a pooled analysis resulted in a prevalence of 1.6% (CI = (1.3–1.9%)) for biopsy-verified CD [185]. The abovementioned low prevalence can possibly be explained by the fact that intestinal biopsies are not performed in all seropositive patients with potential CD. Furthermore, the prevalence of CD was higher among children with ATDs (6.2%, CI = (4.0–8.4%)) compared to adults (2.7%, CI = (2.1–3.4)) [185], whereas another study suggests that the prevalence of CD is higher among patients with ATDs above the age of 65 [167].

A meta-analysis of a systematic review from 2016 revealed a significantly higher prevalence of thyroid disease among patients with CD compared to controls (OR = 3.08, 95%CI = (2.76–3.56)) [186]. Similar results were also found for euthyroid ATD (OR = 4.35, 95%CI = (2.88–6.56)) and hypothyroidism (OR = 3.38, 95%CI = (2.73–4.19)), however, the prevalence of hyperthyroidism among patients with CD did not differ from that in controls (OR = 1.28, 95%CI = (0.37–4.46)) [186]. On the contrary, data from 3209 patients with Grave's disease and 1069 HCs support the idea that the prevalence of CD is higher among patients with Grave's disease (1.1%) compared to HCs (0.3%) (OR = 3.81, 95%CI = (1.17–12.41)) [187]. Additionally, a meta-analysis reports that the prevalence of biopsy-proven CD is higher among patients with hyperthyroidism (2.6%, CI = (0.7–4.4%)) compared to patients with hypothyroidism (1.4%, CI = (1.0–1.9%)) [185]. We hypothesize that the late age of disease onset for hyperthyroidism could be a potential explanation for the above contradictive results. An association between thyroid disease and CD has also been confirmed by more recent studies [188,189]. One calculated that the prevalence of thyroid disease was fourfold higher among 288 patients with untreated CD compared to 250 controls without CD (13.6% vs. 3.2%, *p* < 0.05) [188] and the other calculated the hazard ratio of subsequent hypothyroidism among patients with CD to be 4.64 (95%CI = (2.88–7.46)) [189].

It has been debated whether the late diagnosis of CD and, as a result, the late introduction of a GFD can increase the risk of developing other autoimmune diseases [55,102,103,190]. When a meta-analysis compared treated and untreated patients with CD, no difference was found in the frequency of thyroid disease (OR = 1.08, 95%CI = (0.61–1.92)) [186]. In addition, a study highlights that first-degree relatives of patients with CD also have an increased risk of ATDs [191]. Interestingly, a study reports that the prevalence of ATDs among Irish women with CD has decreased significantly over recent decades [192], whereas another study suggests that the prevalence of autoimmune thyroiditis may be higher among seronegative (26.9%) compared to seropositive (9.7%) patients with CD (*p* = 0.002) [193]. Last, but not least, the prevalence of ATDs has also been reported to be high among patients with non-celiac gluten/wheat sensitivity [194,195] and dermatitis herpetiformis [196].

#### **7. Conclusions**

The current level of evidence is yet not sufficient to recommend a GFD to patients with MS, psoriasis, T1D or ATDs. Larger epidemiological studies and meta-analyses of systematic reviews support that psoriasis, T1D and ATDs are all associated with CD, but this does not seem to be the case for MS. The only clinical trial to have studied the effects of a GFD among patients with MS found positive results on important MS-specific outcomes, however, the publication was subject to major limitations. Further studies are warranted to replicate the results found by Rodrigo et al. [31] and clarify whether any beneficial effects could be restricted to specific subgroups of patients. With regard to psoriasis, the currently available data suggest that patients with gluten-related antibodies or CD may benefit from a GFD, however, larger trials are still missing. The majority of studies failed to reveal an effect of a GFD on diabetes-related autoantibodies, however, it seems likely that a GFD may contribute to normalizing metabolic control in patients with T1D. In addition, some publications report that untreated CD can affect metabolic control and diabetic complications in patients with T1D. On the contrary, studies support the idea that thyroid-related antibodies may respond to a GFD in patients with concomitant CD and ATD, however, no studies have addressed the effects of a GFD among non-celiac patients with ATDs to date. Lastly, in patients with concomitant CD and ATD, a GFD may improve the absorption of levothyroxine.

**Author Contributions:** Conceptualization, M.P., J.C.A., J.L.F. and K.J.; methodology, M.P., J.C.A. and K.J.; writing—original draft preparation, M.P., J.C.A. and K.J.; writing—review and editing, M.P., J.C.A., J.L.F. and K.J.; visualization, M.P. and K.J., funding acquisition, M.P., J.C.A. and J.L.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This article was supported by Scleroseforeningen (the Danish Multiple Sclerosis Society), as well as Kirsten and Freddy Johansen's foundation.

**Conflicts of Interest:** The authors declare no conflict of interest in relation to this paper. 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.


**A1.**MetaboliccontrolandeffectsofaGFDinpatientswithconcomitantT1DandCD.CD:celiacDisease,GFD:gluten-freediet,NS:notsignificant,T1D:

**Appendix**

 **A**



**TableA1.***Cont.*

#### **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* **TG6 Auto-Antibodies in Dermatitis Herpetiformis**

#### **Marios Hadjivassiliou 1,\*, Timo Reunala 2,3, Kaisa Hervonen 2,3, Pascale Aeschlimann <sup>4</sup> and Daniel Aeschlimann <sup>4</sup>**


Received: 24 July 2020; Accepted: 15 September 2020; Published: 21 September 2020

**Abstract:** Dermatitis herpetiformis (DH) is an extraintestinal manifestation of gluten sensitivity, in which an autoimmune response is directed against transglutaminase 3 (TG3), an epidermal transglutaminase. TG2 is the autoantigen in celiac disease (CD), defined by the presence of enteropathy, and TG6 is the autoantigen in neurological manifestations of gluten sensitivity. The interplay between B cell responses to these 3 transglutaminases in developing the clinical spectrum of disease manifestations is not completely understood. Also, the individual or combined diagnostic and predictive value of the respective autoantibodies is not fully explored. We examined the prevalence of TG6 antibodies in a cohort of patients with DH. TG6 positivity was found in 13/33 (39%), with IgA detected in 11 patients, IgG in 3, and both in 1. This was significantly higher compared to what is seen in the classic CD cases (14%) in a Finnish population. TG6 positive baseline samples constituted 60% of DH patients with no enteropathy (*n* = 10), as opposed to 17% positivity in those with overt enteropathy (*n* = 12; Marsh IIIB). Repeat testing after adherence to a gluten-free diet for 1 year showed reduced titers for TG6 antibodies in 11/13 (85%), whereby 7 patients were now TG6 antibody-negative. Four patients seroconverted and tested positive for TG6 antibodies at one year, due to the ongoing exposure to gluten. We report another patient who presented with neurological manifestations (encephalopathy) leading to the diagnosis of CD, who was intermittently adhering to a gluten-free diet. Serological testing at baseline showed him to be positive for antibodies to all 3 transglutaminases. Eleven years later, he developed DH. He also subsequently developed ataxia and peripheral neuropathy. Although TG3 and TG6 autoantibodies are linked to certain disease manifestations, TG2, TG3, and TG6 autoantibodies can be present across the spectrum of GRD patients and might develop years before onset of symptoms of extraintestinal manifestations. This is consistent with gluten-dependent adaptive immunity being a necessary but not sufficient pretext to organ-specific damage. TG6 antibodies appear to develop more frequently in patients where tolerance to gluten was broken but, either there was no development of the molecular state driving the tissue destruction at the level of the gut, or perhaps more likely, there was more resistance to developing this phenotype.

**Keywords:** transglutaminase antibodies; TG2; TG3; TG6; dermatitis herpetiformis; gluten ataxia; celiac disease; gluten encephalopathy; gluten neuropathy

#### **1. Introduction**

Gluten-related disorders (GRD) are a group of immune-mediated diseases with diversemanifestations, triggered by the ingestion of gluten [1]. Enteropathy/celiac disease (CD) is not a prerequisite for the diagnosis of GRD, and some patients exclusively present with extraintestinal manifestations in the absence of enteropathy. Such manifestations include skin involvement in the form of dermatitis herpetiformis (DH) and a diverse range of neurological dysfunction, including cerebellar ataxia, sensorimotor axonal neuropathy, sensory ganglionopathy, and encephalopathy characterized by headaches and cognitive difficulties, often with white matter abnormalities on brain imaging [2].

The identification of TG2 as the autoantigen in CD was an important step in our understanding of the pathophysiology of CD [3]. Assessment of serum anti-TG2 antibodies has since become an important tool in CD diagnosis, as a surrogate marker of disease [4]. Recent success in recapitulating the hallmark features of CD including villous atrophy, plasmacytosis, and anti-TG2 autoantibodies in a mouse model shed light on the interplay between gluten, genetics, and IL-15 driven tissue inflammation in the establishment of CD pathology [5]. Importantly, these studies revealed how overexpression of IL-15 leads to activation of intraepithelial cytotoxic T cells, thereby providing a mechanistic explanation regarding the absence of intestinal tissue destruction ('normal' gut mucosa), even in the presence of adaptive gluten immunity in some patients. However, despite these advances, the mechanism underlying the clinical spectrum of GRD remains incompletely understood [1]. Variations in the specificity of antibodies produced in individual patients appear to be linked to specific extraintestinal manifestations. The epidermal transglutaminase 3 (TG3) was shown to be the autoantigen in DH [6]. The discovery of another transglutaminase primarily expressed in neural tissue (TG6), that shared enzymatic properties with both TG2 and TG3, offered further insights into the pathophysiology of neurological manifestations of GRD [7]. Patient-derived autoantibodies to these different isozymes are not crossreactive [7,8], and their development appears to be linked to the shared enzymatic properties of these enzymes rather than their structural similarity and potential shared epitopes (for review see [9]). This, therefore, suggests that any of these transglutaminases could be the primary immunological target of the gluten-driven immune response, although relative abundance in the gut and sensitivity to regulation by proinflammatory mediators explains the unique association of TG2 with GRD development.

The potential interplay between B cell responses to these 3 transglutaminases in disease manifestation and individual or combined diagnostic and predictive utility of respective autoantibodies was not fully explored. Here, we examine the prevalence of TG6 antibodies in a well-characterized cohort of patients with DH, and we discuss the implications of the findings for interpretation of patient serology in the wider context of GRD. We also discuss an interesting clinical case that highlights the potential predictive value for these antibodies.

#### **2. Methods**

#### *2.1. Patient Selection*

Serology samples from 33 DH patients diagnosed with granular IgA deposits on skin immunofluorescence biopsy were collected at diagnosis, and then 6 months and one year after a gluten free-diet (GFD) at the Tampere University Hospital; 31 patients adhered to the GFD. All patients provided informed consent and the project was approved by the local ethics committee (no specific code was allocated). The project adhered to the ethical principles for medical research, according to the Declaration of Helsinki. Small bowel biopsy was taken at diagnosis and was graded histologically as subtotal or partial villous atrophy with crypt hyperplasia and increased intraepithelial lymphocytes (24 patients), or normal mucosa (9 patients). Sera of a control group consisting of 27 patients with atopic dermatitis and 9 patients with psoriasis were also analyzed.

#### *2.2. Serological Testing*

Determination of anti-TG6 IgA and IgG was done using our in-house ELISA assays. The methodology was described in detail elsewhere [7,10]. Full-length human TG6 was produced in SF9 cells and diluted to 2 μg/mL in 20 mM Tris/HCl, 300 mM NaCl, pH 7.6, for coating of high-capacity protein binding 96-well plates (ImmulonTM 2 HB). All steps to reveal antibody binding were performed according to the published procedure. All serum samples were analyzed in duplicate, on wells containing antigen or only BSA, included on the same plate. The BSA-only background was subtracted from the values for antigen, and the units were calculated from a series of standards run in parallel, whereby a measurement >75 U/mL for IgA or >34 U/mL for IgG was considered to be positive. Standards were calibrated to be consistent with those used in commercial assay produced by Zedira (2nd generation assay). Results are given as the mean of two independent determinations.

#### **3. Results**

The clinical characteristics of the cohort of patients with DH were already published [11]. In brief, TG3 antibody positivity was seen in 88% (29/33) as compared to 24% (19/79) in patients diagnosed with classic CD. Endomysial antibody (EMA) positivity was found in 79% (26/33) of the DH group. The percentage TG3 antibody positivity dropped from 86% (24/28) to 21% (6/28), after a year on strict gluten-free diet.

In the current study, TG6 antibody positivity was found in 13/33 (39%) of DH patients (Figure 1). Eleven were positive for IgA anti-TG6, 3 for IgG, and 1 for both. Sera from patients with unrelated dermatological conditions (36) were also investigated as a control, with one psoriasis patient, testing positive for TG6 IgA (3%). Interestingly, tissue destruction, as determined by the histopathology of intestinal biopsies showed an apparent inverse correlation to serum findings, with TG6 autoantibodies detected in the baseline samples of 6/10 (60%) patients with DH without enteropathy, as opposed to 5/11 (45%) with partial villous atrophy (Marsh IIIA), and 2/12 (17%) in those who had overt enteropathy (Marsh IIIB). Three patients with TG6 antibodies (23%) were serologically negative for anti-TG2 IgA (ELISA & EMA), which is similar to 7/33 (21%) in the cohort as a whole. However, all TG6-positive DH patients also had circulating TG3 IgA.

**Figure 1.** Serum concentration of (**A**) anti-TG6 IgA and (**B**) anti-TG6 IgG in the Dermatitis herpetiformis (DH) cohort (*n* = 33) at baseline. The cut-off limits are indicated by the dashed line. Samples from patients with psoriasis (P; *n* = 9) or atopic dermatitis (AD, *n* = 27) were included as non-Gluten-related disorders (GRD) controls. Note, for one AD sample, no result could be obtained due to unacceptably high non-specific reactivity in the assay.

Repeat testing after adherence to a gluten-free diet (GFD) for 6 months and 1 year (strict *n* = 28, partial *n* = 3) showed a response in 11/13 (85%) patients (Figure 2). Seven patients who were positive at baseline, tested negative for TG6 antibodies, after 1 year on GFD. One of the two patients that failed to respond to the GFD, also remained positive for anti-TG2 and anti-TG3 IgA, whereas the other patient

was only partially compliant with the diet, and although this patient became negative for deamidated gliadin peptide (DGP) antibodies he had persistently high titers of anti-TG3 IgA. Despite this evident effect of gluten withdrawal, and somewhat unexpectedly, the overall TG6 antibody positivity at one year was 8/31 (26%). This was because there were 4 patients who had seroconverted to become positive at one year, 2 for anti-TG6 IgA, and 2 for IgG (Figure 2). This appeared to correlate with ongoing exposure to gluten, as indicated by the other serology (DGP) in two patients (1 IgA and 1 IgG), while no obvious explanation could be found in the other two.

**Figure 2.** Longitudinal analysis of serum concentration of (**A**) anti-TG6 IgA and (**B**) anti-TG6 IgG in DH patients that tested positive for these antibodies at baseline on a gluten-free diet (GFD) (left panel, closed circles). Patients who tested negative at baseline but subsequently developed antibodies are given in the right panel (open circles). Note—two patients with TG6 IgA failed to respond to GFD, one of which also became positive for TG6 IgG after 6 months of GFD.

Given the high prevalence of anti-TG6 autoantibodies in this cohort of DH patients, patient records were retrospectively investigated for relevant history. Of the 31 DH patients that were analyzed longitudinally (GFD), six were dead and 25 were followed up to 2011 (mean 21 years), through questionnaire and hospital records. One death from Alzheimer disease was recorded but no other neurological conditions were found in any patient. Autoimmune diseases occurred in four, three had thyroid disease, and one had type 1 diabetes mellitus.

#### **4. Predictive Value of Di**ff**erent Transglutaminase Antibodies: An Illustrative Case**

A 41-years old man presented to neurology in 1997, with intractable headaches for the previous 2 years. He described them as severe, often unilateral and throbbing, lasting for several days and often associated with focal but transient neurological deficits (e.g., hemisensory disturbance and double vision). He also complained of memory difficulties and inability to concentrate. Brain MRI was done to rule out any ischemic cerebrovascular events. This showed white matter changes not typical of stroke or inflammation but of undetermined clinical significance (Figure 3). He had no vascular risk factors. He was started on aspirin and discharged. Subsequent outpatient clinical review showed ongoing symptoms of headache, and poor concentration and compromised memory function interfering with his everyday activities. Additional investigations at that stage ruled out systemic lupus erythematosus (SLE) and antiphospholipid syndrome. Cardiac echo and 24-h cardiac recordings, as well as vascular imaging were normal. Blood tests (available at that time) found him to be positive for anti-gliadin (AGA) and EMA antibodies. Duodenal biopsy confirmed the presence of gluten-sensitive enteropathy. It was thought that the headaches and MRI changes were secondary to CD (gluten encephalopathy) [12]. The patient was given advice for GFD by an experienced dietitian. He was followed up at regular intervals (every six months). Initial review after being on the diet for 6 months showed significant improvement in his headaches and cognitive difficulties. His adherence to the GFD over subsequent years was intermittent for a number of reasons—he could not afford gluten-free products, family problems and housing issues, and a one-year spell in prison. During this period, his antibody profile remained positive. He continued to be reviewed by the dietitian and attempts were made to ensure strict adherence to GFD. In 2006, he completely abandoned the GFD, but he restarted it a year later. He gave up the diet again in 2009. A few months later, he presented with an itchy vesicular rash over his arms and face. Dermatological review and skin biopsy confirmed the diagnosis of DH. He was still consuming gluten and serological testing for TG2 IgA, EMA, and anti-gliadin antibodies confirmed the presence of CD-related antibodies. He remained symptomatic with frequent headaches. More recently, he developed a degree of gait incoordination and a tendency to fall. He also complained of distal sensory disturbance with a burning feeling in his feet, less so in his hands. Further brain imaging showed evidence of cerebellar atrophy (Figure 4) that was not present in the baseline scan. Neurophysiological assessment including thermal threshold studies confirmed the presence of small fiber neuropathy.

Serum from this patient was stored at the time of the diagnosis of gluten encephalopathy (1998), and was subsequently available for testing for TG2, TG3, and TG6 autoantibodies, when these serological tests became available [7,13]. The tests showed him to be positive for deamidated gliadin peptide antibodies (IgA/IgG) and, interestingly, all 3 types of transglutaminase autoantibodies (TG2 IgA/IgG, TG6 IgA, and TG3 IgA/IgG). The TG2 antibody positivity would be expected on the basis of CD. TG6 antibody positivity would be in keeping with the diagnosis of gluten encephalopathy and the subsequent development of cerebellar ataxia (gluten ataxia) and neuropathy (gluten neuropathy), due to poor adherence to a GFD. The positive anti-TG3 antibody result from the 1998 sample would explain the subsequent development of DH, which however, manifested in 2009, over 10 years after the initial presentation with the neurological complaints. The patient is still under regular review and during his last attendance (August 2020) he tested positive for TG2, EMA, AGA, and TG6 IgG and IgA antibodies. AGA and TG2 antibodies were tested throughout his clinic appointments and were always positive. An observed reduction of the TG2 titer was only seen in 2008 and 2016, whilst he was trying to be strict with his GFD.

**Figure 3.** Brain Magnetic resonance imaging (MRI) scan conducted on presentation in 1997. The patient was diagnosed with gluten encephalopathy, having presented with headaches and cognitive difficulties. Serological testing and subsequent duodenal biopsy confirmed celiac disease (CD). The scan shows white matter changes typical of what is seen in the context of gluten encephalopathy.

**Figure 4.** Brain MRI scan conducted on the same patient, as in Figure 3. The scan was conducted in 2018 and at that point showed evidence of cerebellar atrophy that apparently developed over an interval of 21 years. The patient now displays clinical evidence of cerebellar ataxia.

#### **5. Discussion**

The significance of the serological presence of TG2, TG3, and TG6 antibodies amongst different populations of patients with gluten-related disorders needs clarification. Here, we report a high prevalence of circulating anti-TG6 autoantibodies in DH patients (39%), which was an unexpected finding. It supports the notion that GRD patients, independent of clinical presentation may produce any of these TG isozyme-specific antibodies, or indeed any combination thereof. However, we previously showed that the prevalence of circulating TG6 antibodies in CD patients presenting with ataxia in the UK was much higher than in classic CD patients (73% vs. 40%) [14]. Using the same methodology, the prevalence of TG6 antibodies in Italian pediatric CD patients presenting to gastroenterologists was found to be 25% [15]. Furthermore, in this pediatric cohort of CD patients, a significant correlation between duration of gluten exposure, before the CD diagnosis, and anti-TG6 prevalence/concentration was found [15].

The TG6 antibody prevalence in these 3 groups (patients presenting with neurological manifestations, adult classic CD patients, and pediatric CD patients presenting to the gastroenterologists) is analogous to what was observed in patients with DH. Circulating TG3 antibodies (DH-specific epidermal autoantibodies) were found in up to 87% of patients with DH but in only 24% and 11% of adult and pediatric classic CD patients, respectively [16]. It is noteworthy that in untreated patients

with DH, not all patients have circulating anti-TG3 antibodies, yet 100% have IgA-TG3 deposits in the papillary dermis, the site of the primary manifestation [6]. Therefore, although the presence of these antibodies (to TG2, TG3, and TG6) in the serum is diagnostically helpful, their absence does not preclude a localized response at the level of the target tissue (gut, skin, and brain).

Significant advances in understanding of the events leading to autoantibody development were made over the last decade. DH patients were recently shown to have TG3-antibody secreting plasma cells at the level of the gut. Importantly, a gluten challenge of such patients revealed that gluten exposure drove rapid expansion of the TG3-specific plasma cell population in the lamina propria, and their frequency correlated with the serum titer of the corresponding autoantibodies [17]. Furthermore, there was no evidence that this cell population recognizes TG isozymes other than TG3 [17]. This is in keeping with the results from the analysis of patient-derived immunoglobulins, indicating that cross-reacting antibodies are rare [7,8]. Similarly, we analyzed GRD patients presenting with classical CD or gluten ataxia for TG6-specific plasma cells and demonstrated their presence in the lamina propria (Aeschlimann, Dos Reis, Hadjivassiliou, unpublished results). Hence, it is likely that antibodies are generated at the level of the gut; not just TG2 antibodies but also autoantibodies to other TG isozymes. All TG isozymes implicated in GRD form stable thioester complexes with gliadin peptides [18], leading to the uptake and ultimately the presentation of MHC-gliadin complexes by B cells expressing the respective TG-specific IgD. Their activation can consequently be driven through interaction with gluten-specific T cells. In vitro studies confirmed that this was possible [19], and this explanation is consistent with the gut resident T cell response and exquisite gluten dependence of TG autoantibody production, as demonstrated here for TG6 antibodies in DH patients. Enhanced intestinal permeability might drive the immunological reactions that lead to antibody development and circulatory presence [20,21].

How the development of adaptive immunity leads to extraintestinal manifestations remains a matter of debate, although some evidence that the autoantibodies themselves play a role in this has been put forth [9]. Perivascular antibody deposition, as observed in GRD patients can drive organ-specific inflammatory processes that result in tissue damage. TG3 antibodies within immune complexes formed in the papillary dermis of DH patients retain enzymatic activity and thereby drive the innate immune cell activity [22]. Importantly, the demonstration that circulation-derived anti-TG3 antibodies can induce a dermatitis herpetiformis-like pathology in human skin-grafted SCID mice, supports a central role for TG isozyme-specific antibodies in disease establishment in different organ systems [23]. However, expression of anti-TG2 antibodies by themselves did not precipitate CD-like lesions in the small intestine or overt systemic manifestation akin of GRD [24], identifying that establishment of overt tissue damage might require failure in more than one immune regulatory system. The adaptive immune response is a prerequisite for the development of intestinal villous atrophy but is not by itself sufficient. An interplay with the innate immune system that drives IL-15 overexpression in two distinct tissue compartments in the gut is required to mediate tissue destruction [5]. Perhaps this paradigm also applies to the extraintestinal manifestations, with development of anti-TG3 or anti-TG6 antibodies being required for the respective organ-specific manifestations but not being sufficient to precipitate overt tissue damage. This notion is consistent with the observation that the 3 types of autoantibodies could occur across the different forms of GRD, but without apparent clinical correlation. Despite high prevalence of anti-TG6 antibodies in the DH cohort, retrospective analysis of patient history did not reveal any related neurological problems, as was the case in the pediatric CD patients with the anti-TG6 antibodies analyzed previously [15]. We also present a case report that illustrated that the circulating antibodies could indeed be present for long periods of time (>10 years), prior to the onset of the corresponding extraintestinal manifestations. Late onset of DH was also reported in patients that initially presented with classical enteropathy, and appeared to be connected to intermittent GFD (antigen re-stimulation), as was the case here [25,26]. Extraintestinal manifestations were, however, not the consequence of reaching a threshold in the circulating antibody concentration, as there was no correlation between serum concentration and clinical presentation.

Of interest is the potential significance of these antibodies when present in patients who do not seem, clinically at least, to have any obvious corresponding extraintestinal dysfunction. Indeed, our recent work demonstrated that the subgroup of patients with classical CD who are anti-TG6 antibody positive (40% of UK patients) already show a significant reduction in the volume of specific brain regions, when compared to those who are TG6 antibody negative [27]. This finding also suggests that careful investigation (neurological examination and brain imaging) might reveal neurological deficits that are largely ignored, either because they are not considered by gastroenterologists or because the patients would not discuss neurological symptoms in the context of a gastroenterology consultation. Long-term studies of seemingly asymptomatic individuals screened for GRD (e.g., because of family history) who are serologically positive for these transglutaminase autoantibodies, might be helpful in understanding their lifelong significance and help inform the advice given to such patients.

Molecular etiology might also explain the observation that extraintestinal manifestations appear to be a late phenomenon. For example, patients with gluten ataxia and CD presenting with ataxia have a mean age at presentation of 53 years, as opposed to the mean age of 43 years, observed in patients presenting with the classic CD symptoms to gastroenterologists [28]. The prevalence of anti-TG6 antibodies in this Finnish DH cohort was 39%. This figure differed to what was found to be the prevalence of TG6 antibodies in a Finish cohort (same geographical area as the DH patients) of patients presenting with classical CD, which was 12/86 (14%) [14]. Furthermore, we made the observation that TG6 antibody positivity was more prevalent in those DH patients without enteropathy (60%), when compared to those DH patients with overt enteropathy (17%). This is in line with a significant proportion of patients with gluten ataxia and circulating TG6 antibodies displaying none or only minor signs of intestinal tissue damage [14]. TG6 antibodies, therefore, appear to be developed more frequently by patients who lost oral tolerance to gluten but either did not develop the molecular state that leads to tissue destruction at the level of the gut, or perhaps more likely, were more resistant to developing this state due to their genetics.

There were limitations to this study. Ideally, serological testing for all different GRD markers should take place without intermittent sample storage. Longitudinal analysis, as carried out here, is always a compromise between analysis of samples at the point of collection or at a later point when all linked samples for a cohort are available for simultaneous analysis; the latter approach was taken here. In our hands, there is no indication for sample storage (even long-term) affecting measurements of TG6 autoantibodies, as long as the samples were kept in sealed tubes for storage and repeated freeze–thawing cycles were avoided. Such sera are not easy to collect particularly from patients with DH who are becoming increasingly rare. The case described above is a single example of what we observed, and ideally, it should be followed up in a large cohort of patients who do not adhere to a GFD. However, such cases are not common and most patients with CD are likely to adhere to a GFD. Nevertheless, we believe that individual case reports can provide important insights.

#### **6. Conclusions**

We demonstrated here for the first time that TG6 antibodies are prevalent in patients with DH, and occur at a much higher frequency than what is seen in patients with classical CD. It appears therefore that although linked to certain disease manifestations, TG2, TG3 and TG6 autoantibodies can be present across the spectrum of GRD patients and may develop years before the onset of symptoms of extraintestinal manifestations. Nevertheless, our findings suggest that these antibodies could have a predictive diagnostic value for the future development of specific extraintestinal manifestations. Such autoantibody positivity at diagnosis might further inform the decision to adopt a GFD, particularly for those patients that appear to be asymptomatic but are at risk of developing neurological dysfunction in the longer term.

**Author Contributions:** M.H., T.R., and D.A. conceptualized the project. M.H. looked after the patient reported. T.R. and K.H. looked after all D.H. patients and collected the sera. P.A. and D.A. carried out the autoantibody

measurements. M.H. and D.A. produced the draft and all authors contributed and approved the final manuscript. All authors read and agreed to the published version of the manuscript.

**Funding:** No funding was available for this work.

**Acknowledgments:** This is a summary of independent research supported by BRC and carried out at the National Institute for Health Research (NIHR) Sheffield Clinical Research Facility. The views expressed are those of the authors and not necessarily those of the BRC, NHS, the NIHR, or the Department of Health.

**Conflicts of Interest:** T.R., K.H. and M.H. have no conflict of interest to declare. D.A. serves as a scientific advisor/collaborator to Zedira (without financial incentives) and receives royalties from Zedira for patents.

**Ethics Statements:** The D.H. samples were collected by T.R. and K.H. according to local ethics approval and following informed consent from all participants.

#### **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/).

### *Review* **Gluten and Autism Spectrum Disorder**

**Iain D. Croall 1,\*, Nigel Hoggard <sup>1</sup> and Marios Hadjivassiliou <sup>2</sup>**


**Abstract:** An expanding body of literature is examining connections between Autism Spectrum Disorder (ASD) and dietary interventions. While a number of specialist diets have been suggested as beneficial in ASD, gluten has received particularly close attention as a potentially exacerbating factor. Reports exist suggesting a beneficial effect of the gluten-free diet (GFD) in ameliorating behavioural and intellectual problems associated with ASD, while epidemiological research has also shown a comorbidity between ASD and coeliac disease. However, both caregivers and clinicians have expressed an uncertainty of the value of people with ASD going gluten-free, and as the GFD otherwise receives considerable public attention a discussion which focuses specifically on the interaction between ASD and gluten is warranted. In this review we discuss the historical context of ASD and gluten-related studies, and expand this to include an overview of epidemiological links, hypotheses of shared pathological mechanisms, and ultimately the evidence around the use and adoption of the GFD in people with ASD.

**Keywords:** Autism Spectrum Disorder; gluten sensitivity; celiac disease; coeliac disease; review

**Citation:** Croall, I.D.; Hoggard, N.; Hadjivassiliou, M. Gluten and Autism Spectrum Disorder. *Nutrients* **2021**, *13*, 572. https://doi.org/ 10.3390/nu13020572

Academic Editor: Stefano Guandalini Received: 31 December 2020 Accepted: 27 January 2021 Published: 9 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Motivation and Literature Search Methods**

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterised primarily by deficits in social communication and restricted/repetitive patterns of behaviour (DSM-5) [1]. As the name implies its phenotype exists on a spectrum, and overall it is estimated to affect as many as one in 69 children [2]. Interest in the use of specialist diets in ASD is increasing, as a way to alleviate its behavioural and intellectual outcomes. Though many dietary interventions have been suggested, the gluten-free diet (GFD) is among the most notable. Clinically, the GFD is well recognised as the primary treatment for patients with a gluten-related disorder. The most prominent of these is coeliac disease (CD) which is predominantly expressed as a gastrointestinal (GI) condition. However, physiological sensitivity to gluten is known to exist in other forms. These include other immune-mediated disorders (e.g., dermatitis herpetiformis and gluten ataxia), allergic reactions (wheat allergy), and non-coeliac gluten sensitivity (a condition characterized by self-reported gastrointestinal and extra-intestinal symptoms subjectively improving upon a GFD in subjects in whom other major organic gluten related disorders have been excluded) [3]. The clinical utility of the GFD outside of these contexts is debatable, and elsewhere its adoption within the general public as a sometimes "fad" diet heightens such scrutiny [4]. However, generally increased rates of GI problems have been reported in people with ASD, as has evidence of an apparent comorbidity between ASD and CD specifically. As adoption rates of specialty diets which include a gluten-free component are very high in ASD we were motivated to conduct a literature review which focused specifically on the interaction between ASD and gluten.

Searches were made on Pubmed on the 12th of November 2020. Terms included:


These terms were designed to capture a range of relevant terminology, for example "autism spectrum disorder" or "gluten free diet" would each be picked up by these. Regional variation in spelling of coeliac/celiac would also be accounted for. This returned 237 unique articles. The abstracts of all of these were read to determine eligibility for inclusion in the main review. Criteria for this were that the paper must be original research (i.e., not another review or systematic review) which in some way directly explored links between ASD and GRDs, the use of the GFD in ASD, or any other relevant interaction between ASD and gluten. Case reports were excluded, as were papers where the main article was not available in English. Seventy-nine articles were deemed eligible for inclusion. These were read, and any additional relevant citations found through this were added for discussion. Figure 1 shows these included papers according to their year of publication.

**Figure 1.** A histogram of eligible papers found in the primary pubmed search according to their year of publication.

Throughout the study of the literature, common research methods/sub-topics were noted. The remainder of this review synthesises these papers according to those themes. Additional literature was searched for and referenced where necessary to elucidate on a relevant key concept which was not adequately covered by the initial searches.

#### **2. Historical Context**

The first observation of a possible link between gluten and ASD was reported in 1969 by Goodwin & Goodwin [5], who noted in a cohort of 65 children with ASD that one 6 year old boy also had CD. This child's subsequent treatment with a GFD appeared to improve outcomes relating to his ASD. It is relevant to note that at that time the prevalence of CD was considered to be far less than the 1-in-100 that it is sometimes reported as today [6], largely due to the lack of effective diagnostic methods such as serological testing. In a commentary piece on the then-emerging topic, Dohan references in their 1970 paper [7] that CD is

thought to affect approximately 1 in 3000 people, making suspicion after finding it in one of a cohort of 65 children with ASD understandable. Another point of interest emerges from this paper, which was purely focused on links between CD and schizophrenia but in which Dohan also found it relevant to include an anecdotal report of increased CD rates in ASD groups ("at least 2 coeliac patients among 140 severely autistic children"). While limited phenotypic similarities between schizophrenia and ASD are still discussed now, initial characterisation of the conditions involved such overlap that full clinical separation did not occur until 1980 with the publication of the DSM-III [8]. To a modern reader, this explains why there are a number of early articles which combine ASD with schizophrenia and otherwise draw potentially confusing links between the two.

Goodwin et al. published another study in 1971 [9] which is arguably the first trial investigating how gluten modified behaviour in a cohort of children with ASD. Also included were controls and a group of participants with schizophrenia for further comparison. Here, the participants followed the "sprue diet" (GFD) for a single day, in which they were also given a cherry drink which had added to it either gliadin or a placebo (sugar). They were then subsequently monitored and tested via investigation of blood counts and electrophysiological recordings by trans cephalic direct current. Although the authors note some findings, these predominantly focus on differences which appear to separate ASD and schizophrenia participants, providing only one comment on the effect of gliadin where it appeared to reduce plasma cortisol levels. However, this was observed across both control and ASD participants, and when coupled with the small sample sizes (the ASD group had 9 participants for that portion of the results) and extremely short period of dieting, it is difficult to extrapolate any meaningful conclusions. Further studies in the 1970's included a comparison of serum alpha-1-antitrypsin levels between children with ASD vs. children with CD [10] (finding them comparably abnormal and suggesting a shared pathology), an experiment published in a book [11] describing 72 patients with ASD in whom CD was diagnosed (without biopsy) in 8, and another trial of gluten in eight children with ASD who already followed a GFD and were purportedly better for it [12]. This latter study saw the participants stop their diet to undergo a gluten challenge for 1 month, hypothesising this would worsen their phenotype, but finding no change in bodily measurements (weight, bowel habit etc.) or behaviour (measured by parental reports and observation from a specialist paediatrician).

To summarise, early interest in the topic was driven by essentially anecdotal reports of apparently comorbid cases of CD with ASD. Direct experimentation of this resulted in largely negative findings. These studies had small samples sizes and, when viewed with a contemporary lens, suffered from experimental designs and measurement techniques which would now be considered extremely insensitive in targeting relevant outcomes. Following these papers, little relevant research activity appeared until the mid 1990's when the topic appeared to become more popular once again.

#### **3. Gastrointestinal Symptoms in ASD**

A heightened rate of gastrointestinal (GI) symptoms in people with ASD is well documented. While this particular topic in its entirety falls outside of the scope of the present systematic review, the observation of these symptoms is a major motivator for gluten-specific research. Relevant studies from the review are therefore included here, as well as other key literature.

In 2014 Chaidez et al. [13] conducted a large study in which 499 children with ASD were compared to typically-developing (TD) children (*N* = 324) and children with developmental delay (*N* = 137) in terms of GI symptoms measured by 10 Likert scales (abdominal pain, constipation etc.). After controlling for age, sex, maternal education and medications which may lead to GI side effects, children with ASD had significantly heightened odds ratios (OR) compared to controls for 8 outcomes, the lowest being 3.14 (abdominal pain) and the highest being 8.61 (sensitivity to foods). The children with ASD and developmental delay were not significantly different from one another.

This is one of a number of similar studies who's findings are supported by metaanalyses; a pubmed search of "gastrointestinal autism" found the most recent meta-analysis was performed in 2014 [14]. This included 15 studies which gave a combined sample of 2215 children with ASD in which four variables were included; general GI concerns, diarrhoea, constipation and abdominal pain. Each of these was found to be significantly more prevalent in the ASD group compared to TD children, with overall OR's of 4.42, 3.63, 3.86 and 2.45 respectively.

Relevant studies from the current review include a report [15] of a higher frequency of constipation in children with ASD, a study [16] which found children with ASD and regression more often had abnormal stool than those without regression, and another experiment [17] which found GI symptoms to be more common in children and adolescents with ASD than in TD controls, and for these symptoms to be weakly correlated to behavioural measures. Correlations such as these have been documented elsewhere [18–20]. These studies reference the often non-specific nature of GI symptoms with one explaining that "a GI pathology specific to ASD had not been established" (Babinska et al., 2020 [17]). Regardless, a notable body of literature has investigated for a comorbidity between ASD and specific GI conditions, often finding significant results.

#### **4. The Co-Morbidity between ASD and CD**

Following early literature from the 1970's, the first study to investigate for an increased rate of CD in ASD was Pavone et al. in 1997 [21]. Pavone examined a cohort of children with ASD (*N* = 11) to detect the rate of CD (by antibody and ultimately biopsy testing), and similarly a cohort of children with CD (*N* = 120) to detect the rate of those with features of ASD (as reported by parents and according to the DSM III-R). None of the children with ASD had biopsy-proven CD, while none of the children with CD met criteria for a full ASD diagnosis (though a limited few did show isolated features). This was therefore overall a negative study, but the limitations of examining in such small cohorts as 11 are evident.

Since then, a limited number of large epidemiological studies have been conducted which generally do show an effect indicating CD and ASD to be comorbid to one another. One of 2009 [22] which focused specifically on comorbidities to ASD within parental medical history, used the Danish Civil Registration System to identify all children born between 1993 and 2004 with ASD (*N* = 3325). Here, maternal history of CD led to a significant, overall incidence rate ratio (IRR) of 2.97 in terms of the child having ASD. Other studies examining comorbidities within the same participant have also found significant results while using medical databases. A 2017 study [23] examined for the risk of psychiatric sequalae in children with CD (*N* = 10,903), finding a hazard ratio (HR) of 1.5 (univariate analysis) of being diagnosed with ASD after their CD diagnosis but before the age of 18 (adult data was not included). The same research group has replicated this more recently [24] with a larger cohort of people diagnosed with CD while a child (*N* = 19,189), but which this time did also include psychiatric diagnoses obtained after 18. Here, the HR of developing ASD was 1.47, the highest of all disorders included in analyses.

These papers do however contrast an earlier study, again by the same research group [25], which examined specifically for the likelihood of an ASD diagnosis *preceding* a CD diagnosis in children and adults with CD (*N* = 26,995). This OR was non-significant, though potentially of interest was a finding wherein previous ASD was still associated with an increased risk of having normal mucosa on biopsy, but positive CD serological test results (tissue transglutaminase; TTG, endomysial; EMA or gliadin; AGA antibodies, reported as a single grouping). While the immediate clinical implication of positivity to these antibodies varies, i.e., TTG/EMA positivity indicates CD with high sensitivity/specificity while many generally-healthy individuals may exhibit AGA positivity, it should be highlighted that within the study of wider "gluten sensitivity" heightened rates of any of these may be considered pathologically-relevant when compared to an appropriate "control" such as in this discussed study. This is therefore an important study as it highlights the link between serological markers of gluten sensitivity and ASD in the absence of enteropathy.

Other prevalence research has been conducted with less stringent diagnostic criteria and/or smaller sample sizes, finding mixed results. Studies with significant findings include Calderoni et al. [26] who examined a cohort of children with ASD (*N* = 382) and found the rate of CD within the sample was 2.62%, although it should be noted this sometimes relied only on a positive serological (TTG/EMA) result and formal CD diagnosis was not always confirmed. Valicenti-McDermott et al. [16] found an increased family history of CD and/or inflammatory bowel disease in children with ASD who also exhibited regression (*N* = 24), compared to children with ASD without regression (*N* = 71). In a letter to the editor, Barcia et al. [27] report an experiment where of 91 "randomly selected" children with ASD, 4 had "biopsy-proven" CD (the authors reference diagnostic guidelines for diagnosis where Marsh grade 3 denotes CD). This is a rate of 4.4% which is considerably higher than might be expected. Mazzone et al. [28] found in a cohort of 100 children with CD that 2 had ASD (while none of a control group did); whether this is a positive result or not is arguable.

Studies with negative findings include Alabaf et al. [29] who via parental reporting of 91 children with ASD did not find an association with CD (not reported but this was measured, implying a negative finding). Juneja et al. [30] screened children with ASD (*N* = 150) for CD defined by IgA TTG testing, finding no positive tests. In 2012, Batista et al. [31] examined children and adolescents with either ASD (*N* = 147) or biopsy-proven CD (Marsh grade 3, *N* = 211) for the rate of the other. The ASD group was found to be entirely negative for CD (although one subject did have a weakly-positive TTG result with negative EMA), while two cases of ASD were found in the CD group; this was concluded to not be above chance. Zelnik et al. [32] examined a cohort of CD patients (*N* = 111) for a range of neurological outcomes including ASD, although as this was reported mixed in with other learning disabilities and ADHD (which overall affected 20.7% of the group) how common ASD specifically was is not known. Finally, Black et al. [33] examined medical records to identify 96 children with ASD and reported on all diagnosed gastrointestinal comorbidities, failing to identify any CD cases.

A recent meta-analysis [34] combined some of the above studies (where eligible) to find a significant odds ratio of 1.53 in terms of CD patients having ASD, but a nonsignificant likelihood of ASD participants having CD. Overall therefore, the majority of studies relevant to the question of comorbidity have used variable sample sizes and diagnostic methods, making definitive conclusions difficult in most individual instances. However, the strongest powered are undoubtedly those from Sweden which studied large cohorts and established an increased risk of a subsequent ASD diagnosis in people with CD, therefore showing a convincing comorbidity. This is further supported by the meta-analysis finding, which showed the same. Studies which examine the "reverse" of this, where initial ASD features may increase risk of subsequent CD, have led to more negative findings however do demonstrate an association with the development of gluten antibodies in the absence of clinical CD. It should also be highlighted that a very large epidemiological study which principally studies an ASD cohort for the rate of CD is absent, which may introduce a sampling bias when interpreting the findings of this overall field. Overall therefore, ASD does appear comorbid to CD, and while an increased risk of CD in ASD is not currently supported a suspicion of ASD being linked to subsequent, immune-mediated "gluten sensitivity" may be warranted.

#### **5. Hypothetical Mechanisms of Action**

With a comorbidity between ASD and CD established, a natural question is of what shared pathophysiology may drive these associations. Further, as non-specific GI symptoms are also seen to be generally more prevalent in ASD and that gluten sensitivity is increasingly understood to be a spectrum that extends beyond the clinical criteria for CD specifically, it is important to consider any mechanism of action between gluten and ASD.

From an early point in the literature, hypotheses have frequently related in some way to heightened autoimmunity in ASD. While a predisposition towards autoimmunity was noted as early as 1971 [35], enquiries of this nature gathered pace after a key publication in 2001 [36] which showed children with ASD (with regression) to have increased markers of innate and adaptive immune response (TNF-A, cytokines etc.). This was investigated at the time partially in response to parental reports of children with ASD suffering apparently high rates of reactions to dietary irritants, and this autoimmune phenotype was subsequently hypothesised to be part of the aetiology of ASD. The authors presented this idea in terms of environmental stimuli triggering an immune response which exacerbates ASD features, and in the specific case of their study hypothesised it may stimulate regression.

In the early 2000's a series of publications by Vojdani et al. [37–39] built on this evidence by focusing on more specific dietary triggers and hypothetical knock-on effects they would lead to in terms of molecular pathways. Here, focusing mainly on gliadin (a gluten-specific protein) and casein (a protein in dairy products), it was demonstrated that children with ASD have high rates of antibodies against these (i.e., anti-gliadin and anti-casein) as well as antibodies against DPP4, a digestive enzyme. DPP-4 is important in the processing of gliadin. Initially, gliadin is degraded into various peptides which include gliadinomorphin-7 [40], an immune reactive substance with "opioid activity", i.e., which stimulates opioid receptors in the body [41]. Further degradation of gliadinomorphin-7 is therefore required, which is where DPP4 functions by cleaving such peptides [42]. As Vojdani et al. reported, the existence of anti-DPP4 would hypothetically reduce the amount of circulating DPP4, increasing the abundance of gliadinomorphin-7 and the likelihood of downstream, opioid-like effects. It should be highlighted that casein and other dietary peptides are similarly degraded to intermediary substances with opioid properties (e.g., casomorphin [43]), and together these potentially harmful peptides have been termed "exorphins" [44].

Stimulation of the opioid system has been studied in the context of ASD features. An early proponent of this link, Panksepp outlined a theory in 1979 [45] (based largely on his earlier animal model experiments) wherein excess opioid activity may lead to the decreased social behaviour seen in ASD. This theory has persisted until today, with numerous publications concerned with evidence of opioid overactivity in people with ASD. Animal studies have continued to show the importance of a balanced opioid system in maintaining social behaviours which are similar to those impacted in ASD, while experiments investigating for levels of relevant, opioid-like peptides in the sera, CSF or urine of people with ASD have generally shown raised titres albeit with some notable exceptions where decreases have been reported [46]. Indeed, measurement of urine peptides has become a common tool in this field, with high levels being seen as an indication of insufficient digestion of food which may lead to excess exorphins [47].

An alternative theory has focused on the role that oxidative stress may play in ASD, which may lead to a state of inflammation in the brain. It has for example been reported that people with ASD have an impaired antioxidant defence in the cerebellum [48], while problems metabolising nitrous oxide (which may lead to increased oxidative stress [49]) has also been proposed as driver of ASD pathophysiology [50]. This holds a relevance to gluten sensitivity, where increased oxidative stress has also been demonstrated, for example as triggered by gliadin [51] or as demonstrated generally by raised markers of oxidative stress across untreated children with CD [52].

Studies have also noted the potential for shared genetic predisposition. One recent paper by Bennabi et al. [53] compared genotyping data between ASD and control cohorts, finding that the haplotype HLA-DRB1\*11-DQB1\*07 was more common in the ASD group, with this being more prevalent still in those ASD patients with the most pronounced behavioural symptoms. A different haplotype (HLA-DRB1\*17-DQB1\*02) was conversely more common in the control group. The \*07 haplotype was therefore concluded to be potentially causative and the \*02 one protective. Of relevance is that the \*07 haplotype is additionally recognised as associated with CD, leading to a suggestion that there may be a sub-group of people with ASD holding a genetic risk for both [54]. However, other genetic research has produced negative findings, such as a meta-analysis of genome-wide

association studies [55] which did identify regions associated with ASD but noted only overlap between these and schizophrenia.

The potential of reactivity of antibodies to gluten products should be discussed. Antibodies against tissue-transglutaminase (TTG) have very high sensitivity and specificity in diagnosing CD, meaning that as a comorbidity has been demonstrated they may have a relevance in ASD pathology. These antibodies have been reported to lead to apoptosis of neuroblast cells in vitro [56]. Other antibodies which may indicate gluten sensitivity but not CD specifically include transglutaminase 6 (TG6) antibodies, which have been indicated in the diagnosis of gluten ataxia [57] (where the cerebellum is the primary site of damage), with this supported by animal research showing TG6 to be distributed throughout the central nervous system including brain regions such as the cerebellum and thalamus [58]. TG6 antibodies have been reported at a rate of 4.4% in a group of 77 children with ASD [59]; this experiment lacked a control group and as this is a relatively novel marker it is difficult to evaluate if this is abnormal. Finally, gliadin antibodies have been shown to react with brain blood vessel structures [60], show cross-reactivity with neuronal synapsin 1 [61], and to be associated with rates of depression in people with CD and healthy controls [62]. Gliadin antibodies have been measured across a number of studies in ASD, frequently finding them to be raised. Those found in the current review are summarised in Table 1.



The potential for antibodies and other irritants to travel from the gut to the brain is raised by a number of studies which have demonstrated generally inflamed/abnormal intestinal findings in ASD [68–71], and others showing compromised intestinal permeability specifically [66,72]. Indeed, one other publication [73] examined gene and protein expression of brain and intestinal tissue of human ASD subjects, finding evidence of impaired intestinal permeability in combination with altered blood-brain barrier integrity. It should be noted that not all studies support an impacted intestinal permeability [59,74], but nonetheless these phenomena raise the possibility of a gut-brain axis interaction being relevant in ASD. Here, a negative feedback loop between the brain and the gut would lead to exacerbation of both neurological and GI outcomes. Arguably most studies which have investigated how gluten can impact people with ASD may fall within this broader

concept, where irritants enter the bloodstream from the gut to cause downstream negative consequences for the brain. A loop may be completed if the effect on the brain leads to alterations of behaviour and appetite which may maintain or exacerbate the cycle [75].

In summary, pathological interactions between ASD and gluten have focused on opioid activity from improperly digested gluten products, inflammation caused by oxidative stress and/or reactivity with anti-gluten antibodies, and some indications of shared genetic factors. These hypotheses provide some explanation for the previously discussed comorbidity between ASD and CD and also make it appear reasonable that gluten may exacerbate bodily stress in other groups of people with ASD who do not have CD. However, it remains unclear to what extent ASD populations and sub-populations are affected, and to what degree these findings represent a unique interaction with gluten specifically or are a consequence of a generally-raised autoimmune profile in ASD.

#### **6. Trials of the GFD in ASD**

Establishing that gluten is potentially harmful for people with ASD leads to the question of if a GFD would then bring any benefits. After the early studies of the 1970's, the first trial which involved gluten in any capacity was conducted in 1990 and is detailed in two publications [76,77] by Knivsberg et al. Here, fifteen children with ASD in combination with abnormal urine peptide results engaged with a gluten and casein-free diet (GCFD) for four years. Measurements were generally taken at baseline, one year and four year time points, and included scales which characterised psychotic behaviour in children, psycholinguistic ability and fluid intelligence. However, this data was not all collected consistently (e.g., the psychotic behaviour measurements were not made at 4 years), and authors note variable dietary success between the children. Regardless, significant findings suggested improvement across multiple outcomes, including normalisation of urine peptides. The 1990's saw one other trial [78], the primary analyses of which concerned a group of 22 children with mixed spectrum disorders (the most common being ASD) who undertook a GFD for 5 months. Other groups were also examined, e.g., children with ASD already on a GFD took a gluten challenge, although these sample sizes were very small. Similar to the Knivsberg study, improvement in behavioural outcomes was noted in response to the GFD although no change in urinary peptide levels were seen.

The first randomised trial was conducted in 2002 [79]. Here, 20 children with ASD and abnormal urinary peptides were randomised into parallel groups to receive either the GCFD or a regular diet for 12 months. Following this, improvements were noted across behavioural and intellectual outcomes. A number of randomised trials have been conducted since and those that utilise an intervention that in any way involves gluten are summarised in Table 2.

**Table 2.** A summary of randomised trials which have in some way included a gluten-free diet as an intervention in treating ASD.



**Table 2.** *Cont.*


**Table 2.** *Cont.*


**Table 2.** *Cont.*

Findings described in the final column relate to any analysis which indicates with statistical significance that the intervention affected an outcome. 6-GSI; 6-Item Gastrointestinal Symptom Index, ABC; Abberant Behaviour Checklist, ADHD-IV; Attention-Deficit Hyperactivity Disorder—IV rating scale, ADOS; Autism Diagnostic Observation Schedule, AQ; Autism Spectrum Quotient, ASD; Autistic Spectrum Disorder, ASRS; Autism Spectrum Rating Scale, ATEC; Autism Treatment Evaluation Checklist, AWPC; Approach Withdrawal Problems Composite (a subset of the PDD-BI), CARS; Childhood Autism Rating Scale, CARSA; Conners Abbreviated Rating Scale and Actigraphy, CBC; Child Behavior Checklist, CPRS-R; Connor's Parent Rating Scale-Revised, DIPAB; Diagnose af Psykotisk Atfærd hos Børn, ECO; Ecological Communication Orientation, ERC-III; The Behavioral Summarized Evaluation, EQ-SQ; Empathy and Systemising Quotient, GARS; Gilliam Autism Rating Scale, GI; Gastro-intestinal, ITPA; Illinois Test of Psycholinguistic Abilities, LIPS; Leiter International Performance Scale, MABC; Movement Assessment Battery for Children, MSEL; Mullen Scales of Early Learning AGS edition, PDD-BI; Pervasive Developmental Disorders Behaviour Inventory, RIAS; Reynolds Intellectual Assessment Scales, RRLRS; Ritvo-Freeman Real Life Rating Scales, SAS Pro; Severity of Autism scale-Professional Evaulation, SCAS-P; Spence's Children Anxiety Scale-Parent version, SCQ; Social Communication Questionnaire, SRS; Social Responsiveness Scale, SSP; Short Sensory Profile, VABS-2; Vineland Adaptive Behavior Scale, Second Edition.

> Examining this literature reveals a very mixed picture of findings. Of the 13 RCT's found in the current review, improvements of some kind were noted in 6 [79,82–85,90], no findings were observed in another 6 [80,81,86,88,89,91], while a worsening of GI symptoms (in response to the GCFD) was observed in one study [87]. All 6 studies which noted a positive effect from the interventional diet included improvements in intellectual/behavioural outcomes, and sometimes also in physiological measurements (e.g., GI symptoms).

> Consolidation of these studies is difficult even beyond the mixed findings. One immediate observation is that the exact dietary intervention employed is variable. This increases heterogeneity between studies and makes commenting on the effect of gluten specifically impossible in most cases. Only 3 of the 13 trials had a group design which in some way tested the GFD in isolation; one of these reported improvements in outcomes [85]. Others typically focus on the GCFD (the most investigated of all interventions), while some use unique interventions such as Grimaldi et al. [82] who primarily tested a probiotic mixture (in combination with the GCFD). The study by Adams et al. [83] is also notable for the approach of sequentially accumulating interventions over a year which included the likes of dietary supplementation and epsom salt baths, with the GCFD being added at day 210.

> The majority of studies are also unblinded (8 of 13) which raises a risk of placebo/nocebo effects. Some trials which are blinded use as a placebo gluten-free versions of food (bread etc.) given to participants on the assumption that they will not be able to tell the difference, meaning that a degree of skepticism is warranted even for those with such an experimental

design. Authors of non-blinded trials that achieve significant results acknowledge this limitation but highlight the practical difficulty of effective blinding for a GFD over a long period of time, or blinding of the other mixed interventions employed. This often leads to varying levels of "blindedness" within a trial, depending on the specific intervention/outcome examined. For example Adams et al. [83] write "A strength of the study is that it was a randomized, controlled study, but a major limitation of this study is that implementation of a healthy, HGCSF (healthy/gluten/casein/soy-free diet) does not allow blinding of participants. The RIAS evaluation was single-blinded, and the CARS and SAS-Pro were semi-blinded (the evaluators were blinded, the participants were not), so those results are fairly robust. The parent evaluations certainly are subject to some placebo-effect but provide an upper-bound on possible benefits. The laboratory measurements were conducted in a blinded manner, so those results should be reliable."

Studies variably do or do not use dietary "run-in" periods, which would be generally advisable to account for delays in physiological adjustment between different regimens when taking experimental measurements. Some trials are conducted over very short timeframes, such as Navarro et al. [88] which ran for 4 weeks or Pusponegoro et al. [87] which ran for 1 week. The implication of this will vary depending on the outcomes measured, but regarding gluten it is for example known that resolution of symptoms due to gluten exposure can take a number of weeks in patients with CD [92], while achieving gliadin antibody negativity can take 6 months or longer [93]. This emphasises the need for long term trials if adequate time is to be given for changes to be captured. In terms of assessing change, the measurement scales used are also scarcely replicated between studies. The majority of RCTs employ a set of tools which are unique to that particular trial, further complicating comparisons or synthesis of findings. An effort to arrive at an agreed-upon set of outcomes would benefit these trials greatly, as would purposeful replication of alreadyreported significant findings using the same measurement techniques. Otherwise, it also remains an open question as to how much current findings are driven by e.g., different tool sensitivities.

Taken together, it is very difficult to identify a single trial which arguably addresses all of these concerns. Ghalichi et al. [85] is the largest of those identified (80 subjects randomised), but this ran for 6 weeks and was not blinded. The longest running trials were Whiteley et al. [90] and Adams et al. [83], which took principle measurements over 12 months. Each of these did have modest sample sizes (*N* = 73 and *N* = 67 randomised, respectively), but neither were blinded and as discussed Adams et al. included a wide range of accumulative interventions. This highlights a real gap, wherein a well-powered, long-duration and placebo-controlled trial of either the GFD or GCFD has not yet been conducted. Such an experiment would ideally be run after a community consensus is reached regarding what outcomes should be focused on. Until such a trial is conducted a confident, overall conclusion cannot be made. Currently therefore, the overall pattern of the available literature does not support a proved benefit of the GFD in people with ASD (who do not have a clinical diagnosis of CD).

#### **7. Adoption of the GFD and GCFD in ASD**

Regardless of there being inconclusive evidence of a benefit to the GFD or GCFD in ASD, adoption of speciality diets is high. Studies assessing this also frequently attempt assessment of possible benefits of the diet primarily via cross-sectional analyses utilising symptom scales/survey responses, or anecdotal reporting from caregivers.

Bowers [94] reported that a majority of ASD referrals to their diet service regarded a suggestion to go on a GCFD (54.1%). Two of these 14 referrals later saw families of the patient report a "transformation" following adoption of the GFD diet ("One family described a 90% improvement and another family described an 'awakening' from a different level of consciousness"). A small comparison study [95] of children with ASD who were and were not following the GCFD reported that 7 of 13 children with ASD were already on a GCFD when recruited (outcome measures did not differ significantly from the 6 of 13 who were

not on the diet, however parents of all children on the GCFD reported that it had improved symptoms and behaviour). Babinska et al. [20] found 20.7% of children and adolescents with ASD to follow a diet which in some way restricted gluten (either GFD or GCFD); it was not found that the following of speciality diet correlated with GI symptom severity. Another study [96] found that 12% of their cohort of children with ASD consumed a GCFD, with these children also more likely to take supplements and overall showing better intake of nutrients including vitamin E, D and magnesium.

Hopf et al. [97] surveyed parents of children with ASD to identify reasons why they engaged with "complimentary and alternative medicine" (CAM). The GCFD had been used at some point by 54.8% of responders, although this was not rated among the interventions which were perceived as having had the greatest effectiveness (which included sensory integration therapy, melatonin and prescription antifungal medication). A similar study [98] also focused on the use of CAM in children with ASD as measured by caregiver report. Here, the GFD was followed at a lower rate (10% of the whole group), but was still the most common speciality diet followed. Rubenstein et al. [99] found 20.4% of children with ASD had ever used a GFD; those currently engaging with it had started on the suggestion of a medical professional in 50.7% of cases. Self-reported (from caregivers) data which predicted use of the GFD included GI conditions and developmental regression. Another experiment [100] examined all inpatients at a university medical centre, who did not have CD but who followed a GFD, to find predictors as to why in terms of comorbidities. In this, it was observed that having ASD led to an odds ratio of being on the diet of 23.42; by far the highest of all significant conditions reported (the next being irritable bowel syndrome with an OR of 6.16).

Studies which focus more directly on matching parental reporting of dietary practice to behavioural outcomes include Pennesi et al. [101]. Here, reports from 387 parents/caregivers of children with ASD were examined which focused on GI symptoms, suspected food sensitivities and adoption of speciality diets (primarily GCFD). Within these reports statistical effects were noted wherein greater suspicion of GI problems predicted greater improvement in ASD outcomes following adoption of speciality diets. Strict diet engagement was also observed to be significantly related to better outcomes. Another study [102] found no associations between dietary intake (which included measurement of gluten) and GI symptoms. However, these authors compared intake of gluten in grams against study outcome and a critical observation may be that gluten often needs to be eliminated entirely to usually see any benefit.

Some research has also focused more on the motivation in parents of children with ASD to adopt a GFD or GCFD. Marsden et al. [103] noted that parents who adopted these diets for their children with ASD were most influenced by "anticipated regret, positive outcomes and attitude". Perceived control was also relevant as a factor (with more predicting use of the diet). Tarnowska et al. [104] also investigated a similar question of what influenced parents of children with ASD to purchase GCFD foods. Packing features such as clear labelling that the food was e.g., gluten-free made them more likely to buy, while social issues around following exclusion diets (e.g., going out for a meal) and the expense/limited range of GCFD foods were seen as negative points. A survey study [105] found that approximately three quarters of clinical professionals who care for people with ASD had been asked at some point about the GCFD, while 29.5% of parents reported use of the GCFD specifically. Inadequacies with the knowledgebase regarding the use of speciality diets were noted by respondents.

Adoption of the GFD or GCFD in children with ASD is therefore quite pronounced, with lower estimates starting at 10%, and multiple studies reporting >50%. Motivations to engage with the diet appear to revolve around anticipated regret of negative outcomes should it not be tried, as well as a parent having a higher degree of perceived control. A recurring theme in a number of studies is the anecdotal reporting (e.g., by parents) of improvement in ASD outcomes, which are often isolated in incidence but apparently

dramatic in effect. Caregivers and clinicians each highlight that greater understanding of how these diets interact with ASD is required.

#### **8. Nutritional Considerations**

Limited research has also studied the impact of the GFD or GCFD on the nutritional health of children with ASD. Studies which indicate a positive consequence of following the GFD/GCFD on health include one by Herndon et al. [106]. While this focused on comparisons between (all) children with ASD compared to TD children, a subgroup analysis revealed those with ASD who followed a GCFD had higher vitamin E intake than those who did not. As already discussed, Stewart et al. [96] found following a GCFD led to higher levels of vitamin D, E and magnesium, possibly relating to a higher likelihood of simultaneously using supplements compared to those following a regular diet. Supporting this, another study [107] found that those on a GCFD were far more likely to take vitamin D and calcium supplements; no child who followed the GCFD had a deficiency of 25(OH)D (a marker of bone health), compared to 24% of those on a regular diet who did.

Studies of no or mixed outcomes include one [108] where no difference was found in nutritional intake between children with ASD who did and did not follow a GCFD (although an overall effect of suboptimal intake was noted across the whole cohort). Analysing food diaries, Mari-Bauset et al. [109] also reported mixed outcomes wherein children who followed a GCFD had lower BMI and energy, and lower intake of some nutrients (sodium, calcium, phosphorus and pantothenic acid). Conversely however, they had better intake of fiber, legumes, vegetables and fat.

Studies reporting negative impacts of the diet include Arnold et al. [110], who found a trend for children with ASD who followed the GCFD to have more deficiencies relating to essential amino acids, including tryptophan. This was replicated by another investigation [111] which also detected lower tryptophan in children with ASD compared to TD controls (it was lowest in those with ASD who followed a restricted diet). These authors hypothesised this may lead to a worsening of ASD symptoms.

#### **9. Synthesis of Literature and Future Directions for Research**

The interest of speciality diets in ASD, and particularly the GCFD, has increased markedly in the last 2 decades both in terms of adoption in the ASD community as well as scientific study. Research does convincingly demonstrate certain effects and associations which justify this. Most notably is a modest comorbidity between CD and ASD. Also shown is physiological evidence of inadequate digestion of gluten in people with ASD, leading to elevated "exorphins"/gluten antibodies around which reasonable hypotheses exist regarding downstream negative consequences on the central nervous system. Whether these associations are because of a unique relationship between CD and ASD or because of a predisposition in people with ASD to have a generally higher rate of autoimmune-like features, is yet to be resolved. Further research which directly examines the effects of these gluten products is needed in people with ASD, while additional studies examining shared genetic predispositions would also be warranted and beneficial. There is also a scarcity of epidemiological research characterising the comorbidity of ASD and CD; while one very well powered study does exist and supports this to be the case replication elsewhere is desirable. The availability of newer gluten-related antibodies (e.g., TG6) and the use of native antigliadin antibodies that are known to be sensitive to the whole spectrum of gluten-related disorders, may provide a good opportunity for further large scale epidemiological studies.

Arguably the greatest gap in current literature relates to the lack of tightly designed trials of the GFD, or GCFD in people with ASD. Those that are currently available suffer from very pronounced heterogeneity regarding the intervention followed, sample size, trial duration, blinding and outcomes measured. There does not yet exist an RCT which combines an at-least modest sample size with a placebo-controlled design over a long duration. This would be highly valuable to address, accepting the limitations of the

difficulties of such an intervention (gluten free diet) in the context of what is a behaviourallycomplex cohort of patients.

Adoption of the GFD and GCFD appears to be very high amongst people with ASD. An impression is gained of strong anecdotal evidence of a benefit in relevant studies, although statistical associations do not as often bear this out. It is also unclear if any reported behavioural benefits would be because of a direct interaction between physiological gluten/casein-related impacts on the brain, or if engaging with speciality diets simply reduce non-specific GI symptoms and thus improve quality of life in a more general sense.

The nutritional impact of a GCFD on the child with ASD generally seems to be slight, or even associated with improved intake. However, some studies showing deficiencies of certain nutrients highlight the need to still maintain a balanced diet once on a restricted one. Finally, a general observation is the abundance of research which focuses on children. Of the initial papers found in the literature review, 68 (of 79) studied groups which were exclusively children, or mixed children and adolescents. While this is likely an outcome of opportunity and convenience sampling effects, it is nonetheless a strong bias within the available literature and means that generalisation of anything discussed in this review to adults with ASD is difficult.

#### **10. Conclusions**

This review highlights a modest comorbidity between ASD and CD, and a base of evidence on which reasonable hypotheses may be built to explore if gluten has a generally adverse effect in exacerbating the symptoms and quality of life in children with ASD. However, a negative effect of gluten ingestion in ASD has not been proved. Trials which have sought to demonstrate this are variable in their findings, and suffer from issues with experimental design and execution which means that an overall interpretation cannot yet be made. Efforts should focus on future studies which address limitations detailed here to create an RCT from which confident conclusions can be drawn. As diets which restrict gluten see a very high adoption rate among people with ASD, such further research is certainly warranted.

**Author Contributions:** I.D.C.: methodology, data curation, writing-original draft preparation, writing-review and editing; N.H.: writing-review and editing, supervision; M.H.: conceptualization, writing-review and editing, supervision. All authors have read and agreed to the published version of the manuscript.

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

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

#### **References**


### *Article* **Stiff Person Syndrome and Gluten Sensitivity**

#### **Marios Hadjivassiliou 1,\*, Panagiotis Zis 1, David S. Sanders 2, Nigel Hoggard <sup>3</sup> and Ptolemaios G. Sarrigiannis <sup>1</sup>**


**Abstract:** Stiff person syndrome (SPS) is a rare autoimmune disease characterised by axial stiffness and episodic painful spasms. It is associated with additional autoimmune diseases and cerebellar ataxia. Most patients with SPS have high levels of glutamic acid decarboxylase (GAD) antibodies. The aetiology of SPS remains unclear but autoimmunity is thought to play a major part. We have previously demonstrated overlap between anti-GAD ataxia and gluten sensitivity. We have also demonstrated the beneficial effect of a gluten-free diet (GFD) in patients with anti-GAD ataxia. Here, we describe our experience in the management of 20 patients with SPS. The mean age at symptom onset was 52 years. Additional autoimmune diseases were seen in 15/20. Nineteen of the 20 patients had serological evidence of gluten sensitivity and 6 had coeliac disease. Fourteen of the 15 patients who had brain imaging had evidence of cerebellar involvement. Twelve patients improved on GFD and in seven GFD alone was the only treatment required long term. Twelve patients had immunosuppression but only three remained on such medication. Gluten sensitivity plays an important part in the pathogenesis of SPS and GFD is an effective therapeutic intervention.

**Keywords:** stiff person syndrome; anti-GAD antibodies; gluten sensitivity; coeliac disease; cerebellar ataxia; gluten free diet

#### **1. Introduction**

Glutamic acid decarboxylase (GAD) is the enzyme involved in the synthesis of the inhibitory neurotransmitter gamma-aminobutyric acid (GABA). GAD is found in both the central and peripheral nervous systems and in pancreatic beta cells [1]. GAD antibodies were first detected and characterised in children with newly diagnosed insulin dependent diabetes mellitus (IDDM) [2]. These were shown to be reacting with pancreatic islet cell proteins.

The first neurological disease to be associated with anti-GAD antibodies was stiffperson syndrome (SPS) [3]. SPS is a very rare autoimmune neurological disease, clinically characterised by axial rigidity, often resulting in hyperlordosis, painful spasms and anxiety. It belongs to a spectrum of CNS hyperexcitability syndromes. SPS is often associated with additional autoimmune diseases such as hypothyroidism, IDDM, pernicious anaemia and others. The majority of patients with SPS have anti-GAD antibodies. Anti-GAD antibodies have also been found in some cases of sporadic idiopathic ataxias [4]. Their presence implies an autoimmune pathogenesis raising the possibility of therapeutic interventions with immunosuppressive medication.

We have previously made a connection between anti-GAD associated diseases and gluten sensitivity (GS) including coeliac disease (CD) [5]. We were also able to show considerable overlap between anti-GAD ataxia and gluten ataxia (70% of patients with anti-GAD ataxia are gluten sensitive), and we have demonstrated that gluten free diet

**Citation:** Hadjivassiliou, M.; Zis, P.; Sanders, D.S.; Hoggard, N.; Sarrigiannis, P.G. Stiff Person Syndrome and Gluten Sensitivity. *Nutrients* **2021**, *13*, 1373. https:// doi.org/10.3390/nu13041373

Academic Editor: Giacomo Caio

Received: 30 March 2021 Accepted: 18 April 2021 Published: 20 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

(GFD) is an effective therapeutic intervention in such patients [6]. In this report we share our experience in managing and treating patients with SPS and in particular highlighting the overlap between SPS, gluten sensitivity and CD as well as reporting the therapeutic effect of gluten free diet (GFD).

#### **2. Methods**

This report is based on a retrospective observational case series of patients regularly attending our specialist clinics (GS/neurology, neuroimmunology and ataxia). The South Yorkshire Research Ethics Committee has confirmed that no ethical approval is indicated given that all investigations/interventions were clinically indicated and did not form part of a research study. All patients were identified from these clinics by one of the authors (MH) who is in charge of the clinical care of all these patients. The patients have been looked after for over 25 years and are under regular follow up by the same consultant neurologist. The diagnosis of SPS was based on the typical clinical features (stiffness, axial rigidity, episodic painful spasms) in addition to the presence of high titre of anti-GAD antibodies (>2000 U/mL) and neurophysiological evidence of CNS hyperexcitability (continuous motor unit activity on EMG and/or abnormal blink reflex).

Serological testing in addition to anti-GAD antibodies, included antigliadin antibodies (AGA, Phadia), TG2 (Phadia), endomysium antibodies (EMA, Werfen) and TG6 antibodies (Zedira). Those patients with one or more positive antibodies were offered gastroscopy and duodenal biopsy to establish the presence of enteropathy (triad of villous atrophy, crypt hyperplasia, increased intraepithelial lymphocytes). All patients with positive serology for gluten sensitivity were advised to adopt a GFD irrespective of the presence of enteropathy. They were all reviewed by an experienced dietitian and given detail advice on GFD. Depending on clinical response after GFD some patients were also offered treatment with immunosuppression. This included intravenous immunoglobulins, azathioprine, mycophenolate, rituximab, plasma exchange and cyclophosphamide.

All patients underwent brain imaging with MRI, some also had MR spectroscopy of the cerebellum. Cerebellar involvement in the context of SPS is almost universal but often under-reported by patients because their most disabling symptoms are those of stiffness and painful spasms.

#### **3. Results**

We identified 20 patients with SPS over the last 25 years. There were 11 female and 9 male patients. Mean age at onset of symptoms was 52 (range 37–69 years). The presenting symptoms included primarily leg stiffness in 12, truncal stiffness in 5, painful spasms in 2 (painful spasms became a prominent feature in most patients later on in the disease), ataxia in 2 (later a feature in 17 patients) and one leg stiffness in one. Additional autoimmune diseases (apart from GS and CD) were present in 15 patients with 11 having hypothyroidism, 7 having IDDM, 2 having myasthenia gravis, 2 having Sjogren's syndrome, 1 pernicious anaemia and 1 psoriatic arthropathy (some patients had more than one autoimmune disease). Serological evidence of GS (one or more of AGA, EMA, TG2 and TG6 antibodies) were found in 19 of the 20 patients (95%). The diagnosis of gluten sensitivity was made in Sheffield by the authors and only one of the patients had a preexisting diagnosis of CD. Fourteen patients underwent gastroscopy and duodenal biopsy. Six patients had evidence of CD on biopsy, the remaining 8 had a normal mucosa. Only 3 of the 19 patients with GS/CD had any gastrointestinal symptoms (diarrhoea, bloating, abdominal pain) attributed to GS/CD. All of these 3 had coeliac disease on biopsy and in all 3 the gastrointestinal symptoms improved on a GFD. The above results are summarised in Table 1.


**Table 1.** Clinical characteristics, investigations and outcomes of 20 patients with stiff person syndrome.

Please note that not all patients were tested for TG6 antibodies. (IDDM-insulin dependent diabetes mellitus, AGA-antigliadin antibodies, TG6- transglutaminase 6 antibodies, TG2-transglutaminase 2 antibodies, EMAendomysium antibodies).

Neurophysiology showed continuous motor unit activity in keeping with SPS in 13 of the patients. Three patients did not have neurophysiology but the clinical phenotype in combination of high anti-GAD was sufficient to enable the diagnosis. The remaining 4 patients had normal neurophysiology, but this was done after the patients were established on antispasmodic medication. Abnormal blink reflex was seen in 2 patients, but this was only performed in 4 patients.

GFD was recommended to all 19 patients with gluten sensitivity. In 12/19 patients (5 with CD and 7 with GS and no enteropathy) the GFD was found to be beneficial to their SPS and ataxia symptoms (reduced frequency of spasms, stabilisation of mobility, rigidity and improved ataxia). Five patients did not derive any benefit. Two patients did not go on GFD. In 7 patients (3 with CD and 4 with GS and no enteropathy) GFD alone was the only treatment needed to keep their symptoms under control and stabilise their condition. Immunosuppression was used in 12 patients. This included intravenous immunoglobulins (IVIgs) in 9, mycophenolate in 5, plasma exchange in 2, azathioprine in 1, rituximab in 1 and cyclophosphamide in 1 (some patients had more than one immunosuppressive medication). One patient underwent autologous stem cell transplantation because nothing else was effective. This resulted in stabilisation [7]. Only one of the 9 patients that received immunoglobulins found it beneficial long term. In the remaining, oral immunosuppression was felt to be more effective than repeat IVIg's. All apart from 3 patients were taking antispasmodic medication. Six on 3 different antispasmodics, 5 on 2 and 6 on one. The most commonly used antispasmodics were baclofen (14), dantrolene (9), diazepam (8) and tizanidine (1). One patient with very resistant disease also had botox injections and found entonox very helpful for the painful spasms. Ten patients were on Gabapentin. The addition of this medication seemed to offer some additional benefit in partially controlling/stabilising their symptoms.

MR spectroscopy of the cerebellum was done in 15 patients. Evidence of reduced NAA/Cr ratio in the vermis and/or hemisphere was seen in 14 (mean NAA/Cr from the vermis was 0.9 (range 0.79–1.12) and from the hemisphere 0.95 (range 0.74 to 1.1), normal ratio should be above 1, in keeping with cerebellar dysfunction. Only one of the 15 patients had normal spectroscopy (NAA/Cr vermis 1.03 and hemisphere 1.23). Follow up MR spectroscopy was available in 10 patients who were also GS (5) or had CD (5), after staring GFD (no immunosuppression). Eight showed improved MR spectroscopy (NAA/Cr are ratio) from the vermis whilst 2 showed deterioration.

At the time of writing this report 5 patients had died. Three as a result of SPS (2 in hospital with complications of immobility, one had respiratory arrest resulting in brain hypoxia during a severe chest spasm causing respiratory arrest). One died of myocardial infarction (also had IDDM) and the other due to COVID pneumonia. Of the remaining 15 patients 11 remain mobile with minimal walking aids (use of one stick) and 4 are wheel-chair bound.

Out of the 15 patients who are still alive, 12 are still on a strict GFD, and only 3 are on immunosuppression (1 regular IVIgs, 2 on mycophenolate).

#### *Illustrative Clinical Case*

This 60-year-old man was referred to our centre 4 years ago with an established diagnosis of SPS. He had received high dose steroids on several occasions after being admitted with painful spasms that rendered him bed bound. The episodes of severe spasms had become so disabling that the patient had become understandably anxious and fearful of leaving the house.

At the time of the initial assessment, he was on no regular immunosuppression, but the referring neurologist was planning the introduction of IVIgs. On clinical examination in addition to severe stiffness in both legs and exaggerated lordosis he had evidence of incoordination with nystagmus on lateral gaze, finger nose and heel to sheen ataxia. He also had gait ataxia. He was using a wheelchair when out of his home.

He had a history of hypothyroidism. Investigations showed high titre of anti-GAD antibodies (>2000 U/mL), with neurophysiological evidence of continuous motor unit activity on EMG which, in combination with the clinical presentation led to the diagnosis of SPS.

In addition to the high levels of anti-GAD antibodies he was positive for AGA, EMA, TG2 and TG6 antibodies. Duodenal biopsy confirmed the presence of CD. MR spectroscopy confirmed abnormal NAA/Cr ratio over the vermis in keeping with the clinical findings of gait ataxia (Figure 1). He started on a strict GFD. Within the first 6 months he observed significant reduction in the painful spasms. This was without any other medication. His mobility improved and he was now able to walk using a frame and also go out of his home. After a year on GFD his gluten sensitivity-related antibodies were no longer present. Repeat MR spectroscopy of the cerebellar vermis showed improved NAA/Cr ratio (from 0.84 to 0.90, normal ratio should be over 1) reflecting the improved mobility. He is still stable (3 years after the introduction of GFD) and not requiring any immunosuppression.

**Figure 1.** Magnetic Resonance Spectroscopy of the cerebellar vermis from the illustrative clinical case (see text) showing a significant reduction of the N-Acetyl-Aspartate to Creatine ratio (NAA/Cr) at 0.84 (normal should be above 1). All but one of these patients with stiff person syndrome (SPS) had abnormal spectroscopy of the cerebellum highlighting the fact that cerebellar involvement is universal in SPS.

#### **4. Discussion**

We have previously reported an association between anti-GAD related diseases and GS/CD [5]. We have also recently published our experience in the management of 50 patients with anti-GAD ataxia where we have again shown a significant overlap between anti-GAD ataxia and gluten ataxia (70% of patients with anti-GAD ataxia were gluten sensitive) [6]. Furthermore, we have shown that patients with anti-GAD ataxia who are gluten sensitive respond well to strict GFD, with improvement of the ataxia. In this report we present our experience of managing 20 patients with SPS, primarily to highlight the strong association with gluten GS/CD (95% positive for one or more gluten sensitivity-related antibodies and at least 30% having CD) and demonstrate that GFD has a therapeutic role to play. The prevalence of AGA antibodies in the healthy population was 12% and that of CD 1% [8].

The association between SPS and gluten sensitivity cannot be simply explained on the basis of an association of two autoimmune diseases by chance. Nineteen of these 20 patients (95%) with SPS were found to be gluten sensitive. In addition, we have shown that GFD has an important therapeutic role to play in these patients, suggesting that GS/CD must play a role in the pathogenesis of SPS.

As per the case highlighted in this report, the majority of patients who are on GFD have found this intervention helpful in controlling their SPS symptoms.

In our experience, cerebellar involvement in the context of SPS appears to be universal in these patients. In fact, cerebellar ataxia in isolation is a commoner manifestation of anti-GAD related diseases than SPS based on our experience; the number of patients with anti-GAD ataxia we have treated is 50 as opposed to 20 with SPS. However, there may be some referral bias given that our unit is one of the National Ataxia Centres in the UK.

Cerebellar involvement in the context of SPS may have an important pathophysiological role to play; the output of the cerebellum is all inhibitory. Any dysregulation of such output could potentially result in a state of CNS hyperexcitability. This state of hyperexcitability is particularly prominent in the immune ataxias by contrast to the genetic or degenerative ataxias [9]. It can manifest with rigidity and spasms, as is the case in SPS but also with cortical myoclonus as is often seen in cases of refractory CD [10]. Additional clinical markers of hyperexcitability include brisk reflexes and exaggerated startle. It is possible that the selective involvement of different cerebellar cell populations may explain why the state on brain hyperexcitability is more often seen in immune rather than genetic ataxias.

The association between anti-GAD antibodies and gluten sensitivity merits further consideration. Ventura et al. have made the observation that the prevalence of additional autoimmune diseases in children with CD is significantly lower than in those patients with CD diagnosed in adulthood [11]. They concluded that GFD may reduce the risk of developing additional autoimmune diseases later on in life. This observation echoes our previously observed reduction in anti-GAD antibodies in patients with anti-GAD related diseases and gluten sensitivity who go on a strict GFD [5].

The pathological role of anti-GAD antibodies in the genesis of SPS and ataxia is unclear. Since GAD65 (the GAD isoform implicated in these diseases) is intracellular and is associated with a range of neurological conditions, some have argued that anti-GAD65 antibodies have no pathogenic role to play. On the other hand, recent physiological studies in vitro and in vivo have demonstrated that binding of GAD by anti-GAD antibodies induces loss of GAD functions relating to GABA release, leading to the development of cerebellar ataxia [12]. Given these observations the question remains as to why should GFD be beneficial in those patients with gluten sensitivity and SPS. The response to GFD in those patients who are gluten sensitive suggests that gluten sensitivity may be driving the immune response that results in SPS.

There are no clear-cut evidence-based guidelines for the treatment of SPS. It has been shown that regular IVIgs can be beneficial [13]. This has not been our experience. Only 2 of the 20 patients reported here found regular IVIgs of some sustained benefit. Immunosuppression with mycophenolate has been beneficial for some patients but by far the most effective therapeutic intervention has been the GFD.

In terms of symptom relief most patients require a combination of one or often several antispasmotics of which in our experience dantrolene, baclofen and diazepam offer the best combination. We also have used gabapentin more as a "disease modifier" simply on the theoretical benefit based on its mode of action.

To conclude, 95% of our patients with SPS who have been under regular review at our centre have evidence of gluten sensitivity or CD and have benefited from a strict GFD. The diagnosis of gluten sensitivity relies on a range of antibodies and the use of EMA or TG2 antibodies alone, whilst sufficient to diagnose CD, cannot diagnose gluten sensitivity without enteropathy. This is an important consideration because GFD is beneficial for patients with SPS who are gluten sensitive irrespective of the presence of enteropathy.

**Author Contributions:** M.H. identified the link between gluten sensitivity and SPS and conceptualised this report as well as looked after all the patients. He produced the first draft. P.Z. and P.G.S. performed all the neurophysiology. D.S.S. performed all the gastroscopies and biopsies. N.H. was responsible for the imaging. All authors reviewed and contributed to the final version of the paper. All authors have read and agreed to the published version of the manuscript.

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

**Institutional Review Board Statement:** This article summarizes observational data from patients attending a specialised neurology clinic, Sheffield Teaching Hospitals NHS Trust, Sheffield, UK. The South Yorkshire Research Ethics Committee confirmed that no ethical approval is indicated given that all investigations were clinically indicated and did not form part of a research study.

**Informed Consent Statement:** The South Yorkshire Research Ethics Committee has confirmed that no ethical approval is indicated given that all investigations/interventions were clinically indicated and did not form part of a research study.

**Data Availability Statement:** Anonymised data can be provided on request.

**Acknowledgments:** This is a summary of independent work carried out at the NIHR Sheffield Biomedical Research Centre (Translational Neuroscience). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR.

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

#### **References**

