Baseline levels of dynamic CD4⁺ T cell adhesion to MAdCAM-1 correlate with clinical response to vedolizumab treatment in ulcerative colitis: a cohort study

Despite an increasing therapeutic armamentarium for the treatment of inflammatory bowel diseases (IBD), disease activity can still not be sufficiently controlled in a considerable number of patients [1]. Response to the available agents is observed only in portions of patients [2‐4] and, additionally, patients may lose response over time [5]. Moreover, there is evidence indicating that the probability of response to a subsequent treatment is lower, if previous therapies have failed [6], and health care systems may be encumbered with costs for ineffective therapies [7].

Thus, treatment selection in individual patients remains an important challenge. Since head-to-head studies and biomarkers for the prediction of response to therapy are largely lacking, objective guidance driving such treatment decisions is low.

The anti-α4β7 antibody vedolizumab is successfully used for the treatment of IBD since 2014 [3, 8] and has been shown to inhibit immune cell homing to the inflamed gut [9, 10] indicating that cell trafficking is a central event in the pathogenesis of IBD [11]. Randomized controlled trials [3, 8], as well as several real-world cohorts [12‐14], demonstrated the efficacy and safety of vedolizumab in ulcerative colitis (UC) and Crohn’s disease (CD). Vedolizumab is considered to be rather “slow-acting” [15], which might be explained by its mode of action, not directly targeting intestinal immune cells but only their replenishment by recruitment of cells from the peripheral blood [16]. Therefore, to avoid long periods of ineffective treatment in non-responders, the identification of biomarkers to predict response to vedolizumab therapy is a particularly unmet need. Moreover, since vedolizumab rather acts in the peripheral blood than in intestinal tissue, the drug might provide an especially convenient opportunity for the determination of biomarkers with low invasiveness.

We had previously introduced a dynamic adhesion assay to study the adhesion of immune cells to cell adhesion molecules [17]. In this context, we had reported the anecdotal observation that the extent of dynamic adhesion of peripheral blood CD4⁺ T cells from IBD patients to the α4β7 ligand mucosal vascular addressin cell adhesion molecule (MAdCAM-)1 in this assay before initiation of vedolizumab treatment seemed to correlate with subsequent clinical response to therapy.

Here, we conducted a retrospective cohort study in UC patients treated with vedolizumab to further investigate this hypothesis. We show that dynamic adhesion to MAdCAM-1 is higher in responders than in non-responders and that vedolizumab treatment in vitro leads to a stronger reduction of adhesion in responders compared with non-responders. High levels of dynamic adhesion had a high specificity and positive predictive value for the response to treatment.

Treatment decisions are a constant challenge in IBD. While more and more agents are available to treat UC and CD, there is no clear sequence of preferred treatments [22] and, thus, the choice of treatment often remains an individual consideration. In some cases, the presence of complications like fistulae [23] or associated conditions like arthritis [24] may help to decide, but in most cases, not evidence-based aspects like patient preferences may be even more important. Apart from one very recent example comparing the efficacy of vedolizumab and adalimumab for achieving clinical remission at week 52 in UC [25], prospective head-to-head studies comparing biological treatments for IBD are lacking. Since it is not likely that appropriate additional randomized head-to-head studies will be performed in the near future to sufficiently guide treatment decisions in particular treatment situations, the identification of biomarkers predicting the individual response or non-response of patients to certain agents is an alternative strategy to make IBD therapy more efficient.

The feasibility of this concept has been previously demonstrated in studies showing that the quantification of target cells for the anti-TNF-α antibody adalimumab during confocal endomicroscopy and the baseline expression of αE integrin predicted subsequent response to adalimumab and the anti-β7 antibody etrolizumab, respectively [26, 27]. With regard to vedolizumab, a broad number of observations has been made: While gene expression profiling identified a considerable number of genes that were significantly differentially expressed following vedolizumab therapy, none of them was able to predict treatment outcomes [28]. However, another study reported that response to vedolizumab was associated with pre-therapeutic composition of the intestinal microbiome [29]. Moreover, we had suggested that dynamic changes in the expression of α4β1 on peripheral blood CD4⁺ T cells in the early course of vedolizumab treatment and the baseline dynamic adhesion of CD4⁺ T cells to MAdCAM-1 in vitro might predict subsequent therapeutic response [17, 20].

Here, we further investigated these hypotheses. Using a dynamic adhesion assay suitable to quantify the adhesion of CD4⁺ T cells to MAdCAM-1 under physiological shear stress [18], we show that in our cohort of 21 UC patients, more T cells from responders than from non-responders adhered to MAdCAM-1 prior to initiation of vedolizumab therapy. Consistently, in vitro treatment with vedolizumab reduced adhesion in responders stronger than in non-responders. AUCs determined with ROC analyses were around 0.8. This corresponds to a fair to good test accuracy [30] and is comparable with the accuracy of fecal calprotectin quantification for the prediction of endoscopic remission in UC [31]. Contingency table analyses showed that the ideal cut-off values determined in ROC analyses for the identification of responders either based on adhesion to MAdCAM-1 or on the reduction of adhesion after in vitro treatment with vedolizumab had high specificity, but only moderate sensitivity due to a number of false-negative results. However, this also mirrored in high positive predictive rates. Consistently, lower cut-off values were better suited to identify non-responders. When combining the information on adhesion to MAdCAM-1 and on the reduction of adhesion by vedolizumab, it turned out that high adhesion and high reduction were strongly associated with response to treatment, while low adhesion and low reduction were clearly associated with non-response. On the contrary, medium levels of adhesion and reduction were not conclusive in this regard.

Taken together, our data suggest that determining the baseline adhesion of CD4⁺ T cells to MAdCAM-1 and their response to vedolizumab in vitro with the proposed assay may be suitable as a functional biomarker and help to support treatment decisions in most patients. It is important to mention, that – based on previous observations [17] – differential levels of CD4⁺ T cell adhesion to MAdCAM-1 are most likely not a result of differential α4β7 integrin expression, but rather differential intrinsic functionality of this pathways in responders vs. non-responders.

We also continued to explore the suitability of changes in α4β1 integrin expression on CD4⁺ T cells early during vedolizumab treatment as a marker of subsequent treatment success and report the extension of a previous cohort [20] in this study. In the expanded cohort of 26 UC patients, on average, CD4⁺ α4β1 integrin levels declined in patients with a clinical response at week 16 within the first 6 weeks of therapy, while it increased in non-responders. Again, specificity and positive predictive value for the assessment of clinical response were high. The mechanistic explanation for this finding might be that α4β1 has been shown to represent an alternative pathway for lymphocyte gut homing, when α4β7 is blocked [21]. Thus, the determination of α4β1 integrin expression might present a promising way to predict the response to vedolizumab as a “slow-acting” drug early in the course of treatment.

It has to be underscored that several limitations have to be taken into account, when interpreting these data. For one aspect, they are based on an only small cohort from one single center. Moreover, clinical parameters were determined retrospectively. Thus, our study needs to be understood as a pilot study and the data cannot easily be generalized and conclusions must be drawn carefully.

Yet, these results show the promising potential of these techniques and provide a clear rationale for subsequent prospective studies with higher sample size, which are currently under way.

Analysis of the adhesion of CD4⁺ T cells to MAdCAM-1 on a functional level with dynamic adhesion assays in IBD patients prior to the start of treatment appears to be a promising tool for the identification of patients, which are likely to respond to therapy with vedolizumab. Similarly, changes in α4β1 integrin expression after initiation of treatment might be an early indicator of clinical success in vedolizumab-treated patients.

Provided that these findings can be confirmed and validated in larger and prospective studies, our observations might lay the basis for the prediction of success of vedolizumab treatment with the help of peripheral blood samples.