Up-Regulation of α4β7 Integrin on Peripheral T Cell Subsets Correlates with the Development of Acute Intestinal Graft-versus-Host Disease following Allogeneic Stem Cell Transplantation

Abstract

Acute graft-versus-host disease (aGVHD) is a major complication after hematopoietic stem cell transplantation (HSCT). The pathophysiology of aGVHD involves priming of naïve donor T cells in host secondary lymphoid tissue, followed by migration of effector T cells to target organs. Mediators of lymphocyte trafficking are believed to play a significant role in this migration. In this retrospective case-controlled study, we analyzed the expression of α4β7 integrin and CCR9, 2 surface T cell molecules specific for intestinal trafficking, from blood samples collected previously from 59 patients after HSCT (20 without aGVHD, 20 with skin aGVHD, and 19 with intestinal aGVHD). All samples had been obtained before the onset of aGVHD symptoms (with 1 sample collected on the day of symptom onset). Analysis by flow cytometry demonstrated that α4β7 integrin was significantly increased on both naïve and memory T cells in patients who subsequently developed intestinal aGVHD, with the most significant differences observed in memory subsets. Immunohistochemical staining on rectal biopsy specimens from patients with intestinal aGVHD showed that expression of α4β7 integrin was concentrated on mononuclear cells in blood vessels within the intestinal mucosa. These results suggest that α4β7 integrin likely is involved in lymphocyte trafficking in intestinal aGVHD and may have potential clinical use as a correlative biomarker or as a target for the treatment and prophylaxis of intestinal aGVHD after HSCT.

INTRODUCTION

Acute graft-versus-host-disease (aGVHD) is a major cause of morbidity and mortality after allogeneic hematopoietic stem cell transplantation (HSCT) [1]. It is caused by the recognition of host antigens by donor lymphocytes, with target organ damage mediated by activated donor effector T cells. With its large mucosal surface and extensive secondary lymphoid tissue, the gastrointestinal (GI) tract is not only a major target organ, but also a crucial amplifier of a systemic cytokine response in aGVHD [2]. Intestinal aGVHD is believed to be similar to other immunologic responses in that it has similar requirements for lymphocyte trafficking. Priming and maturation of naïve donor T cells that are targeted for the gut mucosa are believed to be mediated by activated host dendritic cells within gut-associated secondary lymphoid tissue (GALT), including Peyer's patches (PPs) of the small intestine, the appendix, and mesenteric lymph nodes [3]. Recent evidence has suggested that effector T cells acquire an intestinal homing phenotype through interactions with GALT dendritic cells in a process mediated by retinoic acid [4-8]. Once T cells have been educated in GALT through antigen engagement and costimulation, they continue to circulate through the bloodstream, but preferentially migrate to intestinal effector sites—the lamina propria and epithelium of the small and large intestine—as directed by their specific surface molecules [9,10]. This process of specific lymphocyte trafficking represents a potentially novel target for the monitoring, treatment, and prophylaxis of intestinal aGVHD.

Several classes of membrane surface molecules, including selectins, integrins, chemokine receptors, and pertussis-sensitive G-proteins, are involved in lymphocyte trafficking; all of these help ensure organ-specific localization of lymphocytes [11-13]. One specific molecule, α4β7 integrin, plays a crucial role in the recirculation of naïve T cells to intestinal secondary lymphoid tissue, as well as the selective trafficking of specific effector T cells into sites of intestinal inflammation [14-16]. The primary ligand for α4β7 integrin is mucosal addressin cell adhesion molecule 1 (MAdCAM-1), which is selectively expressed in the high endothelial venules and follicular dendritic cells of GALT, with specific up-regulation at sites of active inflammation [17-20]. Chemokines are thought to participate in this process through the activation of α4β7 integrin [21-23]. Although the precise chemokines in this process involved are unknown, the CCR9–CCL25 interaction has emerged as a good candidate [24,25]. CCL25 (TECK) has been shown to be expressed only in epithelial cells of the thymus and small intestine [26], and the vast majority of CCR9-expressing T cells in the peripheral blood express α4β7 integrin as well [27].

In the present study, we analyzed peripheral blood (PB) samples collected previously from 59 human patients after HSCT, to assess the involvement of α4β7 integrin and CCR9 in lymphocyte trafficking in intestinal aGVHD. We analyzed the expression of α4β7 integrin and CCR9 on specific T cell subsets by flow cytometry. α4β7 integrin was found to be significantly up-regulated on both naïve and memory T cell subsets in patients who subsequently developed intestinal aGVHD compared with those who developed primarily cutaneous aGVHD or did not develop aGVHD. Immunohistochemical staining on previously obtained rectal biopsy specimens in patients with intestinal aGVHD showed the presence of α4β7 integrin on inflammatory mononuclear cells aggregated within mucosal blood vessels. Our findings suggest that α4β7 integrin is up-regulated on peripheral blood T cells before the onset of symptomatic intestinal aGVHD, implying a role for this molecule in the pathophysiology of aGVHD.

METHODS

Patients and Samples

This was a retrospective case-controlled analysis using patient samples identified from the stem cell transplantation database at the Dana-Farber Cancer Institute (DFCI). The study design was approved by the Institutional Review Board of the Dana-Farber/Harvard Cancer Center. Using a random number generator, we selected 59 patients, categorized into 3 groups: control, comprising 20 patients without aGVHD after HSCT; skin, comprising 20 patients with primarily cutaneous aGVHD after HSCT (grade III or IV); and gut, comprising 19 patients with intestinal aGVHD after HSCT (grade III or IV). Patients in the gut group were allowed to have skin or liver aGVHD, whereas those in the skin group were not allowed to have greater than grade I disease of the intestine or liver. The diagnosis and grading of aGVHD were recorded previously and were based on the patients’ clinical and pathological features, in accordance with previously published criteria [28]. The patients had been treated with either myeloablative (MA; either cyclophosphamide (Cy)/total body irradiation (TBI) or busulfan (Bu)/Cy) or reduced-intensity conditioning (RIC; Bu/Flu) regimens. Donor stem cells were obtained from either HLA-matched related donors (MRDs) or HLA-matched unrelated donors (MUDs). All patients received calcineurin inhibitor–based (either cyclosporine (CsA) or tacrolimus, dosed to target serum level) GVHD prophylaxis.

Blood Samples

All of the samples used in this study were obtained from a patient sample repository established at DFCI beginning in 2000 and were collected from patients who had signed informed consent to provide blood and bone marrow samples for research purposes. All samples were collected in EDTA tubes from patients after HSCT, and mononuclear cells were isolated by density gradient centrifugation. At the time of collection, cells were cryopreserved in FBS + 10% DMSO at –180°C. After appropriate patients were identified, samples were selected with the goal of using blood samples that had been collected within 30 days before the onset of aGVHD symptoms. After the blood samples were selected, peripheral blood mononuclear cells (PBMCs) were washed before analysis by flow cytometry.

Flow Cytometry

Cells were stained using a 7-color antibody panel comprising conjugated antibodies against CD45, CD3, CD4, CD8, CD45RO, α4β7 integrin, and CCR9. All of these reagents were purchased from Beckman Coulter (Fullerton, CA) except anti-CCR9, which was purchased from R&D Systems (Minneapolis, MN), and anti-α4β7 integrin (ACT-1), which was provided by Millennium Pharmaceuticals (Cambridge, MA) and conjugated at DFCI with phycoerythrin (PE). The cells were then analyzed using a Beckman Coulter FC500 or a FACSAria Special Order Research Platform (BD Biosciences, San Jose, CA) flow cytometer. T cell subsets were defined as follows: naïve CD4⁺ T cells (CD45⁺, CD3⁺, CD4⁺, and CD45RO^-), memory CD4⁺ T cells (CD45⁺, CD3⁺, CD4⁺, and CD45RO⁺), naïve CD8⁺ T cells (CD45⁺, CD3⁺, CD8⁺, and CD45RO^-), and memory CD8⁺ T cells (CD45⁺, CD3⁺, CD8⁺, and CD45RO⁺). Flow cytometry results were analyzed using FlowJo version 7.1.3 (FlowJo,) and CXP Analysis (Beckman Coulter, Miami, Florida).

Immunohistochemistry

Of the 19 patients in the gut group, 17 had previously obtained rectal biopsy specimens available for analysis, all of which revealed aGVHD pathologically. Conditions were defined for a PE-conjugated ACT-1 and anti-CCR9 antibodies, and immunohistochemical staining was performed. Sections (4 μ thick) of formalin-fixed tissue were used for immunoperoxidase analysis after baking at 60°C for 1 hour, deparaffinization, and rehydration (100% xylene × 4 for 3 minutes each, 100% ethanol × 4 for 3 minutes each, and running water for 5 minutes). The sections were blocked for peroxidase activity with 3% hydrogen peroxide in ethanol for 15 minutes, and then washed under running water for 5 minutes. All of the sections were treated in 1 mM EDTA buffer using a digital decloaking chamber (Pacific Southwest Lab Equipment, Vista, CA). The slides were cooled for 15 minutes, and then transferred to PBS. All of the sections were blocked with 1.5% horse serum for 15 minutes in a humid chamber at room temperature. The sections were then incubated with the corresponding primary antibodies: mouse monoclonal anti-CCR9 (clone 248621, 1:100 dilution; R&D Systems) and mouse monoclonal anti–α4β7 integrin (1:5 dilution; Millennium Pharmaceuticals) for 1 hour. Primary antibodies were detected using Vectastain Elite ABC Kit reagents (Vector Laboratories, Burlington, CA). The secondary antibody was used at a 1:200 dilution, made in 2% horse serum, and incubated for 30 minutes. Avidin-biotin complex was incubated for 30 to 40 minutes. The sections were developed using 3,3′-diaminobenzidine (Sigma-Aldrich, St Louis, MO) as a substrate, and counter-stained with Gill's hematoxylin (Fisher Scientific, Pittsburgh, PA). Images were viewed using a Zeiss Axiophot microscope equipped with Zeiss Plan Apo 10× and Zeiss Neofluar 40× objectives (Carl Zeiss, Germany, Carl Zeiss, Inc. Thornwood, NY) and acquired using a SPOT RT color digital camera (Diagnostic Instruments, Sterling Heights, MI). Final images were obtained using SPOT Advanced software (Diagnostic Instruments, Inc., Sterling Heights, MI).

Statistical Analysis

Two-sided Fisher's exact, Kruskal-Wallis, and Wilcoxon rank-sum (WRS) tests were used for comparisons of baseline patient and transplant characteristics. The 2-sided exact WRS test was used for pairwise group comparisons of each peripheral T cell subset. To adjust for multiple comparisons, permutation testing was performed using SAS version 9.1.3 PROC MULTTEST (SAS Institute, Cary, NC), with 10,000 resamples. If the adjusted P value from the permutation test was <.05, then the null hypothesis (of no association) for that particular T cell subset was rejected. In addition, multivariable logistic regression analysis with a generalized logit function was performed to further assess each T cell subset as a predictive biomarker for intestinal aGVHD in the presence of known risk factors for aGVHD. All T cell subsets were log-transformed before logistic regression analysis was performed. Correlation analysis was performed using Spearman's correlation test. The recursive partitioning method and receiver operating characteristic curves (ROCs) were used to identify an optimal cutoff value for phenotype expression level.

RESULTS

Patient and Sample Characteristics

Clinical characteristics of the 59 study patients are summarized in Table 1. Of these 59 patients, 33 received an MA conditioning regimen and 26 received an RIC regimen before HSCT. All 3 study groups were similar in terms of age range at the time of transplantation and underlying conditions. Patients received stem cells from an MRD, and 40 patients received stem cells from an MUD. Most of the patients received peripheral blood stem cells (PBSCs). All 59 patients received calcineurin inhibitor–based (either CsA or tacrolimus) GVHD prophylaxis with additional methotrexate (MTX), sirolimus, corticosteroids, or some combination of these agents. As shown in Table 1, more patients in the control group had received sirolimus; however, there was no significant difference in the use of sirolimus between the gut and skin groups (P = .51 for sirolimus vs no sirolimus).

Table 1.

Baseline Patient Characteristics

Characteristic	Gut (n = 19)	Skin (n = 20)	Control (n = 20)	P
Age, median (range)	44 (18, 68)	53 (18, 66)	46 (19, 65)	.33
Time from day of HSCT to blood sample (day), median (range)	34 (5, 106)	26 (5, 140)	35 (15, 93)	.08
Time from HSCT to onset of symptomatic aGVHD (day), median (range)	36 (18, 156)	40 (10, 166)	N/A	.7
Time from blood sample to onset of symptomatic aGVHD (day), median (range)	14 (0, 63)	8.5 (2, 86)	N/A	.88
Diagnosis				.19
ALL	1 (5%)	1 (5%)	3 (15%)
AML/AML from MDS	5 (26%)	6 (30%)	7 (35%)
CLL/PLL	4 (21%)	4 (20%)	1 (5%)
CML	2 (11%)	1 (5%)	3 (15%)
CMML	0	1 (5%)	0
CTCL	0	1 (5%)	0
HD	1 (5%)	1 (5%)	2 (10%)
NHL*	0	1 (5%)	3 (15%)
MDS	6 (32%)	3 (15%)	1 (5%)
MF	0	1 (5%)	0
Donor				.37
MRD	5 (26%)	9 (45%)	5 (25%)
MUD	14 (74%)	11 (55%)	15 (75%)
Stem cell source				.53
Bone marrow	1 (5%)	2 (10%)	0
PBSCs	18 (95%)	18 (90%)	20 (100%)
Sex mismatch	10 (53%)	10 (50%)	11 (55%)	1
Conditioning				.58
Myeloablative	9 (47%)	11 (55%)	13 (65%)
Reduced intensity	10 (53%)	9 (45%)	7 (35%)
GVHD prophylaxis				.06
Including sirolimus	6 (32%)	9 (45%)	16 (80%)
Tacrolimus-based without sirolimus	7 (37%)	10 (50%)	4 (20%)
Other	6 (32%)	1 (5%)	0
Gut grade
3	5 (26%)
4	14 (74%)
Skin grade
3		20 (100%)
Overall grade
2	0	20 (100%)
3	14 (74%)	0
4	5 (26%)	0

ALL indicates acute lymphoblastic leukemia; AML, acute myelogenous leukemia; MDS, myelodysplastic syndrome; CLL, chronic lymphocytic leukemia; PLL, prolymphocytic leukemia; CML, chronic myelogenous leukemia; CMML, chronic myelomonocytic leukemia; CTCL, cutaneous T cell lymphoma; NHL, non-Hodgkin lymphoma; MF, myelofibrosis; Tac, tacrolimus; rapa, rapamycin; NA, not applicable.

^*NHL includes NHL with MDS.

PB samples were identified based on their relationship to the onset of aGVHD symptoms. In the study patients, the timing of onset of aGVHD varied from day 10 to day 166 after HSCT, with a median time to development of aGVHD of 36 days after HSCT (range, 18 to 156 days) in the gut group and 40 days after HSCT (range, 10 to 166 days) in the skin group (P = .70). Most of the samples (89% of samples in the gut group and 80% of samples in the skin group) were collected within 30 days before the onset of aGVHD symptoms. With the exception of 1 patient in the gut group (sample drawn on the day of symptom onset), all samples were drawn before the manifestation of any clinical evidence of GVHD, as verified by the medical records. The median time from the date of sample collection to the onset of aGVHD symptoms was 14 days (range, 0 to 63 days) in the gut group and 8 days (range, 2 to 86 days) in the skin group (P = .66). Only 2 of the 59 patients had blood samples collected after day 100 of HSCT; most of the samples were collected within the first 50 days after HSCT (74% in the gut group, 85% in the skin group, and 70% in the control group).

Surface Expression of α4β7 Integrin and CCR9 on Peripheral T Cell Subsets following HSCT in Relation to the Development of aGVHD

Flow cytometry analysis of surface expression of α4β7 integrin and CCR9 on peripheral T cell subsets (Table 2) showed significant up-regulation of expression of α4β7 integrin on all studied peripheral T cell populations in the gut group compared with both the control and skin groups, as measured by the absolute cell count (cells/μL) and the percentage of lymphocytes. Permutation testing revealed that the T cell subsets exhibiting the most significant up-regulation of α4β7 integrin expression in the gut group were the memory T cells (CD45⁺, CD3⁺, and CD45RO⁺) in both the CD4⁺ and CD8⁺ cell populations. The median expression level of absolute memory CD4⁺ α4β7⁺ T cells was 2.81 cells/μL in the gut group, 0.28 cell/μL in the control group (P = .001), and 0.13 cell/μL in the skin group (P < .0001). When analyzed according to percent lymphocytes, in the gut group a median of 0.67% of lymphocytes displayed the memory CD4⁺ α4β7⁺ phenotype, compared with only 0.033% in the control group (P < .0001) and 0.03% in the skin group (P < .0001). For CD8⁺ T cells, the median expression level of absolute memory CD8⁺ α4β7⁺ T cells was 1.63 cells/μL in the gut group, 0.04 cell/μL in the control (P < .0001), and 0.021 cell/μL in the skin group (P < .0001). Analysis by relative expression revealed that in the gut group, a median of 0.53% of lymphocytes displayed the memory CD8⁺ α4β7⁺ phenotype, compared with only 0.004% in the control group (P < .0001) and 0.009% in the skin group (P < .0001). Figure 1 illustrates the results of the analyses of α4β7 integrin expression on CD4⁺ and CD8⁺ memory T cells, and Figure 2 shows representative histograms. Because there was heterogeneity in the timing of sample collection, we investigated whether time after HSCT or time before GVHD symptoms affected the expression of α4β7 integrin in the memory subsets. As shown in Table 3, there was no significant effect of the timing of sample collection on the expression of α4β7 integrin in either the gut group or the skin group.

Table 2.

Surface Expression of α4β7 Integrin and CCR9 on Various Peripheral T Cell Subsets, Median (Range)

				P value
Phenotype	Control (n = 20)	Skin (n = 20)	Gut (n = 19)	C vs S*	C vs G*	S vs G*	C vs S†	C vs G†	S vs G†
Abs CD4 memory α4β7	0.28 (0-3.1)	0.13 (0-1.67)	2.81 (0.13-48.4)	.11	<.0001	<.0001	.88	.001	<.0001
Abs CD8 memory α4β7	0.040 (0-3.3)	0.021 (0-5.25)	1.63 (0-57.0)	.51	<.0001	<.0001	1.00	<.0001	<.0001
% CD4 memory α4β7	0.033 (0-0.45)	0.030 (0-0.25)	0.67 (0.02-3.31)	.99	<.0001	<.0001	1.00	<.0001	<.0001
% CD8 memory α4β7	0.004 (0-0.13)	0.009 (0-0.49)	0.53 (0-13.7)	.69	<.0001	<.0001	1.00	<.0001	<.0001
Abs CD4 naïve α4β7	0.028 (0-0.46)	0.021 (0-0.29)	0.13 (0-8.33)	.37	.0014	<.0001	1.00	.01	.008
Abs CD8 naïve α4β7	0.069 (0-4.35)	0.017 (0-4.51)	0.78 (0-12.8)	.13	<.0001	<.0001	1.00	.01	<.0001
% CD4 naïve α4β7	0.005 (0-0.044)	0.004 (0-0.086)	0.03 (0-1.10)	.95	<.0001	.001	1.00	.002	.05
% CD8 naïve α4α7	0.010 (0-0.092)	0.005 (0-0.42)	0.30 (0-4.56)	.60	<.0001	<.0001	1.00	<.0001	<.0001
Abs CD4 memory CCR9	0.30 (0.05-10.2)	0.22 (0-23.2)	2.23 (0.03-38.8)	.70	.08	.11	1.00	.57	.70
Abs CD8 memory CCR9	0.17 (0.007-6.30)	0.23 (0.009-1.56)	0.42 (0-38.9)	.35	.06	.13	1.00	.34	.61
% CD4 memory CCR9	0.04 (0.004-1.10)	0.087 (0-11.2)	0.30 (0.007-9.14)	.29	.050	.67	.75	.17	1.00
% CD8 memory CCR9	0.017 (0.002-0.33)	0.056 (0.008-0.40)	0.17 (0-2.70)	.007	.032	.52	.51	.006	.78
Abs CD4 naïve CCR9	0.12 (0-2.47)	0.065 (0-14.1)	0.31 (0-1.89)	.92	.30	.81	1.00	1.00	1.00
Abs CD8 naïve CCR9	0.12 (0-8.82)	0.054 (0-1.01)	0.15 (0-15.7)	.07	.56	.04	.94	1.00	.60
% CD4 naïve CCR9	0.009 (0-0.43)	0.007 (0-4.65)	0.026 (0-3.53)	.38	.21	.86	.58	1.00	1.00
% CD8 naïve CCR9	0.02 (0-0.21)	0.017 (0-0.10)	0.032 (0-1.91)	.97	.44	.46	1.00	.71	.52

Abs indicates absolute number of cells/μL; %, percent lymphocytes; C, control; S, skin; G, gut.

Note. Significantly different comparisons are in bold.

^*From the exact WRS test.

^†From the permutation test.

Figure 1.

Expression of α4β7 integrin on memory CD4⁺ and CD8⁺ T cells. A, Expression of α4β7 integrin (cells/μL) on memory CD4⁺ T cells. B, Expression of α4β7 integrin (% lymphocytes) on memory CD4⁺ T cells. C, Expression of α4β7 integrin (cells/μL) on memory CD8⁺ T cells. D, Expression of α4β7 integrin (% lymphocytes) on memory CD8⁺ T cells. All values shown are median values. P values from comparisons between groups derived from permutation testing are listed in the text and in Table 2.

Figure 2.

Expression of α4β7 integrin on peripheral blood T cell subsets in patients from the control, skin, and gut groups. Thawed mononuclear cells were analyzed for cell surface expression of CD45RO (y-axis) and α4β7 integrin (x-axis). Plots are gated on CD3⁺ lymphocytes based on forward and side scatter and CD3 staining. The numbers indicate the percentage of CD3⁺ lymphocytes for each quadrant. These values are representative patients from within each patient group.

Table 3.

Correlation between Expression of α4β7 Integrin on Activated T Cell Subsets with Time from Transplantation or Time before Onset of aGVHD Symptoms in Patients in the Gut and Skin Groups

	Gut		Skin
Potential Correlation	r	P	r	P
Expression of α4β7 on memory CD4+ T cells and time before onset of aGVHD	0.22	.37	0.29	.22
Expression of α4β7 on memory CD8+ T cells and time before onset of aGVHD	0.3	.21	0.27	.24
Expression of α4β7 on memory CD4+ T cells and time from HSCT to sample collection	0.11	.64	0.3	.2
Expression of α4β7 on memory CD8+ T cells and time from HSCT to sample collection	0.23	.34	0.33	.15

aGVHD indicates acute graft-verus-host disease.

Using the recursive partitioning method and ROC, we proposed cutoff values of 1 cell/μL and 0.5 cell/μL and applied them to the subsets of memory CD4⁺ α4β7⁺ T cells and CD8⁺ α4β7⁺ T cells, respectively (Table 4). Using these cutoff values, 79% of patients in the gut group had ≥ 1 cell/μL with the absolute memory CD4⁺ α4β7⁺ phenotype, compared with 10% in both the control and skin groups (P = .00002), indicating 79% sensitivity and 90% specificity. Similarly, 74% of patients in the gut group had ≥ 0.5 cell/μL with the absolute memory CD8⁺ α4β7⁺ phenotype, compared with 5% in the control group (P = .00001) and 10% in the skin group (P = .00007), indicating 74% sensitivity and a 95% specificity.

Table 4.

Proposed Cutoff Values for Memory CD4⁺ α4β7⁺ and Memory CD8⁺ α4β7⁺ T Cells

	ORGAN
	CONTROL	GUT	SKIN	Cvs S	C vs G	S vs G
CD4+ (cells/ul)
<1	18 (90%)	4 (21%)	18 (90%)	1	.00002	.00002
>=1	2 (10%)	15 (79%)	2 (10%)
CD8+ (cells/ul)
<0.5	19 (95)	5 (26%)	18 (90)	1	.00001	.00007
>=0.5	1 (5%)	14 (74%)	2 (10%)

C indicates control; S, skin; G, gut.

Note. Using the recursive partitioning method and ROC, cutoff values of 1 cell/μL and 0.5 cell/μL were applied to the subsets of memory CD4⁺ α4β7⁺ T cells and memory CD8⁺ α4β7⁺ T cells, respectively. P values are 2-way comparisons.

In contrast to the significant differences observed in α4β7 integrin expression among the 3 groups, the only T cell subset that exhibited a significant difference in CCR9 expression for the gut group relative to both the skin and control groups was the memory CD8⁺ population when measured as percent lymphocytes. Surface expression of both CLA and CCR10 also were analyzed in these T cell populations, because both are considered specific mediators of lymphocyte trafficking in the skin. No significant differences in their expression was found in the 3 patient groups or in any of the peripheral T cell populations studied (data not shown).

We then performed a multivariable logistic regression analysis for each T cell subset to explore whether surface expression of α4β7 integrin and CCR9 were confounded by clinical characteristics, such as patient age, donor type, conditioning regimen, recipient–donor sex mismatch, and GVHD prophylaxis regimen (sirolimus-containing or not). After adjusting for these clinical characteristics, the absolute number of memory CD4⁺ and CD8⁺ α4β7⁺, percent memory CD4⁺ and CD8⁺ α4β7⁺, absolute naïve CD4⁺ α4β7⁺, percent naïve CD4⁺ and CD8⁺ α4β7⁺, and percent memory CD8⁺ CCR9⁺ T cells in the gut group remained significantly associated with development of intestinal aGVHD (P < .05) (Table 5). The absolute number of naïve CD8⁺ α4β7⁺ T cells was borderline significant (P = .06). This finding was consistent with the results of univariate analysis. In contrast, none of these phenotypes were predictive for cutaneous aGVHD (Table 5).

Table 5.

Multivariable Logistic Regression Analysis for Each T Cell Subset

	Gut		Skin
	OR (95% CI)	P	OR (95% CI)	P
Abs CD4 memory α4β7	3.37 (1.56-7.28)	.002	0.73 (0.46-1.15)	.17
Abs CD8 memory α4β7	1.91 (1.24-2.94)	.004	0.77 (0.5-1.18)	.23
% CD4 memory α4β7	3.96 (1.84-8.55)	.0004	0.98 (0.54-1.76)	.93
% CD8 memory α4β7	2.91 (1.52-5.58)	.001	0.94 (0.54-1.63)	.83
Abs CD4 naive α4β7	1.97 (1.17-3.34)	.01	0.97 (0.6-1.55)	.89
Abs CD8 naïve α4β7	1.51 (0.99-2.29)	.06	0.70 (0.45-1.09)	.12
% CD4 naïve α4β7	4.08 (1.67-9.95)	.002	1.93 (0.89-4.2)	.10
% CD8 naïve α4β7	2.2 (1.25-3.89)	.007	0.91 (0.53-1.58)	.75
Abs CD4 memory CCR9	1.39 (0.96-2.02)	.08	1.09 (0.76-1.55)	.64
Abs CD8 memory CCR9	1.50 (0.97-2.34)	.07	1.07 (0.68-1.68)	.76
% CD4 memory CCR9	1.64 (1.1-2.47)	.02	1.47 (0.99-2.18)	.05
% CD8 memory CCR9	2.49 (1.35-4.59)	.003	1.63 (0.93-2.86)	.09
Abs CD4 naïve CCR9	1.06 (0.77-1.47)	.71	1.07 (0.78-1.47)	.66
Abs CD8 naïve CCR9	1.01 (0.66-1.55)	.97	0.72 (0.46-1.12)	.15
% CD4 naïve CCR9	1.17 (0.78-1.74)	.44	1.36 (0.93-1.98)	.11
% CD8 naive CCR9	1.53 (0.8-2.92)	.19	0.82 (0.43-1.57)	.55

OR indicates odds ratio with the control group as the reference group; CI, confidence interval; Abs, absolute number of cells/μL; %, percent lymphocytes.

Note. Clinical characteristics adjusted for patient age, donor type, conditioning regimen, GVHD prophylaxis (sirolimus-containing or not), and recipient–donor sex mismatch.

Spearman Correlation

We carried out Spearman correlation analysis to investigate a possible association between the naïve and memory CD4⁺ and CD8⁺ phenotypes for α4β7 expression in patients with intestinal aGVHD. As shown in Table 6, a strong correlation was found between the CD4⁺ and CD8⁺ memory and naïve cells that expressed α4β7 integrin. Indeed, the association between memory CD4⁺ and CD8⁺ T cells was quite close (correlation coefficient, r = 0.85; P < .0001). Similar findings were found when this analysis was carried out with measurements of percent lymphocytes (r = 0.81; P < .0001) (data not shown).

Table 6.

Spearman Correlation for Measurements of Absolute Number of Cells per μL Expressing α4β7 Integrin in Memory and Naïve CD4⁺ and CD8⁺ T Cell Populations

	Memory CD4+	Memory CD8+	Naïve CD4+	Naïve CD8+
Memory CD4+	– –	ρ = 0.85	ρ = 0.70	ρ = 0.81
		P < .0001	P = .001	P < .0001
Memory CD8+	ρ = 0.85	– –	ρ = 0.54	ρ = 0.78
	P < .0001		P = .018	P < .0001
Naïve CD4+	ρ = 0.70	ρ = 0.54	– –	ρ = 0.59
	P = .001	P = .018		P = .008
Naïve CD8+	ρ = 0.81	ρ = 0.78	ρ = 0.59	– –
	P < .0001	P < .0001	P = .008

Influence of Clinical Factors on α4β7 Integrin and CCR9 Expression

We examined memory CD4⁺ and CD8⁺ T cell surface expression of α4β7 integrin in relation to the intensity of the conditioning regimen (MA vs RIC), donor type (MRD vs MUD), GVHD prophylaxis regimen (sirolimus-containing vs not), and recipient age (<50 years vs ≥ 50 years). The results for the control and skin groups were pooled and compared with the results for the gut group. For recipient age, in the gut group, the median expression level of absolute CD4⁺ memory α4β7 was 3.77 for age < 50 years and 1.7 for age ≥ 50 years (P = .90 using permutation testing). In the control and skin combined group, the median level was 0.3 for age < 50 years and 0.09 for age ≥ 50 years (P = .88). Similarly, intensity of conditioning regimen, donor type, and GVHD prophylaxis regimen showed no influence on memory T cell surface expression of α4β7 integrin and CCR9.

We performed linear regression analysis to further examine the differences in expression level between the gut group and the other groups in the presence of patient age, conditioning regimen, donor type, GVHD prophylaxis regimen, and recipient–donor sex mismatch. The development of intestinal aGVHD continued to be the most significant factor affecting expression of α4β7 integrin in both memory CD8⁺ and CD4⁺ T cells (P < .0001) (data not shown).

We reviewed patient records from the gut group for any evidence of common infections of the gastrointestinal tract after HSCT, specifically adenovirus, cytomegalovirus (CMV), and Clostridium difficile (Cdiff). No patients in the gut group exhibited evidence of adenovirus or direct intestinal infection with CMV. Three patients were found to have blood CMV reactivation, as detected by blood antigen or polymerase chain reaction–based assays within 1 month after onset of aGVHD symptoms, none of which progressed to intestinal disease. Two patients were found to have evidence of Cdiff infection, as detected by toxin assays, within 1 month after symptom onset. Neither CMV nor Cdiff infection was found to have an effect on α4β7 integrin expression (data not shown).

Expression of α4β7 Integrin in Intestinal Biopsy Specimens from Patients with aGVHD

Sigmoid biopsy specimens were obtained from 17 of the 19 patients in the gut group at the time of the initial diagnosis of aGVHD. All specimens exhibited pathological findings consistent with aGVHD. Sufficient paraffin-embedded samples for all 17 patients were available for immunohistochemical staining. All 17 biopsy specimens showed areas of α4β7⁺ mononu-clear cells infiltrating within the lamina propria in areas of active aGVHD; however, the staining was heterogeneous and patchy in distribution. Representative areas of infiltration are shown in Figure 3A and B. Interestingly, the highest concentration of α4β7 integrin–expressing cells was found as aggregates within small mucosal blood vessels (Figure 3C-E). Intestinal biopsy specimens from patients in both the skin and control groups had not been performed, given the lack of GI symptoms in these patients. We found several intestinal biopsy specimens obtained from patients after HSCT who did not have aGVHD in our pathology archives, and performed immunohistochemical staining on them. Mucosal blood vessels in these biopsy specimens did not exhibit the same aggregation of α4β7⁺ mononuclear cells found in specimens from patients in the gut group (Figure 3F and G).

Figure 3.

Expression of α4β7 integrin in sigmoid biopsy specimens from patients with intestinal aGVHD. Immunohistochemical staining for α4β7 integrin (brown stain) shows patchy expression of α4β7 integrin on mononuclear cells within the lamina propria (A and B) and as aggregates within small mucosal blood vessels (C, D, and E). Specimens from patients after HSCTwho did not have intestinal aGVHD do not show such staining (F and G). See the Immunohistochemistry section of the text for complete descriptions of techniques and image acquisition information.

DISCUSSION

Recently, several murine models have illustrated that lymphocyte migration after HSCT involves sequential trafficking of naïve donor T cells to host secondary lymphoid tissue, followed by specific infiltration into target organs to cause aGVHD. Data from these models suggest that the initial afferent phase of this response occurs within the first few hours to days after allogeneic HSCT and well before clinical manifestations of aGVHD appear [29,30]. Therefore, defining the specific adhesion molecules involved in the lymphocyte trafficking of either the afferent or efferent phase of aGVHD may provide novel opportunities for the monitoring, prophylaxis, and treatment of aGVHD after HSCT; for instance, reliable prediction of the development of intestinal aGVHD may allow institution of specific preemptive therapy or a different schedule of immunosuppression withdrawal.

In this retrospective case-controlled study, we analyzed the surface expression of both α4β7 integrin and CCR9 on various T cell subsets from 3 groups of patients after HSCT (no aGVHD, skin-predominant aGVHD, and gut-predominant aGVHD) from blood samples obtained before the onset of aGVHD symptoms. Our primary aim was to determine whether up-regulation of α4β7 integrin and CCR9 on peripheral blood T cells is correlated with the subsequent development of intestinal aGVHD. Our results show that α4β7 integrin was significantly up-regulated on both memory and naïve T cells, most strikingly in the memory subsets in patients who went on to develop intestinal aGVHD. Indeed, expression on CD4⁺ and CD8⁺ memory T cell subsets was at least 10-fold higher in patients with intestinal aGVHD than in those with cutaneous or no aGVHD. In contrast, CCR9 expression was less useful, not clearly different between the 3 groups. A recent study in mice suggested that CCR9 expression is a marker of immature dendritic cells, which seem to induce tolerance and to protect against GVHD implying that the overall role of CCR9 remains unclear [31]. We also analyzed CLA and CCR10 expression in the present study, because these are considered specific mediators of lymphocyte trafficking to the skin, but we found no significant differences in the expression of these molecules in the 3 patient groups.

After correction for multiple comparisons, α4β7 integrin expression remained significantly increased on memory T cells in both the CD4⁺ and CD8⁺ subsets from patients in the gut group. In adjusted multivariable logistic regression analysis, these phenotypes were predictive for the development of intestinal aGVHD using samples obtained a median of 14 days before the onset of aGVHD symptoms. Analysis of the correlation between the subsets demonstrated that expression of α4β7 integrin on memory CD4⁺ and CD8⁺ T cells had a strong association with the other phenotypes, suggesting that the CD4⁺ or CD8⁺ CD45RO⁺ α4β7⁺ phenotype could be considered for evaluation as a potential biomarker for intestinal aGVHD in future investigations. Based on our results, we propose a cutoff value of 1 cell/μL for absolute memory CD4⁺ α4β7⁺ T cells and 0.5 cell/μL for absolute memory CD8⁺ α4β7⁺ T cells, to maximize specificity.

Given that this was a retrospective analysis, there was significant heterogeneity in terms of patient characteristics and the timing of sample collection. But even with such variation in donor source, conditioning regimen, recipient age, and GVHD prophylaxis regimen, a significant up-regulation of α4β7 integrin expression on activated T cells was observed in patients who later developed intestinal aGVHD. This finding was not unexpected, because the trafficking of specific immunologic responses should continue to be driven by the same mediators regardless of such clinical characteristics. Although the timing of the sample collection (a median of 14 days before the onset of aGVHD symptoms; most within 30 days) may seem relatively remote from the onset of clinical aGVHD, it is consistent with the previously described murine models showing an early initial activation phase with trafficking to secondary lymphoid tissue, followed by delayed clinical manifestations [29,30]. Clearly, the true association and predictive value of measuring α4β7 integrin expression can be defined only in a prospective trial involving serial samples from patients after HSCT; such a study is underway at our institution. Another possibility is that before HSCT, stem cell donors have naturally higher levels of T cell expression of α4β7 integrin, which may predispose recipients to develop intestinal GVHD. No pre-HSCT donor samples were available for this cohort in the present study, although such a study also is underway at our institution.

The histological findings of intestinal aGVHD range from mild to severe. Mild cases show spotty apoptosis with focal infiltrates of lymphocytes, plasma cells, or eosinophils, whereas severe cases show diffuse crypt cell necrosis [32,33]. Mediators of intestinal lymphocyte trafficking have never been analyzed in pathological specimens of human patients with aGVHD. In this study, all evaluable rectal biopsy specimens consistent with intestinal GVHD exhibited evidence of α4β7⁺ mononuclear cells located within small terminal mucosal blood vessels, which may represent an early phase of lymphocyte trafficking. These cells were not observed in intestinal tissue from healthy patients or from biopsy specimens obtained from patients after HSCT who did not have intestinal aGVHD. These data support the idea that the increased expression of α4β7 integrin on peripheral blood T cells mirrors what is occurring in the intestine, the principle target tissue.

To date, very few human studies have investigated lymphocyte trafficking molecules associated with aGVHD. One such study found that patients with aGVHD had increased expression of certain adhesion molecules, such as ICAM-1 and VLA-4, on their peripheral blood lymphocytes compared with controls who had undergone HSCT without developing aGVHD [34]. Three recent studies have analyzed specific chemokines and their receptors in the setting of aGVHD. One of these studies suggested a correlation between higher levels of CCR7⁺ CD4⁺ T cells in donor grafts and the incidence and severity of aGVHD [35]. Another study that used PCR to analyze chemokine receptor expression on peripheral blood lymphocytes found a correlation between up-regulation of CCR5, CXCR3, CCR1, and CCR2 with episodes of aGVHD [36]. Most recently, a prospective analysis of 34 patients after HSCT showed a 2-fold up-regulation in serum levels of chemokine CXCL10 in patients with aGVHD relative to control patients. As part of their analysis, the investigators also analyzed CLA and β7 integrin expression among activated CD4⁺ and CD8⁺ T cell subsets and found no significant up-regulation of these molecules in patients with aGVHD [37]. But that study included only 34 patients and, unlike our study, did not stratify patients based on the specific organ of involvement.

Although to date no human studies have investigated specific mediators of intestinal lymphocyte trafficking in aGVHD of the intestine, several experiments have been performed in mouse models, with mixed conclusions. Murai et al. [38] reported that Peyer's patch–deficient host mice did not develop aGVHD after bone marrow transplantation, and that blocking access to Peyer's patch via a monoclonal antibody (mAb) against MAdCAM-1 protected mice from aGVHD. In separate studies, investigators from Memorial Sloan Kettering found that purging donor cells of α4β7⁺ T cells or using a selective β7 integrin knockout strain as the source of donor stem cells resulted in significant reductions in the overall incidence, clinical and histological manifestations, and mortality from intestinal aGVHD [39,40]. In contrast, another study suggested that knockout of β7 was insufficient to modify intestinal aGVHD, and that additional knockout of L-selectin was required to bring about a significant reduction in aGVHD [41]. Beilhack et al. [8] performed several experiments in murine models in which entry to different secondary lymphoid tissues was restricted by the use of mAb directed against specific mediators of lymphocyte trafficking. They found that blocking entry to specific secondary lymphoid organs (either peripheral lymph nodes or Peyer's patches) had no effect on aGVHD, suggesting redundancy in secondary lymphoid tissue compartments [8]. Most recently, Anderson et al. [42] investigated the importance of secondary lymphoid tissue in aGVHD initiation using mice deficient in either CD62L or CCR7 or using peripheral lymph node–and Peyer's patch–deficient mice as recipients, and found that clinically significant GVHD developed in both of these settings.

Drawing general conclusions from these individual studies is difficult, because they used different mouse donor–recipient combinations and different models of HSCT. Overall, the results support the theory that aGVHD is initiated by allogeneic activation in secondary lymphoid tissue, with induction of a specific trafficking phenotype dependent on the location of the secondary lymphoid tissue [8]. Whether blocking access to certain secondary lymphoid tissues has a clinically significant effect on aGVHD is unclear, however. In the murine studies that showed a benefit of disruption of the α4β7–MAdCAM-1 interaction, the incidence of intestinal aGVHD was not dramatically reduced; rather, aGVHD seemed to be delayed and decreased in severity, which translated into less morbidity and mortality. These benefits need to be interpreted in the proper context, given that antibiotics, immunosuppressive therapy, and other standard supportive care measures of clinical care after HSCT are not usually included in mouse experiments. Nevertheless, inhibiting early trafficking to GALT may allow time for the intestine to heal after injury from the conditioning regimen, which may result in decreased inflammatory cytokine and chemokine production, leading to a moderation of the allorecognition reaction when it does occur [43,44].

In summary, our data demonstrate that up-regulation of α4β7 integrin on the surface of activated T cells appears to be correlated with the subsequent development of intestinal aGVHD in patients after HSCT. These data need to be validated in a prospective study, which is ongoing at our institution. This prospective study also will allow us to better analyze the effect of infections and other clinical factors, as well as to include patients who develop lower grades of intestinal GVHD. Furthermore, collection of samples at standard prospective times both before and after the onset of aGVHD will indicate whether a specific time after HSCT or interval before the onset of symptoms is important and demonstrate whether expression level correlates with treatment response. Most importantly, a prospective analysis will confirm whether expression of α4β7 integrin on memory T cells can be a predictive biomarker for intestinal aGVHD. Having a predictive biomarker for intestinal aGVHD should prove very valuable in clinical practice, aiding the delivery of preemptive targeted therapy before the onset of clinical manifestations or refinement of the schedule of immunosuppression withdrawal. In addition, identification of the adhesion molecules involved in intestinal aGVHD can provide novel targets for more specific therapy. Such therapies are currently under study for the treatment of inflammatory bowel disease [45], and these studies support the further evaluation of these agents in intestinal aGVHD.