Regulation and Role of αE Integrin and Gut Homing Integrins in Migration and Retention of Intestinal Lymphocytes during Inflammatory Bowel Disease

Key Points Adhesion molecules are upregulated in inflamed intestinal mucosa in IBD patients. Baseline β7 expression does not impact αE induction or gene expression in T cells. Phospho-SMAD3 is increased in inflamed mucosa in IBD.

T he immune system relies on a complex network of mechanisms to enable immune surveillance of tissues. Migration of leukocytes into intestinal tissues involves chemokines, adhesion molecules, and integrins to mediate rolling along the endothelium, arrest following integrin activation, and binding to adhesion molecules and transmigration (1). Circulating T cells in the peripheral blood can enter the gut through an endothelial barrier in one of two ways: 1) naive T cells migrate to gut-associated lymphoid tissue or 2) memory T cells previously activated in gut-associated lymphoid tissue enter the lamina propria (LP) via systemic circulation (2).
Under normal homeostatic conditions, migration into the gut occurs through interactions between a4b7 integrin expressed on both naive and memory T cells and MAdCAM-1 on the endothelium (2,3). To a lesser extent, cells are thought to use a4b1 (very late Ag 4 ) interacting with vascular adhesion molecule-1 (VCAM-1) and fibronectin (4), or aLb2 (lymphocyte function-associated Ag) interacting with ICAM-1 to migrate to the intestine (5,6). Recent data suggest that a4b1 may be particularly important for migration to the small intestine in comparison with the colon (7). Once lymphocytes have entered the intestinal mucosa, a subset of T cells upregulates aEb7 (CD103) (8). aEb7 binds to E-cadherin and increases the retention of cells in the intestinal mucosa (9).
Aberrant innate and adaptive mucosal immune responses to microbes in the gut lumen characterize both Crohn's disease (CD) and ulcerative colitis (UC), the two types of inflammatory bowel disease (IBD) (10). During inflammation, intestinal vascular endothelium can be activated by inflammatory cytokines, including TNF-a and IL-1b, and other stimuli, such as LPS, to upregulate expression of VCAM-1 and ICAM-1 (11), which have the potential to alter normal lymphocyte migration patterns into the intestine. Early studies in IBD showed that MAdCAM-1 is upregulated in IBD samples (12), whereas results for VCAM-1 and ICAM-1 are mixed (13,14). Investigating inflammation-induced changes in lymphocyte migration is important for understanding current anti-integrin therapies for IBD that target lymphocyte migration to the intestine.
Differences between blockade of a4 integrin (natalizumab) (15), a4b7 integrin heterodimer (vedolizumab) (16,17), and b7 integrin (etrolizumab) (18,19), which blocks both a4b7 and aEb7 heterodimers, may be key to understanding differences in therapeutic efficacy. Both a4b7 and a4b1 are expressed by a large proportion of circulating T cells (20). aEb7, in contrast, is expressed by very few circulating T cells but is highly expressed by intestinal T cells both in the LP and intraepithelial space (21). TGF-b1 is a multifunctional cytokine present in the intestinal mucosa that can signal to T cells, is associated with commitment to the Th17 cell lineage, and also induces aE expression. Disease-associated risk alleles in ITGAE, the gene that encodes aE integrin, have been associated with differential levels of surface expression of aE following TGF-b1 stimulation in vitro (22). It is unclear if circulating T cells require pre-existing expression of a4b7 integrin to express aEb7 integrin in the gut, perhaps by switching the pairing of the a-subunit on b7 1 cells from a4 to aE through increased aE surface expression. This is of particular importance, as data from our group and others suggest that a subset of potentially pathogenic T cells expresses aEb7 integrin in UC (21,23,24).
The present study aims clearly to define inflammation-induced changes in lymphocyte migration that are relevant to interpreting clinical data for therapies that inhibit integrin-mediated lymphocyte migration to the intestine. Blockade of a4b7/MAdCAM-1 interaction has been shown to reduce lymphocyte trafficking. As we show in this study, induction of aEb7 integrin on a4b7 À cells is possible, suggesting that cells trafficking to the gut via inflammation-induced alternative mechanisms (e.g., via a4b1/VCAM-1 or aLb2/ICAM-1) may become aEb7 1 effector T cells. Our findings support both altered adhesion molecule patterns in the intestine in active IBD as well as differentiation of aE 1 T cells from cells that initially enter the intestinal mucosa by non-a4b7/MAdCAM-1 means.

Human subjects
Healthy volunteers and IBD patients were enrolled in separate studies at Newcastle University, and in a previously described Mayo Clinic (25) and multicenter observational IBD study (Emerging Biomarkers [EMBARK]) (26). Patient characteristics for all cohorts are shown in Supplemental Table I. In addition, peripheral blood and PBMCs were collected from healthy donors participating in the Genentech blood donor program with written, informed consent. Saliva was collected from Genentech donors and genotyped on the Global Screening Array (Illumina, San Diego, CA) and confirmed by targeted PCR genotyping.
For the Newcastle cohort, none of the IBD patients were taking any medications for their disease, including biologics, immunomodulators, steroids, or 5-ASA. Active disease in the ileum or distal colon was defined as endoscopically visible mucosal inflammation and active inflammation was confirmed via histological examination by a pathologist. Inactive disease in the ileum or distal colon was defined as absence of visible inflammation at endoscopy and absence of histological inflammation by pathologist examination.

Ethics
Written informed consent was obtained in accordance with research and ethics committee (REC) (Newcastle and North Tyneside 1 REC 10/H0906/41 and Newcastle and North Tyneside 2 REC 22/02; Mayo Clinic Institutional Review Board 10-006628) approval. These studies were performed according to the principles of the Declaration of Helsinki. The Western Institutional Review Board provided research and ethics oversight for the Genentech blood donor program and the EMBARK cohort has been previously described (26).

Microarray
RNA was isolated as previously described (23). Microarray data (Gene Expression Omnibus https://www.ncbi.nlm.nih.gov/geo; accession number GSE 179285) were preprocessed and normalized using packages from the Bioconductor project (http://bioconductor.org), supplemented with scripts written in the R programming language (http://r-project.org). Feature-extracted data were background subtracted using the normal 1 exponential method from the limma Bioconductor package (27). Log-ratios for the test and reference channel were normalized using loess normalization (28). Datasets were independently normalized across samples using quantiles normalization (29). We filtered out probes from the microarrays that did not map to known Entrez Genes, and further reduced the data by selecting a single probe per gene, selecting the gene with the highest interquartile range as the exemplar for a given gene (30). Individual genes were plotted for each patient sample in the EMBARK cohort (26).

Flow cytometry
Expression of integrin b7, b1, and aE on CD3 1 peripheral homing (CD45RA b7 low), mucosal homing (CD45RA b7 high), and naive (CD45RA 1 b7 intermediate) T cell subsets were assessed by flow cytometry. In brief, whole blood samples from Mayo Clinic UC, CD, and non-IBD patients were collected into sodium heparin vacutainer tubes and shipped overnight at ambient temperature to Genentech. To lyse the RBCs, 1Â BD Pharm Lyse buffer (BD Biosciences, San Jose, CA) was added to the whole blood sample and allowed to incubate at room temperature for 20 min. Samples were washed three times with ice cold staining buffer (PBS 1 2% FBS) followed by addition of normal mouse serum to block potential nonspecific staining. Samples were incubated on ice for 20 min prior to staining aliquots with Ab mixtures containing fluorescently conjugated anti-CD3 (SP34-2), CD45RA (HI100), b7 (FIB504), b1 (MAR4), aE (Ber-ACT8), and/or the respective isotype controls (BD Biosciences, San Jose, CA). Samples were incubated on ice for 30 min, washed twice with staining buffer, and resuspended in 1% paraformaldehyde prior to acquisition on a BD LSR (BD Biosciences, San Jose, CA). A total of 75,000100,000 lymphocyte gated events were acquired per sample. Flow cytograms were generated to establish the fraction of cells positive for each surface marker. Geometric mean fluorescent intensity (MFI) and percentage of gated cells for each T cell subset were assessed and reported accordingly.
In vitro cell culture experiments T cells were purified from PBMCs by immunomagnetic negative selection using the EasySep Human T Cell Enrichment Kit (STEMCELL Technologies) and cultured with CD3/CD28 Dynabeads (Invitrogen) in complete RPMI medium-1640 (RPMI) containing 10% FCS at a 1:1 cell/bead ratio with or without 10 ng/ml TGF-b1 (R&D Systems). In some experiments, an ALK5 inhibitor (SB-505124; Sigma-Aldrich) was included. For CFSE dilution experiments, purified T cells were labeled with 10 mM CFSE, stimulated at a 2:1 cell/bead ratio, and harvested at times indicated.

Comparison of gene expression in T cells with or without baseline b7 expression
PBMCs from seven donors were phenotyped and sorted using a BD FACS Aria Fusion cell sorter (BD Biosciences) into CD3 1 b7 À aE À and CD3 1 b7 1 aE À populations (Abs: anti-CD3 clone OKT3, anti-b7 clone Fib504, anti-aE clone Ber-ACT8, anti-aL clone HI111, anti-b1 clone TS2/ 16). A total of 5 Â 10 5 cells were cultured in 1 ml of complete RPMI (20% FCS) supplemented with 10 ng/ml TGF-b1 in a 1:1 ratio with anti-CD3/CD28 Dynabeads. Media was changed every 23 d with FCS reduced to 10% at the first media change. Following 10 d of culture, four cell populations, CD4 1 aE 1 , CD4 1 aE À , CD8 1 aE 1 , and CD8 1 aE À (Abs: anti-CD4 clone RPA-T4, anti-CD8 clone HIT8a), cultured from the b7 1 and b7 À progenitor populations were sorted using a BD FACS Aria Fusion Cell Sorter into 1% 2-ME RLT RNeasy lysis buffer (Qiagen). RNA was extracted using RNeasy Kits (Qiagen). Fifty-six total tubes were sorted based on the original b7 surface expression in combination with CD4, CD8, and aE expression at the end of culture. Eight of these tubes had <1000 cells per group and two failed to yield enough RNA for analysis. First-strand cDNA was generated from 34 ng of total RNA using iScript (Bio-Rad Diagnostics, Hercules, CA). cDNA (2.125 ng) was preamplified using the TaqMan PreAmp Master Mix (Life Technologies, Carlsbad, CA) along with diluted TaqMan assays (Life Technologies, Carlsbad, CA), according to the manufacturer's instructions. Quantitative PCR was performed using the Biomark HD system (Fluidigm 96.96 format).

Comparison of selected genes in sorted in vitro differentiated cells
A linear model was used to evaluate the effect of baseline expression of b7 integrin on expression of selected genes following 10 d of culture with or without TGF-b1 stimulation as described. Both fold change and statistical significance was determined for all genes tested and a false discovery rate (FDR) was calculated using the method of Benjamini and Hochberg (31). The analysis was performed independently for CD4 1 and CD8 1 T cells, with day 10 aE integrin expression included as a covariate.

Results
Adhesion molecules ICAM-1, VCAM-1, and MAdCAM-1 are upregulated in IBD Previous work has shown that ICAM-1 and VCAM-1 have low expression whereas MAdCAM-1 is highly expressed on vascular endothelium in normal intestine (1214). Although changes in patterns of expression have been described in IBD, shifts in overall gene expression have not been extensively evaluated. We therefore measured ICAM-1, VCAM-1 and MAdCAM-1 gene expression levels in healthy subjects and IBD patients. Consistent with historical protein data, we observed low but detectable levels of adhesion molecules in colonic and ileal biopsies from healthy subjects (Fig. 1A, 1C, 1E). ICAM-1 was highly upregulated in inflamed tissue from both UC and CD patients in comparison with uninflamed and healthy control tissue in both ileum and colon (Fig. 1A). In contrast, VCAM-1 gene expression was only moderately increased in inflamed CD, but not UC, samples (Fig. 1C). MAdCAM-1 upregulation was observed in both inflamed UC and CD tissue (Fig. 1E). We found no difference in gene expression between uninflamed UC or CD and healthy subject samples for any of the three adhesion molecules (Fig. 1A, 1C, 1E).
Because migration patterns change during inflammation, gene expression changes may reflect differences in expression between distinct cellular compartments. Inflamed biopsies from both ileum and colon exhibited consistent increases in ICAM-1 gene expression (Fig. 1A) and widespread upregulation of ICAM-1 protein (Fig. 1B) compared with uninflamed biopsies and controls (Supplemental Fig. 1A, 1B). These findings are consistent with previous reports demonstrating ICAM-1 expression on monocytes and plasma cells as well as endothelial cells in inflamed tissue (28). VCAM-1 protein levels were less altered (Fig. 1D, Supplemental Fig. 1C, 1D). MAdCAM-1 staining showed increased staining limited to endothelial cells (Fig. 1F,  Supplemental Fig. 1E, 1F), consistent with previous reports (32).

Expression of b1 and b7 integrins is unchanged on peripheral blood T cell populations in IBD
To better understand changes in integrin expression on lymphocytes from patients undergoing intestinal resection because of UC, CD, and diverticulitis (non-IBD), we evaluated b1 and b7 integrin expression on the surface of peripheral blood T cells.
Peripheral blood T cells were separated into three groups based on high surface expression of CD45RA (CD45RA 1 ; naive T cells), low surface expression of CD45RA and high b7 integrin (CD45RA b7 high ; mucosal homing T cells) or low b7 integrin expression (b7 low ; peripheral homing T cells) (example of gating strategy is shown in Fig. 2A). The frequency of these different subpopulations in peripheral blood was similar between UC and CD and between IBD and non-IBD groups (Fig. 2B).
Expression of b1 and b7 integrins was observed on all T cell subsets with the exception of b7 integrin on peripheral homing T cells (Fig. 2C, 2D). On mucosal homing T cells, both b1 and b7 integrin levels were similar between all patient groups (Fig. 2C, 2D). b1 expression was 3-fold higher on peripheral blood homing T cells in comparison with naive and mucosal homing T cells (Fig. 2C), whereas b7 expression was 4-fold higher on average on mucosal homing T cells in comparison with naive T cells (Fig. 2D). Similar findings were observed for all subsets for both CD4 1 and CD8 1 T cells, with the exception of b1 expression, which was only 2-fold higher on peripheral blood homing CD8 1 T cells in comparison with naive and mucosal homing cells CD8 1 T cells (Supplemental Fig. 2).

Induction of aE integrin on T cell populations through TGF-b1 stimulation is not dependent on baseline b7 expression
The aEb7 integrin is expressed by a very low proportion of peripheral blood T lymphocytes. T cell expression of aE integrin is known to be TGF-b inducible (3032), as was confirmed in our experiments (Fig. 3). Both CD4 1 and CD8 1 T cells responded to increasing concentrations of TGF-b1 with increasing frequency of surface aE expression (Fig. 3A, 3B). A dose-response relationship was observed for aE induction, with a higher frequency of CD8 1 aE 1 T cells at all doses tested in comparison with CD4 1 aE 1 T cells. Treatment with an ALK5 (TGF-b receptor I kinase) inhibitor demonstrated that aE upregulation in both CD4 1 and CD8 1 T cells required TGF-bRI signaling (Fig. 3A, 3B). CFSE dye was used to follow aE upregulation in dividing CD4 1 and CD8 1 T cells following stimulation with anti-CD3/CD28 and 10 ng/ml TGF-b1 (Fig. 3C). A lower overall percentage of CD4 1 T cells underwent aE upregulation, yet achieved maximal frequency of aE positivity more rapidly than CD8 1 T cells (Fig. 3D). aE expression levels were also higher on CD8 1 T cells and continued to increase over later cell divisions, whereas CD4 1 T cells had lower MFIs, yet achieved these expression levels at earlier cell divisions (Fig. 3E). aE MFI on CD4 1 aE 1 lymphocytes were not dependent on the number of cell divisions, whereas aE MFI of CD8 1 aE 1 T cells increased with successive cell division (Fig. 3E). A similar level of upregulation of aE in response to anti-CD3/CD28 and 10 ng/ml TGF-b1 was observed on both CD4 1 and CD8 1 T cells from peripheral blood of IBD patients (Supplemental Fig. 3A). No effect on aE induction was observed when therapeutically relevant anti-integrin Abs (anti-a4b7 or anti-b7) were included in the culture for the duration of the stimulation (Supplemental Fig. 3B).
TGF-b1driven induction of aE expression on CD4 1 T cells has been reported to be reduced in sarcoidosis patients carrying a genetic polymorphism in the promoter of the ITGAE gene tagged by the rs2981 single nucleotide polymorphism (22). We evaluated reported genetic risk alleles in the ITGAE gene in genotyped healthy donors to assess effects on surface aE expression. In contrast to previous studies, we found no association between rs2981 and aE expression by frequency or surface levels in either CD4 1 (Fig. 3F) or CD8 1 T cells (Fig. 3G). We also observed no significant differences in aE expression by frequency or surface levels for described intronic and exonic alleles (data not shown). Although TGF-b1 can induce aE expression on both CD4 1 and CD8 1 T cells, the baseline requirement for b7 expression prior to upregulation of aE has not been extensively evaluated (33). We sorted T cells by b7 expression and CD4 or CD8 lineage prior to stimulation with anti-CD3/CD28 and TGF-b1. Following in vitro stimulation, both b7 1 and b7 À T cells upregulated aE expression (Fig. 3H, 3I). Both b7 1 and b7 CD4 1 and CD8 1 T cells upregulated aE expression in response to TGF-b1 by day 68, with $20% relatively fewer aE 1 cells observed in baseline b7 cells (Fig. 3H,  3I). These data confirm that both b7 À CD4 1 and CD8 1 T cells can be induced to upregulate aE following TGF-b1 stimulation in vitro, although the frequency of cells with inducible aE expression is consistently lower than for cells that already have surface b7 expression.

pSMAD3 and increased TGF-binducible gene expression in IBD intestinal mucosa
TGF-b regulates transcription of the aE and b7 integrin genes (ITGAE and ITGB7, respectively) (3335). Serine/threonine phosphorylation of SMAD2 and SMAD3 following TGF-b activation of ALK5 (TGF-bR1) as well as NFAT-1 activity have been shown to drive expression of ITGAE (36,37). Indeed, both NFAT-1 and SMAD binding sites have been identified in the ITGAE promoter and enhancer (36). The ITGB7 promoter contains two functional TGF-b response elements (35). Conversely, TGF-b leads to reduced transcription of the a4 integrin subunit (35).
We used IHC to evaluate changes in SMAD3 phosphorylation (pSMAD3) in biopsies from IBD patients in comparison with healthy subjects. In healthy subjects, pSMAD3 was largely limited to the epithelium in both the colon (Fig. 4A) and ileum (Fig. 4B), with a few scattered mononuclear cells also staining positive. The majority of epithelial cells displayed positive staining for pSMAD3 in the ileum, with a lower frequency of positive cells in the colon. pSMAD3 1 mononuclear cells were also more frequent in the ileum than in the colon of healthy subjects. Colonic biopsies from UC patients with active disease showed increased infiltration of mononuclear cells, many of which were pSMAD3 1 (Fig. 4C). Epithelial staining for pSMAD3 in UC biopsies was more intense than in healthy colonic tissue. Similarly, colonic biopsies from CD patients showed increased epithelial staining intensity (Fig. 4D). Ileal biopsies from CD patients had localized increases in pSMAD3 1 mononuclear cells in inflamed areas (Fig. 4E) with overall similar intensity of pSMAD3 in the epithelium compared with healthy ileum (Fig. 4B).
Analysis of TGF-b family members and receptor expression in isolated cells using single-cell RNA-sequencing data (38) showed that TGF-b1 is the predominant family member expressed in the colon and is expressed primarily by immune cells, whereas TGF-b receptor expression was found on both immune and epithelial cells (Fig. 4F). TGF-b3 was detected at low levels whereas TGF-b2 was not detectable. Expression of TGF-b1 and both TGF-b receptors was similar between healthy, inflamed, and uninflamed biopsies. TGF-b induces a much broader gene expression program beyond ITGAE and ITGB7, including common and distinct elements in different cell types (39). A common gene induced across cell types is EGR1, a transcription factor (37) with binding sites near TGF-b1 target genes, including SMAD7, ID1, and SKIL (40). Expression of the TGF-binducible genes EGR1 and EGR2 in intestinal biopsies from healthy subjects and IBD patients (Supplemental Fig. 3E, 3F). Increased gene expression of both EGR1 and EGR2 was observed in inflamed colonic biopsies versus healthy colon or uninflamed biopsies from IBD patients. There was no significant difference in EGR1 or EGR2 gene expression in ileal biopsies, consistent with the stronger pSMAD3 staining observed in the colon in comparison with the ileum. Gene expression in induced aE 1 T cells is independent of baseline b7 expression status pSMAD3 levels in the inflamed colon suggest that TGF-b signaling is increased in active colonic disease. We hypothesized that upregulation of aE integrin by local active TGF-b would not require preexisting expression of the a4b7 dimer and that both b7 À and b7 1 peripheral blood T cells can induce aE following migration into the intestinal mucosa. To evaluate potential differences in the aE 1 T cell phenotype due to initial b7 integrin expression, we developed a two-step in vitro culture and cell sorting approach to enable comparison of gene expression between these populations. Data from seven donors confirmed expression of aE integrin on <1% of peripheral blood CD4 1 T cells and $3% of peripheral blood CD8 1 T cells, with 80% of CD4 1 T cells and 60% of CD8 1 T cells showing high expression of b7 integrin (Fig. 5A). Ubiquitous expression of aL integrin was observed on peripheral CD4 1 and CD8 1 T cells (Fig. 5B). b1 integrin was observed on $90% of CD4 1 T cells and nearly all CD8 1 T cells (Fig. 5C). Initial cell sorting was based on b7 expression, with low/negative b7 (peripheral homing) and high b7 (mucosal homing) T cells sorted from seven healthy donors (Fig. 5D, 5E). Following 10 d of culture with anti-CD3/CD28 beads and 10 ng/ml TGF-b1, a subset of T cells had upregulated aE surface expression. Higher induction of aE was again observed on CD8 1 in comparison with CD4 1 cells, and cells were sorted into aE 1 CD4 1 , aE À CD4 1 , aE 1 CD8 1 , and aE À CD8 1 fractions (Fig. 5F). Consistent with upregulation of aE surface expression in response to TGF-b1, there was a clear increase in b7 MFI in all cell groups over the 10 d of culture (Fig. 5G). Analysis using a linear model approach to evaluate the effect of baseline b7 expression was performed independently for CD4 1 and CD8 1 T cells using aE surface expression at day 10 as a covariate (Fig. 5H). T cell lineage genes, TGF-b1 response genes, and genes associated with aEb7 integrin (18,20) were selected for evaluation based on expression in cell types of interest. For CD4 1 T cells, baseline b7 integrin expression had a nominally significant (unadjusted p < 0.05) effect on the following genes: ITGA4, GATA3, GZMA, BATF, HIFA, ICOS and IRF4, although none of these genes had an FDR below 0.1. For CD8 1 T cells, STAT1, ITGB1, ITGA4, CD127, and CD4 demonstrated a significant (p < 0.05) effect of baseline b7 integrin expression. Both STAT1 and ITGB1 had an FDR below 0.05. The majority of genes showed no difference in expression related to initial b7 expression levels. (F and G) PBMCs from ITGAE rs2981 genotyped healthy subjects with either C and/or T alleles were stimulated with anti-CD3/CD28/TGF-b1 for 6 d, then CD4 1 and CD8 1 T cells were evaluated for aE expression. (H and I) Peripheral blood T cells were sorted by baseline expression of b7 integrin and stimulated with anti-CD3/CD28 ± TGF-b1 (10 ng/ml) for 8 d. aE integrin expression was observed following TGF-b1 stimulation on both b7 À and b7 1 CD4 1 and CD8 1 T cells. Data are representative of three independent experiments.

Discussion
Inflammatory events have the potential to change established patterns of cellular migration and lymphocyte differentiation in patients with IBD. Under homeostatic conditions, lymphocytes activated within mucosal lymph nodes to express high levels of a4b7 integrin as well as naive cells with intermediate a4b7 integrin expression circulate through the intestinal mucosa. In IBD, activation of inflammatory cascades can alter expression patterns of adhesion molecules that serve as entry points into mucosal tissues. Increased levels of adhesion molecules on the intestinal endothelium may alter lymphocyte migration into the gut. We evaluated VCAM-1, ICAM-1, and MAdCAM-1 gene and protein expression in healthy mucosa from normal subjects and uninflamed and inflamed mucosa from patients with IBD. Although inflamed mucosa had generally increased expression of adhesion molecules, ICAM-1 was broadly upregulated, consistent with its known expression by infiltrating monocytes and plasma cells, whereas VCAM-1, and in particular MAdCAM-1, were more restricted in expression to endothelial cells. Subsets of peripheral blood T cells can express both a4b7 and/or a4b1, consistent with previous data showing that T cells express multiple integrins for entry into mucosal tissues (20). Once cells have migrated into the mucosa, induction of other integrins can occur.   ITGB1  ITGA4  IL7R  CD4  CCR7  SMAD3  THBS1  KLRB1  GZMA  PRF1  GPR15  IL2RA  IRF4  18SRDNA  PTPN22  SELL  AHR  HIF1A  ITGAV  IL23A  ITGB2  FAS  TGFB1  GATA3  IL12RB2  IFNG  TMEM55B  TCF7  CD8a  FOXM1  FoxP3  ITGB7  BATF  FOXO1  STAT3  CD69  SMURF2  JAK1  CXCR6  TGFBR1  ICOS  CCL20  SMAD7  KLRG1  SMURF1  CD3e  RUNX1  cMAF  RORA  SOCS1  GZMB  JAK2  TBX21  CTLA4  TGFBR2  SMAD2  RORC  STAT4  JUN  ITGAL  TNFA  LAG3  IKZF2  SMAD4  predominant TGF-b family member by gene expression in colonic macrophages, T cells and DCs, whereas TGF-bR1 and TGF-bR2 are expressed on monocytes, T cells and epithelial cells. Finally, following the induction of aE expression on T cells that were either a4b7 1 or a4b7 À by TGF-b1, gene expression was largely not dependent on initial b7 expression. This suggests that T cells that undergo a4b7-independent migration into mucosal tissue sites can give rise to inflammatory aE 1 T cells and therefore that a4b7directed therapy alone may be insufficient to modulate aEb7 1 lymphocyte retention in vivo. How frequently this occurs in the human gut cannot be measured using current clinical techniques. IBD may increase the likelihood that both a4b7 1 and a4b7 À T cells enter the intestinal mucosa through compensatory homing mechanisms. Increased expression of ICAM-1 and VCAM-1 (Fig. 1), which has been shown to resolve at the gene expression level following anti-TNF treatment (41), may enable entry of these cells into the intestinal mucosa in IBD. Our data in healthy subjects were from two age-matched cohorts and showed increased expression of adhesion molecules at the gene expression and protein level. We found that peripheral blood T cells coexpress b1 and a4b7 on a significant fraction of mucosal homing T cells and naive T cells (Figs. 2 and 5). Previous reports show an increase in peripheral blood a4 1 b1 high cells and a corresponding increase in a4 1 b1 high T cells in intestinal mucosa from CD patients (7). In the absence of a human Ab specific for the a4b1 dimer, it remains to be determined whether a4b1 is expressed on the cell surface of circulating cells, or whether b1 is paired with another of the 10 a-integrin chains that are known to form b1 heterodimers. Experiments using humanized mice and mouse models show that lymphocytes use compensatory homing mechanisms to enter the mucosa; this process can be blocked using a small molecule inhibitor specific for a4b1 (7).
Once T cells have entered the LP, TGF-b1 can induce upregulation of aE through SMAD phosphorylation. Our data suggest that TGFbinduced gene expression may occur more readily upon entry into the colon than the ileum during active IBD where increases in pSMAD3 and TGF-binducible genes were observed. Previous work has shown that colonic aE 1 CD4 1 T cells express increased inflammatory cytokines such as IL-17 and IFN-g (21). The frequency of aE 1 CD4 1 T cells is increased in UC, which is consistent with our observations of increased TGF-b signaling. When we induced aE expression by treating activated T cells in vitro, we detected few gene expression differences for inflammatory cytokines or effector molecules between cells that were originally a4b7 1 and a4b7 À . These data suggest that cells that migrate into the mucosa using compensatory homing mechanisms can develop an aE-associated inflammatory phenotype. Abs that block aE integrin suppress accumulation of inflammatory lymphocytes in the intestinal mucosa (9,24), indicating that once aEb7 is induced, it increases cellular retention in the mucosa. The decrease in cellular accumulation was more pronounced at later time points in a humanized mouse model, consistent with effects on retention rather than initial homing (24).
Recent studies have specifically addressed migration of cells into tissue, which is challenging to directly evaluate in clinical studies. The effect of vedolizumab on lymphocyte trafficking was recently described in a study that followed labeled cells using imaging, but observed no change in T cell migration to the intestinal mucosa (42). Flow cytometry within the same study protocol also failed to demonstrate a change in T cell frequency following 14 wk of vedolizumab treatment, supporting the idea that alternative migratory pathways following treatment with vedolizumab may be active. A separate study observed decreased a4b7 1 within intestinal T cell subsets but no change in aEb7 1 T cells in the colon, whereas also noting incomplete occupancy on intestinal CD8 1 T cells (43). Our in vitro differentiation studies show that initial b7 expression is not required for upregulation of aE. Indeed, patients treated with vedolizumab have been found to have an increase in aE 1 CD8 1 T cells in peripheral blood during maintenance time points, suggesting that blockade of a4b7 does not impair aE upregulation (43). Taken together, these findings suggest that blockade of a4b7 alone may not be sufficient to inhibit T cell trafficking to the intestinal mucosa, and b7 À negative cells may undergo induction of aEb7 once in the mucosa, increasing retention of inflammatory T cells.
Inhibition of lymphocyte trafficking has proven to be an effective strategy for treating both UC and CD, yet in a significant fraction of patients, medication fails to induce and maintain remission. Although therapies designed to inhibit migration through blockade of a4b7 or MAdCAM-1 have shown efficacy, studies showing that blockade of a4b7/MAdCAM-1 interactions alone may leave additional homing and retention pathways intact. Consistent with this, studies evaluating changes in inflammatory gene expression following treatment with anti-integrins show that few genes are significantly changed by vedolizumab (anti-a4b7) treatment in comparison with inhibition of TNF-a at early time points in UC patients (44). Our more recent phase 3 study showed that etrolizumab significantly decreases lymphocyte gene expression at week 14 in CD patients (9), consistent with data from a phase 2 study of etrolizumab in UC (18). Continued migration of lymphocytes through a4b7-independent compensatory mechanisms and/or increased retention of lymphocytes following aEb7 upregulation might yield insights into patients who do not respond to vedolizumab treatment.
In conclusion, we demonstrate that aEb7 can be induced on both a4b7 1 and a4b7 À T cells following exposure to TGF-b1, and we show evidence for increased TGF-b activity in the colon of patients with active IBD. These data, in combination with an increase in adhesion molecules in active IBD, highlight the potential for increased compensatory migration mechanisms to the gut mucosa beyond a4b7/MAdCAM-1 in active inflammation that can result in aE 1 T cell accumulation. Inflammatory gene expression in aE 1 T cells was largely unchanged between T cell populations that were originally a4b7 1 and a4b7 À T cells. These data suggest that IBD patients treated with Abs that block a4b7 or MAdCAM-1 may benefit from additional integrin and/or inflammatory cytokine blockades to further inhibit inflammatory cell migration to the gut mucosa.