Lymphocytic colitis can be transcriptionally divided into channelopathic and inflammatory lymphocytic colitis

Abstract

Background

The pathobiology of the non‐destructive inflammatory bowel disease (IBD) lymphocytic colitis (LC) is poorly understood. We aimed to define an LC‐specific mucosal transcriptome to gain insight into LC pathology, identify unique genomic signatures, and uncover potentially druggable disease pathways.

Methods

We performed bulk RNA‐sequencing of LC and collagenous colitis (CC) colonic mucosa from patients with active disease, and healthy controls (n = 4–10 per cohort). Differential gene expression was analyzed by gene‐set enrichment and deconvolution analyses to identify pathologically relevant pathways and cells, respectively, altered in LC. Key findings were validated using reverse transcription quantitative PCR and/or immunohistochemistry. Finally, we compared our data with a previous cohort of ulcerative colitis and Crohn's disease patients (n = 4 per group) to distinguish non‐destructive from classic IBD.

Results

LC can be subdivided into channelopathic LC, which is governed by organic acid and ion transport dysregulation, and inflammatory LC, which is driven by microbial immune responses. Inflammatory LC displays an innate and adaptive immunity that is limited compared to CC and classic IBD. Conversely, we noted a distinct induction of regulatory non‐coding RNA species in inflammatory LC samples. Moreover, compared with CC, water channel and cell adhesion molecule gene expression decreased in channelopathic LC, whereas it was accentuated in inflammatory LC and associated with reduced intestinal epithelial cell proliferation.

Conclusions

We conclude that LC can be subdivided into channelopathic LC and inflammatory LC that could be pathomechanistically distinct subtypes despite their shared clinical presentation. Inflammatory LC exhibits a dampened immune response compared to CC and classic IBDs. Our results point to regulatory micro‐RNAs as a potential disease‐specific feature that may be amenable to therapeutic intervention.

Key summary.

Summary of the established knowledge on this subject

• Lymphocytic colitis (LC) is a diarrhea‐driven inflammatory bowel disease (IBD) that groups together with collagenous colitis (CC) as microscopic colitis, and its pathology is poorly characterized.

What are the significant and/or new findings of this study?

• LC can be subcategorized into channelopathic LC and inflammatory LC according to their transcriptomic profile.

• Inflammatory LC transcriptome portrays a restricted deregulated inflammation compared with CC and classic IBDs Crohn's disease and ulcerative colitis.

• Both channelopathic LC and inflammatory LC display a unique water channel and cell adhesion molecule expression that might disturb intestinal epithelial homeostasis.

• MicroRNA upregulation in inflammatory LC points towards regulatory mechanisms that could be exploited to develop new disease‐specific biomarkers and treatments.

INTRODUCTION

Lymphocytic colitis (LC) is a debilitating colonic inflammatory bowel disease (IBD) exhibiting persistent non‐bloody watery diarrhea despite a macroscopically normal mucosa. LC histopathologic hallmarks are intraepithelial lymphocytosis (>20 intraepithelial lymphocytes/100 enterocytes), and infiltration of lymphocytes and plasma cells with few eosinophils and neutrophils in the lamina propria. ¹ , ² , ³ The superficial epithelium can display focal damage including flattening, mucin depletion and vacuolization with occasional Paneth cell hyperplasia. ² LC estimated incidence is 4.0–6.1 cases/100,000 persons in Europe and the USA, with potentially increasing rates. ² , ⁴ LC is more frequent in older women but can affect younger individuals, including children. ² , ⁴ , ⁵ Based on guidelines, the first‐line treatment is the oral corticosteroid budesonide, which is effective in 88% of patients. ¹ , ² , ³ Still, disease often relapses after budesonide discontinuation. ⁶

Dense genotyping failed to identify genes associated with LC despite the mucosal immune cell infiltration and co‐occurrence with celiac disease. ⁷ , ⁸ Its counterpart collagenous colitis (CC)—collectively referred as microscopic colitis—, has been associated with several human leukocyte antigen (HLA) genetic variants, which was confirmed by enrichment of adaptive immune transcriptional signatures. ⁸ , ⁹ Besides clinical overlaps with CC, LC shares features with IBDs ulcerative colitis (UC) and Crohn's disease (CD) such as Th1/Th17‐driven immune responses, tight‐junction claudin dysregulation, and water channel aquaporin (AQP) downregulation. ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ Interestingly, the incidence of CD and UC increases in patients previously diagnosed with LC and CC. ⁶ , ¹⁹ , ²⁰ Indeed, LC and CC have been suggested to be attenuated IBD forms where an overt mucosal inflammation remains restrained. ⁹ , ¹³ , ²¹ Still, microscopic colitis can progress to classic IBD upon immune system hyperactivation as shown in LC patients overexpressing higher levels of interferon (IFN)‐γ, and Th2 transcription factor GATA3; or CC patients overexpressing tumor necrosis factor (TNF)‐α and Th1 transcription factor T‐BET. ²¹ Furthermore, LC displays mild and focal crypt architectural distortions (branched, dilated and disorganized crypts) and scattered non‐necrotizing epithelioid granulomas and abscesses, similar to initial presentations of CD. ²⁰ , ²² , ²³ Still, the extent of mucosal dysfunction and immune responses contributing to LC pathophysiology are unclear.

Water and sodium malabsorption have been suggested as pathomechanisms leading to diarrhea in CC. ²⁴ , ²⁵ , ²⁶ However, epithelial surface injury is more prominent in CC than in LC. ⁶ Thus, the mechanisms leading to diarrhea in LC remain obscure and potential biomarkers or druggable targets are yet unknown. To characterize LC pathophysiology, we investigated the transcriptome of LC colonic mucosa and compared it with that of CC and classic IBDs. Our patient cohort enabled us to subclassify LC into channelopathic and inflammatory LC, identify LC diarrhea‐driving pathomechanisms, and propose targets for the development of new treatments for LC patients.

MATERIALS AND METHODS

Study population

Descending colonic biopsies were collected from healthy volunteers and patients with LC and CC who were diagnosed based on clinical history and histopathology and sampled at Linköping University Hospital, Sweden (Table 1). Adjacent biopsies were stored in AllProtect (Qiagen) for RNA extraction or in PBS for paraformaldehyde fixation and paraffin embedding (FFPE).

TABLE 1.

Clinical and demographic characteristics of the patient cohort and controls.

Variable	Hc	LC (all)	LC 1	LC 2	CC
Total number of subjects	10	8	4	4	4
Gender, % females	60.00	62.50	50.00	75.00	75.00
Average age (range)	57 (55–63)	53 (20–80)	43 (20–67)	63 (44–80)	70 (49–86)
Average stools/day (range)	‐	7.13 (2–15)	5.5 (2–7)	8.75 (4–15)	7.50 (6–10)
Average watery stools/day (range)	‐	7 (2–15)	5.5 (2–7)	8.5 (3–15)	7.50 (6–10)
Average collagenous band, μm (range)	‐	‐	‐	‐	34.00 (12–52)

Note: LC was diagnosed by intraepithelial lymphocytosis of ≥20 lymphocytes/100 surface epithelial cells, increased inflammatory infiltrate in lamina propria and normal collagenous band (10 μm), and none of the patients presented any rare endoscopic abnormalities and/or visible lesions at the time the biopsies were taken. CC was diagnosed by subepithelial collagen band of >10 μm thickness. All CC and LC were active at sampling time (>3 bowel movements/day or ≥1 watery bowel movement/day during 1 week), were not taking any medication and did not suffer from enteric infections. Healthy volunteers displaying normal macro‐ and microscopic findings with normal bowel movements, no medication, and not suffering from gastrointestinal or autoimmune diseases were enrolled from the colon cancer screening program for healthy control sampling.

Abbreviations: CC, collagenous colitis; Hc, healthy controls; LC, lymphocytic colitis.

Genome‐wide mRNA sequencing (RNA‐seq) and analyses

Biopsies in AllProtect were homogenized in RLT buffer with 1% 2‐mercaptoethanol in TissueLyser II, and RNA was isolated with the RNeasy Mini Kit (all from Qiagen). Sequencing libraries were constructed with TruSeq Stranded total RNA RiboZero GOLD and paired‐end sequenced (2 × 150bp reads, 300 cycles, 1000 × 10⁶ base reads) on NovaSeq6000 (all from Illumina). FASTQ files were generated using bcl2fastq v2.20, and analyzed using R Bioconductor v3.5.1 (R Core Team 2018), including SARTools v1.6.6 and DESeq2 v4.1.3 packages. ²⁷ , ²⁸ , ²⁹ Reads were aligned to Ensembl GRCh38 release 92. RNA‐seq data are available at Gene Expression Omnibus: GSE245764 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE245764).

Differential gene expression was calculated using DESeq2 linear models (significance by Benjamini‐Hochberg false discovery rate, FDR, adjusted p‐values <0.05). Raw count data were normalized using DESeq2. Principal component analysis was computed after making the data homocedastic. Differentially expressed genes (DEG) were filtered to >|1.5| log₂ fold‐change for gene ontology (GO) biological process and WikiPathways enrichment analyses using clusterProfiler v4.2.2. Intersection plots were created using VennDetail v1.10.0. Potential microRNA‐gene targets were computed using miRTarBase and g:Profiler, ³⁰ , ³¹ and deconvolution using CIBERSORTx (https://cibersortx.stanford.edu/). ³²

Reverse transcription quantitative polymerase chain reaction (RT‐qPCR)

RNA was quantified using NanoDrop ND‐2000 and reverse‐transcribed with High‐Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific). Reverse transcription quantitative polymerase chain reaction (RT‐qPCR) was performed using iTaq Universal SYBR Green Supermix (BioRad, Hercules) with the in‐house designed primers (see ²⁴ and Table S1) in a CFX96 Touch Real‐Time PCR system (BioRad). Gene expression was obtained after relative ‐ΔCt quantification with reference gene HPRT1. Each sample was analyzed in duplicate.

Immunohistochemistry

FFPE 4  μm sections were deparaffined in Xylol substitute (Carl Roth GmbH) and antigens retrieved with citric acid (pH6.0, 20 min). Endogenous peroxidase activity was blocked with 3% H₂O₂ for 10 min before blocking with 1% bovine serum albumin (BSA) with 0.2% TritonX100 for 30 min. Anti‐claudin 2 (dilution 1:250, 36–4800, Invitrogen), anti‐claudin 4 (dilution 1:750, 32–5600, Invitrogen), anti‐ZO1 (dilution 1:500, 339,100, Invitrogen), anti‐Ki67 (dilution 1:500, 556,003, BD Biosciences), and biotinylated secondary antibodies (ab6788, Abcam; BA ‐2000, Vector Laboratories) were used to stain samples in 1% BSA. VECTASTAIN Elite ABC reagent (Vector Laboratories) and hematoxylin served to develop and counterstain samples before dehydration and mounting. Images were acquired on a Zeiss AxioImager.Z1 microscope (Zeiss). Semi‐quantitation of claudin 2, 4 and Zonula Occludens (ZO)‐1 in epithelial or stromal areas was blindly estimated by two analysts (0 = no staining, 4 = very strong staining).

Statistical analyses

Immunohistochemistry (IHC) semi‐quantifications and Ki67 staining percentage in colonic crypts were analyzed with non‐parametric Kruskal‐Wallis test. RT‐qPCR data were analyzed with non‐parametric Kruskal‐Wallis and Mann–Whitney tests and FDR adjusted. All analyses were performed using R.

RESULTS

Transcriptomic profiling distinguishes two lymphocytic colitis subgroups

Lymphocytic colitis is characterized by chronic non‐bloody diarrhea and macroscopically intact mucosa despite increased lymphocyte infiltration in epithelia and lamina propria (Figure 1a). Since LC pathophysiology is underexplored, we investigated its mucosal transcriptome. Principal component analysis split LC into a subgroup that was transcriptionally similar to healthy controls, Hc (LC1), and a subgroup transcriptionally different from Hc and its closely related disorder CC (LC2) (Figure 1b). LC1 mucosa displayed 214 DEGs compared to Hc, 22.43% of which displayed an absolute fold‐change greater than 1.5 (48 DEGs with FC>|1.5|) and were related to temperature homeostasis, lipid transport/localization, and mitosis. On the other hand, LC2 displayed 7475 DEGs compared to Hc, and the 1582 genes with FC>|1.5| (21.16%) were related to mitosis, ion transport, and metabolism (Figure 1c–h). Still, organic acid and ion transport were more upregulated in LC1 than in LC2 subgroup, and the microbial immune response was more increased in LC2 than in LC1 subgroup (458 out of 2590 DEGs had FC>|1.5|, Figure 1c,g,h). Therefore, LC patients could be subclassified as channelopathic LC (LC1) and proinflammatory LC (LC2).

FIGURE 1.

Lymphocytic colitis mucosal gene expression subclassifies the disease into channelopathic and inflammatory LC. (a) Representative histology of hematoxylin and eosin–stained paraffin‐embedded sections of the human colonic mucosa in a healthy control (Hc) subject, a LC patient, and a CC patient. (b) Principal component analysis of the RNA‐seq expression profiles of Hc (green), LC (orange), and CC (blue), where a subset of LC patients transcriptionally resemble Hc (LC1), and a subset separates from other patient groups (LC2). (c) Venn diagram displaying the DEGs between the two LC patient subgroups and Hc samples. (d) Dot plot of enriched biological process gene ontology (GO:BP) in LC1 compared to Hc. (e) Heatmap showing normalized log2‐transformed transcript counts (log TC) of the leading 50 (Top50 DEGs) between LC1 and Hc samples. (f‐g) Dot plots of enriched GO:BP terms in LC2 compared to Hc (f) and LC1 compared to LC2 (g). (h) Heatmap showing log TC of the top50 DEGs between LC2 and Hc samples combined with the top50 DEGs between LC1 and LC2. All results are based on n = 4–10 samples per group. Hc subjects are shown in green, LC1 in orange, and LC2 in brown. The dot size in the dot plots indicates the gene ratio as the ratio of significant DEGs to annotated genes in each gene set term (total number of sets in brackets), while the colors display adjusted p‐values. Heatmap columns are split according to hierarchical clustering. CC, collagenous colitis; DEGs, differentially expressed genes; LC, lymphocytic colitis.

Lymphocytic colitis groups display a moderate immune response compared to collagenous colitis

Compared to its clinically related disorder CC, LC differed in the expression of 685 (LC1) and 1524 (LC2) genes. Of those, 80 DEGs (37 shared for LC1 and LC2) belonged to immunoglobulin (IG) genes (IGH, IGHV, IGK, IGKV, IGL or IGLV). All IG genes displayed a lower gene expression in LC groups compared to CC, indicating a less prominent humoral immune response in both LC subgroups (Figure 2a). Enrichment analyses also pointed to a less prominent cellular immune response in LC (more evident in LC1) as pathways related to chemokine and cytokine activity displayed a diminished gene expression compared to CC (Figure 2b,c). To estimate immune cell abundances, gene expression profiles were imputed and deconvoluted. While no immune cells seem to be altered in channelopathic LC, proinflammatory M1 macrophages could be increased in LC2 and CC, while naïve CD4⁺ T‐cells and memory B‐cells could be underrepresented in LC2 (Figure 2e,f).

FIGURE 2.

The immune response in lymphocytic colitis is limited compared to CC. (a) Heatmap displaying normalized log2‐transformed transcript counts (log TC) of the immunoglobulin (IG)‐related differentially expressed genes (DEGs). (b) Dot plot of enriched biological process gene ontology (GO:BP) in lymphocytic colitis (LC) groups compared to (CC). (c) Heatmap displaying log TC of cytokine and chemokine DEGs between controls (Hc), LC1, LC2, and CC samples. (d) Heatmap displaying log TC of the 50 leading DEGs between Hc, LC1, LC2, and CC samples, excluding IG‐related DEGs. (e) RNA‐seq data deconvolution computed for different immune cells displaying the estimated relative percentage in each sample. (f) Plots of estimated immune cells according to RNA‐seq data deconvolution that were statistically significant in any of the LC groups. Median with interquartile range is shown. (g) Intersection plot of DEGs from comparisons of LC1, LC2 and CC samples against Hc samples. (h) Dot plot of enriched GO:BP that are unique for LC1 group compared to Hc and CC. (i) Dot plot of enriched GO:BP in LC2 group compared to Hc and CC. (j) Correlation matrix of leading DEGs related to inflammation and transport of molecules. All results are based on n = 4–10 samples per group. Hc samples are shown in green, LC1 in orange, LC2 in brown, and CC in blue. The dot size in the dot plots indicates the gene ratio as the ratio of significant DEGs to annotated genes in each gene set term (total number of sets in brackets), while the colors display adjusted p‐values. Heatmap columns are split according to hierarchical clustering. CC, collagenous colitis. Statistically significant differences are shown as *p < 0.05, **p < 0.01, and ***p < 0.001.

In the analyses of LC1, LC2 or CC compared to Hc, we found that LC1 uniquely regulated one of three natriuretic peptide receptors NPR3 responsible for secretion volumes; and with LC2 and CC, LC1 shared dysregulation of ion transport (Figure 2g,h). Regarding proinflammatory LC (LC2) compared with CC, LC2 overexpressed mitotic genes and underexpressed hormone metabolism genes. In addition, the immune response was found again to be lower in LC2 compared to CC (Figure 2i).

Due to the inflammation grade ranging from absent (LC1), towards minor (LC2) and moderate (CC), we correlated leading inflammatory DEGs to channel and transporter DEGs. Here, we found that the inflammatory DEGs IFNG, NOS2, PLA2G2A, HLA‐DRB1, CXCL1‐3, LCN2, and REG1 correlated with the gene expression of several solute transporters (Figure 2j).

Taken together, our results indicate that LC1 results from ion channel dysregulation, while LC2 has an immune nature, which is less developed than the inflammation in CC. Despite the lymphocyte recruitment into LC mucosa, immune cells might remain relatively inactive yet capable of disrupting epithelial solute transport.

Lymphocytic colitis gradually downregulates water channel aquaporins and tight‐junction claudins differentially from collagenous colitis

Both channelopathic and inflammatory LC featured enrichment of organic acid and ion transport DEGs (Figures 1 and 2), which could contribute to diarrhea. Since AQP downregulation seems key for both non‐destructive and classic IBD, ¹⁷ , ³³ we next evaluated AQP gene expression. RNA‐seq showed a significant downregulation of AQP7 and AQP8 in LC1 and LC2, and of AQP10 in LC1 (Figure 3a); however, read counts were low for other AQPs. Thus, we analyzed AQP expression using RT‐qPCR. RT‐qPCR verified APQ8 downregulation in both LC subtypes and identified a notable downregulation of many other AQP genes, including AQP0‐4, AQP6‐7, AQP10, and AQP12, in LC2. Even though LC1 followed the same downregulation pattern, gene expression did not reach statistical significance (Figure 3b). To further explore alternative mechanisms of paracellular permeability, we analyzed tight‐junction protein (TPJ) 1 and the most relevant claudins (CLDN) for intestinal biology. RNA‐seq counts pointed to upregulation of CLDN1 and CLDN12, and downregulation of CLDN8 and TJP1 in LC2 (Figure 3a). RT‐qPCR validated CLDN8 and TJP1 downregulation and identified significant downregulation of several claudin genes, including CLDN3, “tight” CLDN4, CLDN5, CLDN7, and CLDN15 in LC2. Again, LC1 displayed a similar dysregulation pattern as LC2 and reached statistical significance for CLDN3, CLDN4, CLDN8 and TJP1 downregulation (Figure 3b). Semi‐quantification of CLDN2 and CLDN4 staining on tissue samples demonstrated a normal protein expression in intestinal epithelia and stroma, yet CLDN2 distribution in LC was more extended in the stroma than in Hc and CC samples (Figure 3c,d). Interestingly, TJP1 product ZO‐1 was downregulated in the stroma of both LC subtypes, and the staining was more diffuse than that of Hc, where ZO‐1 was mostly localized towards the epithelial barrier (Figure 3e).

FIGURE 3.

Epithelial cell homeostasis is gradually disrupted in lymphocytic colitis. (a‐b) Normalized log2‐transformed transcript counts (log TC) (a) or log₂ fold changes (–ΔCt, log₂ values) in gene expression as analyzed by RT‐qPCR (b) of water channel aquaporins (AQP), claudins (CLDN) and tight‐junction protein 1 (TJP1) genes (n = 2–10 samples per group). HPRT1 was used as a housekeeping control gene for RT‐qPCR analyses. All primers detected all coding transcript variants of the indicated gene. (c‐e) Representative IHC images of longitudinally sectioned epithelial glands stained for claudin 2 (c), claudin 4 (d), and ZO‐1 (e) (left). Semi‐quantitative analysis of claudin 2, 4, or ZO‐1 is shown on the right panels (n = 3–9 samples per group). (f) Log TC of the MKI67 proliferation marker. (g) Representative IHC images of longitudinally sectioned epithelial glands stained for Ki67 proliferation marker (brown) in paraffin‐embedded sections (left). Analysis of Ki67 relative staining to total crypt length is shown on the right (n = 3–9 well‐oriented crypts per group, median of 8 crypts/patient). Hc samples are shown in green, LC1 in orange, LC2 in brown, and CC in blue. All plots represent median with interquartile range. Statistically significant differences are shown as *p < 0.05, **p < 0.01, and ***p < 0.001.

Since we previously described intestinal epithelial hyperproliferation in colonic crypts from active CC mucosa ⁹ and the proliferation marker MKI67 was upregulated in LC compared with Hc (Figure 3f), we quantified the extent of Ki67 protein staining. This analysis showed that epithelial cells in LC1 proliferated at a normal rate, while LC2 epithelia hyperproliferated at a level similar to CC (Figure 3g). The correlation of gene expression or protein levels did not match clinical data (not shown). Collectively, these results point to an epithelial cell hyperproliferation, paracellular and transcellular water transport dysfunction, and tight‐junction altered distribution within submucosa in proinflammatory LC2 subtype, while changes are more modest in the LC1 subtype.

Posttranscriptional regulatory mechanisms may dampen the immune response in lymphocytic colitis

To find an explanation for the macroscopically intact mucosa in inflammatory LC, we compared the LC2 transcriptome to active classic IBD using a patient cohort we previously published. ³⁴ LC2 differed from colonic CD in 3703 DEGs (82.80% with FC>|1.5|), and from UC in 4079 DEGs (83.62% with FC>|1.5|). As expected, CD and UC displayed a stronger immune response than LC2 and exhibited reduced control to angiogenesis and endothelial cell proliferation pathways (Figure 4a). Interestingly, regulatory transcriptional and translational mechanisms were enriched in LC2, which indicated a role for miRNAs (Figure 4a,b). Enrichment analyses of miRNA primary transcript DEGs involved the DNA damage response and extracellular matrix and transmembrane receptors that mediate cell‐to‐cell communication (Figure 4c).

FIGURE 4.

Lymphocytic colitis immune response might be restrained by posttranscriptional regulatory mechanisms. (a) Dot plot of enriched biological process gene ontology (GO:BP) in inflammatory lymphocytic colitis (LC2) compared to Crohn's disease (CD) and ulcerative colitis (UC). (b) Heatmap displaying normalized log2‐transformed transcript counts (log TC) of leading differentially expressed genes (DEG, log₂ fold‐change > |5|) of primary precursors of micro‐RNAs (MIR). (c) Dot plot of leading enriched primary precursor MIR DEGs according to WikiPathway database. (d) Intersection plot of DEGs from comparisons of disease (LC2, CD, or UC) samples against healthy control (Hc) samples. (e) Dot plot of enriched GO:BP between disease (LC2, CD, or UC) compared to Hc, including only unique DEGs for each of the comparisons. (f) Dot plot of enriched GO:BP for shared DEGs between LC2, CD, or UC comparisons to Hc. (g) Log TC in gene expression of leading genes from the dot plot in F between all studied groups (Hc, LC1, LC2, CD, and UC samples). All results are based on n = 4–10 samples per group. Hc subjects are shown in green, LC1 in orange, LC2 in brown, CD in light purple, and UC in dark purple. Heatmap columns are split according to hierarchical clustering.

To further explore the similarities and differences with classic IBDs, we compared each disease with Hc. Among the unique DEGs for each of them, hormone and neurotransmitter responses were decreased in CD, while phagocytosis and immune response were more prominent in UC, and mitosis was upregulated in LC2 compared with Hc (Figure 4d,e). When intersected comparisons were analyzed, LC2 shared with UC a lower response to metal ions, and with CD a decreased amino acid metabolism and increased tolerance response and negative regulation of humoral responses. The three disorders shared an upregulation of metabolic processes when compared with Hc (Figure 4d,f). When checking the expression of the leading DEGs including all patient groups, we observe a progressive decrease in glucuronidation (UGT1A5/A7/A9), ATP‐binding (ABCB11), vasoconstriction (EDN1) and organic cation and drug metabolism (SLC22A4, CYP2C9); and a progressive increase in inflammatory markers (S100A9/calprotectin, CXCL1/2/3, SOCS1), epithelial protector SLPI and antimicrobial PI3 during the surge of inflammation (being LC1 the less inflammatory disorder followed by LC2, then CC, CD, and finally UC). Also, antimicrobial DEFA6 outstanded for its increased expression in all disorders but LC1; and MIR186 for its normal expression in non‐destructive IBDs but downregulation in classic IBDs (Figure 4g).

In summary, the LC2 immune response could be restrained by miRNA‐driven regulatory events affecting the DNA damage response and tissue remodeling, which could be exploited for the development of new treatments for IBD.

DISCUSSION

Lymphocytic colitis etiology and pathophysiology are poorly understood. Here, we identify two subgroups of LC patients according to their transcriptome, which we classified as channelopathic LC due to altered ion transport (LC1 subtype), and inflammatory LC due to additional antimicrobial immune response (LC2 subtype). In addition, LC2 alteration point towards moderate immune responses when compared to its counterpart CC.

Although microscopic colitis, including LC, has been associated with autoimmune disorders, ³⁵ dense genotyping studies have discarded any link to HLA genes with LC. ⁸ A report shows increased HLA‐A1 and decreased HLA‐A3 in LC. ³⁶ In our data, we found the upregulation of HLA genes only in inflammatory LC, with activation of humoral IG immunity. This could match the increased prevalence of anti‐nuclear, anti‐Saccharomyces cerevisiae, and thyroid peroxidase antibodies in LC. ³⁷ Interestingly, Epstein‐Barr virus is nearly always detectable in microscopic colitis biopsies, so it could explain the increased microbial immune responses we found in LC—and previously in CC ⁹ —, which link both diseases with autoimmunity due to its ability to mimic host molecules. ³⁸ , ³⁹ However, whether this is valid for channelopathic LC remains unknown.

We report an active cellular immune response in inflammatory LC, but to a lesser extent than CC, CD and UC. Günaltay et al. noted an increased expression of granulocyte, Th1 and CD8⁺ T cell‐associated chemokines in LC mucosa ¹⁵ that we partially corroborate. However, we found that the expression levels of IG‐related genes and key chemoattractants such as CXCL1, CXCL10 or CXCL11 was lower in inflammatory LC compared with CC (and absent in channelopathic LC). Going forward, deconvolution analysis pointed to a relative increase in proinflammatory M1 macrophages and a decrease in CD4⁺ naïve T‐cells and memory B‐cells in inflammatory LC, which was comparable to CC. In contrast, Carrasco et al. disclosed more robust results using flow cytometry, demonstrating an increase in CD3⁺ T‐cells, CD3⁺CD8⁺ T‐cells, CD3⁺CD4⁺TCRγδ⁺ T‐cells, double negative CD3⁺CD4⁻CD8^‐ T‐cells, and a decrease in CD3⁺CD4⁺IFN⁺ Th1 and CD3⁺CD4⁺IL‐17A⁺ Th17 cells in LC. ¹³ Despite of lower Th1/Th17 cell numbers, IFNG, TNFA, IL17A, IL21, IL23 gene expression levels were increased, which were also reported by Kumawat et al. ⁴⁰ , ⁴¹ We only corroborated IFNG increase in inflammatory LC (FC = 5.96 vs. Hc) probably due to individual variability and cohort differences. More interesting, none of these reports found clear associations between cytokine gene expression and protein levels. Thus, LC intensively recruited immune cells could have a limited activity, remaining as recruited ready‐to‐react sentinels. Still, immune chemoattraction due to microbial infiltration seems plausible only for inflammatory LC. Oppositely, channelopathic LC could be explained by exposure to non‐steroidal anti‐inflammatory drugs, proton pump and selective serotonin reuptake inhibitors or aspirin, ⁶ but additional research is needed to confirm these two disease subtypes.

The extensive water channel AQP downregulation in LC—especially in inflammatory LC—, contrasts the unique downregulation of AQP8 in CC. ²⁴ Since tight‐junction claudin gene expression is downregulated, a combined paracellular and transcellular permeability is possible in LC. Furthermore, the increased calcium channel CLCN1 and sodium‐hydrogen exchanger SLC9A1/NHE1 expression, and limited epithelial sodium channel (ENaC) γ expression ⁴² could contribute to chloride and sodium disbalance in LC, exacerbating water malabsorption in the colon, hence, leading to diarrhea. In contrast, downregulation of sodium‐hydrogen exchangers SLC9A2/3 (NHE2/3), downregulated in adenoma (SLC26A3/DRA), and putative anion transporter 1 (SLC36A1/PAT1) seem to be responsible for misbalanced ion exchange in CC. ²⁴ , ⁴⁰ , ⁴¹ Altogether, these results point towards increased water and ion permeability in LC that differs from CC diarrheal mechanisms. However, whether the ion channel and tight‐junction leakage initiate mucosal exposure to luminal insults in LC or are a consequence of pre‐activated immune responses around hyperproliferative epithelia and underlying immune cells remain unknown.

When comparing inflammatory LC to classic IBD, we found miRNAs as potential protectors against overt immune responses contributing to DNA damage repair and tissue remodeling. However, LC increased expression of primary miRNA transcripts could also be confounded by the effects of different sequencing chemistry versions between datasets. Importantly, we did not assess mature miRNA profiles. Since the only miRNA study with sufficient samples found that the CD phenotype‐linked stress‐responsive miR‐31 was upregulated in IBD and CC, but not in LC, ⁴³ , ⁴⁴ further studies and validation are required to describe the role of miRNAs in LC.

In summary, our results transcriptionally subclassify LC into a subtype solely dysregulating ion channel expression (channelopathic LC), and a subtype additionally inducing microbial immune responses (proinflammatory LC). In comparison to non‐destructive IBD CC, and classic IBDs CD and UC, the inflammatory LC immune response is mild and could be restricted by miRNA modulation. Diarrheal mechanisms are distinct as water channel and tight‐junction claudin downregulation is generalized in both LC subtypes, compared to the exclusive AQP8 downregulation in CC. This, together with an increased proliferative rate of epithelia, demonstrates gradual epithelial cell dysfunction in LC, with a mild effect in channelopathic LC and progressive, increased alterations in inflammatory LC.

CONFLICT OF INTEREST STATEMENT

C.E.H., S.K. and A.M. received financial support from Ferring Pharmaceuticals (Switzerland). A.M. has received a salary for consultancies from Tillotts Pharma AG, Ferring, Vifor and Dr Falk Pharma and speaker's honoraria from Tillotts Pharma AG and Vifor. The remaining authors declare no conflicts of interest.

ETHICS APPROVAL

Informed written consent was obtained from all participants, and their data were handled according to current regulations and guidelines (EU2016/679, corrigendum 23 May 2018; and Declaration of Helsinki). Ethical approval was issued by Linköping's regional ethical committee to conduct studies in microscopic colitis, including collagenous and lymphocytic colitis (Dnr 2015/31‐31).

Supporting information

Table S1

Bhardwaj A, Münch A, Montague J, Koch S, Rosenstiel P, Escudero‐Hernández C. Lymphocytic colitis can be transcriptionally divided into channelopathic and inflammatory lymphocytic colitis. United European Gastroenterol J. 2024;12(6):737–48. 10.1002/ueg2.12531

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.