Case-control study examining the composition of the gut microbiome in individuals with and without immune-mediated dry eye

Abstract

Purpose

Gut microbiome alterations have been associated with various autoimmune diseases. There is limited data, however, on relationships between gut dysbiosis and immune-related dry eye (DE). Our aim was to compare the gut microbiome composition of individuals with early and late markers of Sjögren’s Syndrome (SS) to controls without DE.

Methods

We compared 20 individuals with positive early markers (anti-salivary protein 1 (SP1), anti-parotid secretory protein (PSP), anti-carbonic anhydrase 6 (CA6) IgG, IgA, IgM, n=19) or late markers (anti-Ro/SS-A, anti-La/SS-B, n=1) of SS with no comorbid autoimmune diagnoses and 20 age- and sex-matched controls. Collected stool samples underwent deep RNA sequencing. The main outcomes measured included gut microbiome composition and diversity.

Results

A total of 20 cases (Dry Eye Questionnaire-5 (DEQ-5) 15.2±3.4, Ocular Surface Disease Index (OSDI) 55.1±22.8, Schirmer 7.1±5.2 mm) were compared to 20 controls (DEQ-5 4.8±3.8, OSDI 14.2±12.3, Schirmer 20.4±9.2 mm). No differences were observed in α-diversity (p=0.97) or overall community structure (p=0.62). Between groups, 32 species were differentially abundant (p<0.01). Among cases, 27 were relatively more abundant, including 10 Lactobacilli and 4 Bifidobacteria species. A relative depletion of 5 species was found in cases compared to controls, notably Fusobacterium varium and Prevotella stercorea.

Conclusion

Differences in gut microbiome composition were found in individuals with mostly early markers of SS compared to controls. However, their clinical significance to DE manifestations remains unclear. Further studies are needed to elucidate the role of gut dysbiosis on immune dysregulation and disease activity in the various forms of immune-mediated DE.

INTRODUCTION

Dry eye (DE) is a multi-factorial ocular surface disease that often has inflammatory contributors.¹ Inflammation in DE can be localized to the ocular surface or be comorbid with a systemic immune disease, most commonly Sjögrens syndrome (SS). Unfortunately, many patients with SS develop vision-threatening complications before diagnosis and treatment initiation.² In addition, treatment goals for immune-mediated DE are often unmet with the current therapeutic options, highlighting the need to research new interventions.³

In recent years, numerous studies have demonstrated that individuals with a variety of autoimmune diseases, including rheumatoid arthritis (RA), inflammatory bowel disease (IBD), spondylarthrosis, Behcet’s, and systemic lupus erythematosus (SLE), have alterations in their gut microbiome composition compared to controls without disease.⁴ Gut dysbiosis has also been demonstrated in individuals with primary SS (pSS) (Table 1), a condition closely linked to DE.^5–11 However, many individuals with DE do not fulfill criteria for primary or secondary Sjögrens, yet have evidence of an immune contribution to disease in the form of early SS markers,⁹ including autoantibodies to salivary protein 1 (SP1), parotid secretory protein (PSP), and carbonic anhydrase 6 (CA6).¹² In fact, early marker positivity is more common than late marker positivity (antibodies to Ro and La) in individuals with DE, found in up to 73% of individuals with DE symptoms and suggestive clinical signs (e.g., aqueous tear deficiency, ocular surface inflammation).¹³

Table 1:

Literature review of studies on the gut microbiome in immune-mediated dry eye, including Sjögren’s syndrome

Author, yr	Case group(s)	Control group	Sequencing method	Diversity	Increased relative abundance in cases	Decreased relative abundance in cases
de Paiva, 2016	pSS (n=10), all female	Healthy controls from Human Microbiome Project (n=45), majority male and significantly younger	16S rRNA, V4	NA	Genera: Pseudobutyrivibrio, Escherichia/Shigella, Blautia, Streptococcus	Genera: Bacteroides, Parabacteroides, Faecalibacterium, Prevotella
Mandl, 2017	pSS (n=42)	Age- and sex-matched controls (n=35)	16S rRNA; 54 bacterial probes	NA	None	Genera: Alistipes, Bifidobacterium
van der Meulen, 2019	pSS (n=39); SLE (n=30)	General population controls (n=965), significantly younger and significantly more males	16S rRNA, V4 and 806R	Decreased alpha diversity; No difference in beta diversity	Phyla: Bacteroidetes, Proteobacteria; Genera Bacteroides, Alistipes; Species: Bacteroides vulgatus, Bacteroides uniformis, Bacteroides ovatus	Species: Clostridium sensu stricto
Cano-Ortiz, 2020	pSS (n=19), all female	Age-, sex-, BMI-matched healthy controls (n=19)	16S rRNA, V2-V9	Decreased alpha diversity; Beta diversity showed tighter clustering of cases compared to controls	Phyla: Bacteroidetes, Proteobacteria; Families: Clostridiaceae, Veillonellaceae, Prevotellaceae, Rikenellaceae, Odoribacteraceae, Enterobacteriaceae; Genera: Prevotella, Escherichia, Clostridium, Enterobacter, Veillonella, Streptococcus; Species: Clostridium clostridioforme, E. coli, Prevotella copri	Phyla: Firmicutes, Actinobacteria; Families: Ruminococcaceae, Lachnospiraceae, Bacteroidaceae, Porphyromonadaceae, Bifidobacteriaceae; Genera: Bacteroides, Alistipes, Dorea, Parabacteroides, Blautia, Lachnospira, Roseburia, Faecalibacterium, Ruminococcus, Bifidobacterium; Species: Bacteroides fragilis, Parabacteroides distasonis, Dorea longicatena, Ruminococcus lactaris, Faecalibacterium prausnitzii
Mendez, 2020	pSS (n=10), sSS (n=3); early markers (n=8)	Healthy controls from stool bank (n=21), all male and significantly younger	16S rDNA, V4-V5	No difference in Shannon’s alpha diversity; Increased alpha diversity using Faith’s phylogenetic diversity; Beta diversity showed distinct clustering of controls compared to cases	Phyla: Bacteroidetes, Actinobacteria, Proteobacteria; Families: Actinomycetaceae, Eggerthellaceae, Lactobacillaceae, Akkermansiaceae, Coriobacteriaceae, Eubacteriaceae; Genera: Prevotella, Megasphaera, Parabacteroides	Phyla: Firmicutes; Orders Clostridiales, Bacteroides; Families: Ruminococcaceae, Lachnospiraceae; Genera: Faecalibacterium, Veillonella
Moon, 2020	pSS (n=10); environmental DE (n=14)	Healthy controls (n=12), unmatched but no significant difference in sex or age	16S rRNA, V3-V4	No difference in alpha diversity; Beta diversity showed separate clustering between cases and controls	Phyla: Bacteroidetes, Actinobacteria; Genera: Veillonella, Prevotella, Odoribacter, Alistipes	Classes: Clostridia; Genera: Bifidobacterium, Blautia, Dorea, Agathobacter, Subdoligranulum
Watane, 2021	pSS or sSS (n=5); early markers (n=5)	Healthy, young stool donor for FMT (n=1)	16S rDNA, V4-V5	Increased alpha diversity; Beta diversity showed separate clustering between cases and the control	Phyla: Actinobacteria, Bacteroidetes, Cyanobacteria, Firmicutes, Proteobacteria, Verrucomicrobia; Genera: Alistipes, Streptococcus, Blautia	Phyla: Euryarchaeota, Fusobacteria; Genera: Faecalibacterium, Prevotella, Ruminococcus

Abbreviation: pSS, primary Sjogren’s syndrome; sSS, secondary Sjogren’s syndrome; SLE, systemic lupus erythematosus; FMT, fecal microbial transplant; yr, year; NA, not available

Overall, intestinal dysbiosis can manifest as a loss of beneficial microbes, expansion of pathogenic microbes, and/or loss of microbial diversity. The role of the dysbiosis in promoting autoimmunity is not fully understood,¹⁴ but it is theorized that dysbiosis may lead to an imbalance in T-cell homeostasis, with excessive effector T-cell responses and inflammation. Understanding the role of gut dysbiosis on disease manifestations in various DE sub-types (e.g., early markers, pSS, and secondary SS (sSS)) is important as modulating the gut microbiome through dietary changes, probiotics, or fecal microbial transplant (FMT) may be a potential therapy in the proper population. In this study, we compared the gut microbiome composition of cases with immune-associated DE and no comorbid auto-immune disease to age- and sex-matched controls without DE.

METHODS

Study population

A total of 40 individuals seen at the Miami Veterans Affairs (VA) Hospital eye clinic participated in the study between January 2019 and September 2021. No formal sample size calculation was performed, and 20 patients per arm was chosen based on clinical feasibility. Cases included individuals with DE symptoms and/or signs, who had pSS as defined by the 2016 American College of Rheumatology criteria¹⁵ (total weighted-score of ≥4: (a) anti-SSA/Ro and anti-SSB/La antibody positivity; focal lymphocytic sialadenitis with focus score ≥1 foci/4mm², each scoring 3; (b) ocular staining score of ≥5; Schirmer ≤5 mm/5 minutes; unstimulated salivary flow rate of ≤0.1 ml/minute, each scoring 1, n=1), or had an elevation of at least one of nine early SS markers (IgM, IgA, or IgG antibodies against SP1 (n=5), PSP (n=7), or CA6 (n=13) at a level of 20 EU/ml or greater, n=19).¹³ In this study, 12 cases had one elevated early SS marker, six cases had two elevated markers, and one case had four elevated markers. Controls included age- and sex-matched individuals without DE symptoms or signs, including TBUT ≥5 seconds, corneal staining score ≤2, and Schirmer score ≥5 mm in both eyes. Individuals with sSS, comorbid thyroid disease, or autoimmune diseases, such as RA, IBD, or SLE, were excluded from the study, as the presence of these diseases could confound the gut microbiome profile. Additionally, individuals who were taking antibiotics at the time of the study were excluded for similar reasons.

Clinical metrics

Demographic information for each participant was collected including age, gender, race, and ethnicity. In addition, information was collected on past ocular and medical history, current medications, and current and past usage of probiotics.

DE symptoms

Participants completed two standardized DE symptom questionnaires: the Dry Eye Questoinnaire-5 (DEQ 5, score 0-22)¹⁶ and the Ocular Surface Disease Index (OSDI, score 0-100).¹⁷

DE signs

Participants underwent an ocular surface exam of both eyes which included:

1. Tear breakup time (TBUT) using fluorescein stain measured three times in each eye and averaged.

2. Corneal staining using fluorescein graded to the National Eye Institute (NEI) scale which assesses 5 areas of the cornea on a 0-3 scale (total scale 0-15).¹⁸

3. Basal tear production after anesthesia placement (measured in mm at 5 min) using Schirmer’s strips.

Sample collection and processing

All subjects collected approximately one gram of his or her solid stool and placed the sample in a DNA/RNA Shield^™ fecal collection tube (Zymo Research, Irvine, California). Subjects then mailed the sample to the Miami VA at ambient temperature (4°C-25°C). Samples were then packaged in dry ice and shipped to the Proctor Foundation. Before sample processing, samples were de-identified and all laboratory personnel were masked to the identity of the samples.

Sequencing and statistical analyses

All 40 fecal samples underwent deep RNA sequencing to evaluate their microbiome. We intentionally used deep RNA sequencing because the technique interrogates all actively replicating genomic sequences, and therefore has less risk of varying results compared to 16S-targeted metagenomics which only interrogates specific hypervariable regions.^19,20 Sequencing libraries were prepared and sequenced as previously described.²¹ Total RNA was extracted using the ZymoBIOMICS DNA/RNA Miniprep kit (Zymo Research, Irvine, California). 25ng of extracted total RNA was converted to double-stranded complementary DNA (cDNA). The cDNA was converted to Illumina sequencing libraries using the NEBNext Ultra^™ II DNA library preparation kit (New England Biolabs (NEB), Ipswich, Massachusetts) per manufacturer’s recommendations and amplified with 10 polymerase chain reaction (PCR) cycles. Samples were sequenced on the NovaSeq 6000 instrument (Illumina, San Diego, California) using 125 base paired-end sequencing. Sequencing data were analyzed using a rapid, in-house computational pipeline to classify sequencing reads by comparison to the entire National Center for Biotechnology Information nucleotide reference database.^21,22 DESeq2 was used to perform differential abundance analysis on the microbiota between groups.²³ A false discovery rate (FDR) less than 0.01 and log₂ fold change (FC) greater than 1.5 were considered as notable. Topconfects algorithm was used to determine the confident effect sizes.²⁴ A permutational multivariate analysis of variance (PERMANOVA) on the Euclidean distance (L₂-norm) was performed to determine a difference in bacterial community structures. In addition, α-diversity was compared at the species level using inverse Simpson’s and Shannon’s indices.

For clinical characteristics, statistical analyses were performed using SPSS Statistics version 25.0 (IBM Corp, Armonk, NY). Tables and graphs were created using Microsoft Excel version 1904 (Office 365 ProPlus, Redmond, Washington). Normality of the data was assessed using the Shapiro-Wilk test. Descriptive statistics for categorical variables included means and standard deviations. Group differences for categorical variables were assessed with Pearson chi-square or Fisher’s exact tests, and independent-sample t-tests for continuous variables. Spearman’s rank correlation was used to assess the relationship between DE signs and the abundances of five selected microbiota for both the case and control groups. A p-value less than 0.05 was considered statistically significant. In this paper, we opted to give information on all variables being compared as opposed to correcting the p-value (e.g. Bonferroni) since the latter methodology has its own limitations.²⁵

RESULTS

Study population

A total of 40 individuals were included in the study: 19 cases with elevated early SS markers, one case meeting pSS criteria, and 20 sex- and age-matched controls. The patient demographics are summarized in Table 2. Overall, the groups were matched with respect to age (mean 60±12 years), gender (68% male), race (83% White), and ethnicity (38% Hispanic). Comorbidities included hypertension (n=15) and diabetes mellitus (n=7). No significant differences were found for demographics or comorbidities between case and control groups.

Table 2:

Clinical characteristics of the study population

Variable	No. of patients		p-value	All individuals (n=40)
Demographics	Cases (n=20)	Controls (n=20)
Age, years, mean age ± SD (range)	61 ± 12.8 (37-80)	59 ± 11.7 (40-77)	0.65	60 ± 12 (37-80)
Gender, male, n (%)	14 (70%)	13 (65%)	0.74	27 (68%)
Race, white, n (%)	18 (90%)	15 (75%)	0.41	33 (83%)
Ethnicity, Hispanic, n (%)	8 (40%)	7 (35%)	0.74	15 (38%)
Hypertension, n (%)	6 (30%)	9 (45%)	0.33	15 (38%)
Diabetes mellitus, n (%)	1 (5%)	5 (25%)	0.18	6 (15%)
Former smoker, n (%)	7 (35%)	10 (50%)	0.34	17 (46%)
Current smoker, n (%)	1 (5%)	1 (5%)	1.00	2 (5%)
Current probiotic use a, n (%)	5 (26%)	1 (6%)	0.18	6 (16%)
Dry eye symptoms, mean ± SD (range)
DEQ5	15.2 ± 3.4 (9-20)	4.8 ± 3.8 (0-12)	< 0.0001***	10.0 ± 6.3 (0-20)
OSDI	55.1 ± 22.8 (12.5-90)	14.2 ± 12.3 (0-47.9)	< 0.0001***	34.6 ± 27.5 (0-90)
Dry eye signsb, mean ± SD (range)
Tear break up time	4.9 ± 3.4 (1-15)	13.3 ± 2.4 (9-15)	< 0.0001***	9.0 ± 5.2 (1-15)
Corneal staining	4.4 ± 4.1 (0-12)	0.3 ± 0.6 (0-2)	0.0001***	2.3 ± 3.6 (0-12)
Schirmer score	5.8 ± 4.2 (0-15)	18.6 ± 9.5 (6-35)	< 0.0001***	12.2 ± 9.7 (0-35)
Dry eye therapiesc, n (%)
Artificial tears	18 (90%)	--	--	--
Autologous serum tears	9 (45%)	--	--	--
Nocturnal lubricating ointment	3 (15%)	--	--	--
Topical cyclosporine	8 (40%)	--	--	--
Topical lifitegrast	4 (20%)	--	--	--
Systemic immunosuppressant	1 (5%)	--	--	--
Punctal plugging	1 (5%)	--	--	--

Abbreviation: DEQ5, Dry Eye Questionnaire 5, OSDI, ocular surface disease index, SD, standard deviation (range).

^*p < 0.05,

^**p < 0.01,

^***p < 0.001;

% = percent; n = number

^anumber of subjects for probiotics data was n=19 for cases, n=18 for controls, and n=37 for all individuals

^bmore abnormal value between the two eyes for each subject was included in the analysis

^ctherapies were used by cases at the time of the study

As expected, based on our inclusion criteria, individuals in the case group had DE symptoms and signs, with 85% reporting severe DE symptoms (DEQ5≥12) and 65% having aqueous tear deficiency (Schirmer ≤5 mm in either eye). The mean ± standard deviation (SD) for DEQ5, OSDI, and corneal staining scores in the case group were 15.2±3.4, 55.1±22.8, and 4.4±4.1, respectively. All measurements for DE symptoms and signs were statistically different between the case and control groups (p<0.001) (Table 2).

Gut microbiome profiles in cases compared to controls

Firmicutes and Bacteroidetes were the dominant phyla in the gut microbiomes of all individuals. The predominant genus in all subjects was Bacteroides, with a mean normalized abundance of approximately 25% in both case and control groups (Figure 1A). All other genera had a mean normalized abundance of 10% or below. After Bacteroides, the most abundant genera in both groups were Romboutsia, Faecalibacterium, Eubacterium, and Blautia. Overall, there was no difference in α-diversity, meaning the richness and evenness of species within each sample was similar (Shannon’s index, p=0.97; inverse Simpson’s, p=0.65; Figure 1B). The gut microbiome profiles in cases compared to controls also demonstrated no difference in overall bacterial community structure (Euclidean PERMANOVA, p=0.62).

Figure 1: No difference in overall community structure and bacterial diversity between cases and control groups.

(A) Gut microbial composition at the genus level between the case and control groups (B) Density plots for α-diversity indices at the species level using inverse Simpson’s index (p=0.707) and Shannon’s index (p=0.487) for the case (orange) and control (gray) groups. All p-values are permutated with 10,000 simulations.

Using DESeq2, we identified 32 species and five genera whose abundance differed between the case and control groups (with a FDR<0.01 and log₂FC>1.5) (Figure 2). Among the five genera, Sneathia, Pyramidobacter, Parascardovia, and Veillonella had significantly higher abundance in the case group (4/5, 80.0%), while Octadecabacter had significantly lower abundance. Faecalibacterium was not different between groups (p=0.95). Among the 32 species, 27 had significantly higher abundance in the case group compared to the control group (27/32, 84%), 10 of which belonged to the genus Lactobacillus (10/32, 31%), and 4 of which belonged to the genus Bifidobacterium (4/32, 13%). The species with the highest expansion in the case group compared to the control group included Lactobacillus salivarius, Lactobacillus gasseri, Sneathia amnii, Lactobacillus reuteri, and Bifidobacterium animalis. The species that were most significantly depleted in the case group compared to the control group included Fusobacterium varium, Desulfovibrio piger, Candidatus melainabacteria bacterium, and Prevotella stercorea.

Figure 2: Differentially abundant bacterial species between case and control groups.

Circles on the left show significantly more abundant species in the case group, and on the right show significantly more abundant species in the control group (FDR<0.01 and log₂FC>1.5). BaseMean compares the relative abundance of each species, and was rounded to 10, 100, or 1000 (size of the circle).

Correlations between select genera and species and DE and SS indices

Correlational analyses were conducted between abundances of the genus Veillonella and the species Lactobacillus salivarius, Bifidobacterium animalis, Prevotella stercorea and Fusobacterium varium and clinical metrics as these microbiota were differentially abundant between groups and because Veillonella, Prevotella stercorea, and Fusobacterium varium are opportunistic pathogens previously associated with autoimmune disease,^26–29 and Bifidobacterium animalis and Lactobacillus salivarius are important beneficial species.³⁰ Correlational analyses were performed with DE symptoms and signs in the entire population and with early SS markers (levels of IgM, IgA, and IgG antibodies against SP1, PSP, and CA6) in the 19 individuals with early marker positivity. Spearman correlations between DEQ-5, OSDI, TBUT, Schirmer, corneal staining and abundances are shown in Tables 3A and 3B for the case and control groups, respectively. Spearman correlations between concentrations of early SS markers and abundances are shown in Table 3C.

Table 3A:

Spearman correlations between abundance of selected gut microbiota and dry eye indices among cases

	DEQ-5	OSDI	TBUT	Schirmer	Corneal staining
Genus
Veillonella
rho=	0.19	0.28	−0.35	−0.12	0.34
p=	0.43	0.24	0.13	0.62	0.15
Species
Lactobacillus salivarius
rho=	−0.27	0.22	−0.12	0.08	−0.03
p=	0.91	0.34	0.61	0.73	0.89
Bifidobacterium animalis
rho=	0.18	0.01	−0.28	0.25	−0.29
p=	0.94	0.97	0.23	0.30	0.22
Prevotella stercorea
rho=	0.24	−0.02	−0.27	−0.08	0.23
p=	0.30	0.92	0.26	0.75	0.34
Fusobacterium varium
rho=	0.41	0.00	0.33	0.72	−0.27
p=	0.07	0.99	0.15	< 0.0001 ***	0.24

Abbreviation: DEQ5, Dry Eye Questionnaire 5, OSDI, ocular surface disease index, TBUT, tear break up time, Schirmer, Schirmer score, Stain, corneal staining

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Table 3B:

Spearman correlations between abundance of selected gut microbiota and dry eye indices among controls

	DEQ-5	OSDI	TBUT	Schirmer	Corneal Staining
Genus
Veillonella
rho=	0.39	0.20	−0.08	0.14	−0.25
p=	0.09	0.40	0.74	0.56	0.28
Species
Lactobacillus salivarius
rho=	−0.20	0.37	0.36	0.33	−0.22
p=	0.41	0.10	0.13	0.16	0.35
Bifidobacterium animalis
rho=	−0.05	0.22	0.38	0.31	−0.43
p=	0.85	0.36	0.10	0.19	0.06
Prevotella stercorea
rho=	−0.27	0.23	0.20	0.75	−0.39
p=	0.25	0.32	0.41	< 0.0001 ***	0.09
Fusobacterium varium
rho=	−0.26	−0.05	−0.15	−0.06	0.28
p=	0.26	0.83	0.51	0.82	0.23

Abbreviation: DEQ5, Dry Eye Questionnaire 5, OSDI, ocular surface disease index, TBUT, tear break up time, Schirmer, Schirmer score, Stain, corneal staining

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Table 3C:

Spearman correlations between abundance of selected gut microbiota and early Sjogren’s Syndrome markers among cases

	SP1 IgG	SP1 IgA	SP1 IgM	PSP IgG	PSP IgA	PSP IgM	CA6 IgG	CA6 IgA	CA6 IgM
Genus
Veillonella
rho=	−0.34	−0.20	−0.14	0.07	0.10	0.20	−0.12	0.11	0.21
p=	0.15	0.41	1.0	0.79	0.68	0.41	0.64	0.65	0.38
Species
Lactobacillus salivarius
rho=	−0.39	−0.19	−0.46	0.10	−0.13	0.07	−0.25	0.16	−0.07
p=	0.10	0.44	0.048*	0.69	0.59	0.79	0.30	0.53	0.76
Bifidobacterium animalis
rho=	−0.16	−0.03	0.09	0.04	0.33	0.34	−0.03	0.33	0.30
p=	0.52	0.91	0.71	0.89	0.17	0.15	0.91	0.17	0.22
Prevotella stercorea
rho=	0.32	−0.11	0.05	−0.04	0.14	0.03	−0.12	0.19	0.32
p=	0.19	0.65	0.84	0.88	0.56	0.89	0.66	0.45	0.19
Fusobacterium varium
rho=	0.26	−0.29	−0.52	0.26	−0.05	0.27	−0.44	0.14	−0.20
p=	0.29	0.24	0.02*	0.28	0.85	0.27	0.06	0.58	0.41

Abbreviation: SP1, salivary protein 1, PSP, parotid secretary protein, CA6, carbonic anhydrase 6, Ig, immunoglobulin

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

In the case group, a statistically significant positive correlation was noted between Fusobacterium varium and Schirmer scores (rho=0.72, p<0.0001), and an insignificant positive association between its abundance and DEQ-5 (rho=0.41, p=0.07). In the control group, a statistically significant positive correlation was noted between Prevotella stercorea abundance and Schirmer scores (rho=0.75, p<0.0001). With respect to early markers, negative correlations were noted between SP1 IgM and both Lactobacillus salivarius (rho=−0.46, p=0.048) and Fusobacterium varium (rho=−0.52, p=0.02).

DISCUSSION

In our study, cases with DE and primarily early SS markers had gut microbiome alterations compared to age- and sex-matched controls without DE. A total of 27 species were relatively more abundant in the case group, many of which were beneficial species from Lactobacillus and Bifidobacterium genera. A pathogen that was expanded in the case group was Veillonella, a producer of lipopolysaccharides (LPS) that has also been found to be elevated in autoimmune hepatitis and Crohn’s disease.^26,29 Five species were relatively depleted in the case group, most notably Prevotella stercorea, a bacteria that can possess sialidases which can promote inflammation,³¹ and the opportunistic pathogen Fusobacterium varium implicated in IBD.²⁸ While our study demonstrated these changes in relative abundances of species and genera, the overall bacterial community structure and diversity were not significantly different between the case and control groups.

When comparing our results to prior studies, it is important to keep in mind differences in the composition of cases (e.g., early markers vs. pSS vs. sSS). With this caveat in mind, we found that some species were differentially abundant between cases and controls similar to prior studies, but the specific species involved and whether they were expanded or depleted were not consistent.^5–11 For example, most studies examining cases with pSS demonstrated a relative depletion of Bifidobacterium and no differences in Lactobacillus spp.,^5,7,8 while we found expansions in these beneficial genera in our case group of mostly early SS markers. In addition, while most prior studies demonstrated a depletion in Faecalibacterium,^6,8,9,11 our study found similar abundances of this beneficial microbe between groups. Given these differences across studies, it is possible that gut microbiome composition may differ between immune-mediated DE sub-types (e.g., early markers vs. pSS vs. sSS). However, even when comparing data of studies involving pSS cases, a clear microbiome signature is not apparent.^5–11 For example, both decreases^6,11 and increases^5,8,9 in Prevotella have been noted in cases compared to controls. Similarly, heterogeneity has been noted with respect to Veillonella abundance.^5,8,9 Finally, phylogenetic diversity results have also been variable, with some studies reporting increased α-diversity,^9,11 others reporting decreased α-diversity,^8,10 and one study noting no difference between pSS and controls,⁵ similar to the current study.

Other factors must also be considered when comparing across studies. There is evidence that age and sex may influence microbiome composition,^32,33 yet several prior studies used publicly available stool microbiota databases to create control groups that were significantly younger and more male compared to cases.^6,9,10 A strength of our study was the use of an age- and sex-matched control group, with a similar strategy employed in prior studies as well.^7,8 We additionally excluded individuals with comorbid autoimmune diseases which may also confound results, but some prior studies did not implement such exclusion criteria.^9,11 Another factor to consider is differences in bioinformatics processes.^19,20 In particular, the specific hypervariable regions (V1-V9) selected to be targets for 16S sequencing were not uniform across studies.^5–11 To reduce such bias, we instead used RNA-seq, a technique which interrogates all genomic sequences of actively replicating organisms in a sample rather than narrow sequencing regions. Finally, it is possible that greater utilization of probiotics in our case group, although not statistically significant, may have influenced gut microbiome composition results. Furthermore, differences in dietary patterns, socioeconomic background, and lifestyle may have impacted relative abundances in all studies, as these factors have previously been shown to have influences on gut microbiome composition.^34,35

Despite the unexpected expansion of certain beneficial bacteria, all cases in our study demonstrated DE symptoms and signs. Interestingly, on correlational analyses, only pathogenic microbes correlated with some DE indices. For example, positive correlations were noted between Schirmer scores and the pathogens Fusobacterium varium in the case group (rho=0.72, p<0.0001), and Prevotella stercorea in the control group (rho=0.75, p<0.0001), indicating higher tear production in individuals with higher abundances of these pathogens. The factors driving these relationships are not clear and to our knowledge, a relationship between these pathogens and healthier tear parameters has not previously been reported in the literature. In our study, a marginally significant positive correlation was noted between Fusobacterium varium abundance and DEQ-5 scores in the case group, demonstrating a pathogen whose higher levels correlated with DE symptoms. When examining early marker levels, a negative correlation was noted between IgM antibodies to SP1 and both Fusobacterium varium and Lactobacillus salivarius in the case group. This indicates that higher marker levels were associated with lower abundances of both a pathogen (Fusobacterium) and a beneficial microbe (Lactobacillus). Overall, while various bacteria correlated with various aspects of DE in our dataset, relationships varied based on bacterial species, group type (case vs. control), DE parameter, and specific SS marker.

Despite the noted correlations, there remain knowledge gaps on how the gut microbiome composition impacts immune-mediated DE. Prior studies have identified that specific microbes can impact the balance between pro-inflammatory (Th17) and anti-inflammatory (Treg) T-cells.^36–38 For example, certain commensal Bacteroides fragilis strains express polysaccharide A which can promote Treg differentiation.³⁷ There is interest in manipulating the gut microbiome, through dietary changes, probiotics, or fecal microbial transplant (FMT), to promote expansion of beneficial organisms and decrease abundance of pathogenic ones, specific to a particular disease. In RA, reduced disease activity and reduction of pro-inflammatory cytokines was noted with diet modification and probiotic supplementation, though data is overall limited.^39–41 In pSS, mice models have found improvement in DE metrics following FMT,^42,43 and we previously demonstrated that FMT is safe in individuals with immune-mediated DE.¹¹ However, further research is needed to define which manipulations, and in which patient groups, may be therapeutic in immune-mediated DE.

Several limitations must be considered when examining our study results, including the sample size, selection of controls, sequencing techniques, and unmeasured confounders (e.g., diet, measurement of metabolic products such as short chain fatty acids (SCFA) and butyrate). In addition, stratification analyses of cases by DE severity would be impactful but was not feasible due to sample size considerations. Despite these limitations, our study adds knowledge regarding gut microbiome composition in individuals with DE and early marker positivity. While we similarly found compositional differences between cases and controls, our results vary from studies focusing on pSS, suggesting that gut microbiome compositions may differ among immune-mediated DE sub-types. However, even across pSS studies, an immune-mediated DE signature is not prominent, as it is with Prevotella copri expansion in RA.³⁸ These findings have therapeutic implications as different approaches to microbe-based treatment may need to be explored in different disease sub-types. Further studies are also needed to investigate direct and indirect mechanisms that may link gut dysbiosis to clinical features of DE.