Tumor-Infiltrating Lymphocytes Expressing the Tissue Resident Memory Marker CD103 Are Associated with Increased Survival in High-Grade Serous Ovarian Cancer

Epithelial ovarian cancer (EOC) is the fifth most frequent cause of death due to cancer in women worldwide (1). The overall 5-year survival rate is 44% (2) and depends on stage at diagnosis, histologic subtype, and extent of surgical debulking. In addition, a number of studies have demonstrated that the presence of tumor-infiltrating lymphocytes (TIL) in surgically resected tumor tissue correlates with a favorable prognosis, particularly in patients with high-grade serous ovarian cancer (HGSC). For example, in a landmark study by Coukos and colleagues, retrospective analysis of 186 stage III–IV EOC patients revealed that the 5-year overall survival rate was 38% for patients with CD3⁺ tumor-infiltrating lymphocyte (TIL)-positive tumors, compared with only 4.5% for CD3⁺ TIL-negative tumors (3). Follow-up studies revealed that the CD8⁺ subset of TILs correlates most strongly with survival (4). Moreover, although TILs are found in all EOC subtypes, their beneficial effects are most pronounced in HGSC (5). Importantly, the prognostic effect is strongest for CD8⁺ TILs that are localized within the epithelial component of tumors rather than the associated stroma (5). Despite this association, the mechanisms that promote the localization of TILs to tumor epithelium remain poorly understood.

Retention of T cells within epithelial tissues under a variety of normal and pathologic conditions is known to be mediated, at least in part, through expression of the α_E(CD103)/β₇ integrin molecule (referred to hereafter as CD103; ref. 6). In healthy individuals, CD103 is expressed on less than 2% of circulating peripheral blood lymphocytes but widely expressed on intraepithelial lymphocytes (IEL) of the gut mucosa and skin (7, 8). The only known ligand for CD103 is the epithelial cell surface molecule E-cadherin and adhesive interactions between CD103 and E-cadherin are thought to be responsible for retention of antigen-specific lymphocytes within epithelial tissues (6). Interestingly, CD103 is also expressed by tissue-infiltrating, allo-reactive CD8⁺ T cells during transplant rejection (9) and graft-versus-host disease (10). In the infectious disease setting, CD103 is increasingly being recognized as a definitive marker of “tissue resident memory” CD8⁺ T cells (CD8⁺ T_RM; ref. 11). Virus-specific CD8⁺ T_RM have been found in tissues such as skin (12), sensory ganglia (13), brain (14), lung (15), and tonsils (16). CD8⁺ T_RM are stably retained at the site of primary infection for long periods of time after disease resolution, even in the complete absence of persisting antigen (17) and are thought to play a pivotal role in conferring immunity against re-exposure to the same pathogen (13, 18).

Recent evidence suggests that CD103 also plays an important role in immunity against cancer, and in particular, cancers of epithelial origin (19–23). For example, CD103:E-cadherin interactions have been implicated in the recognition of tumor cells by cytotoxic T cells in colon (19), lung (20), and pancreatic cancer (21). We recently reported that the ascites compartment of some patients with HGSC contains large numbers of CD8⁺CD103⁺ T cells with an effector memory phenotype (24). CD103 is reported to be recruited to the immunologic synapse during tumor cell recognition, where it mediates adhesive interactions with E-cadherin on the target tumor cell, resulting in polarization of lytic granules and potentiated cytotoxic activity (20). In a lung cancer xenograft model, CD103 surface expression was upregulated on TILs after migration to the tumor and coengagement of TCR and TGF-β receptors at that site (25). Together, these findings lead us to hypothesize that CD103 expression might demarcate intraepithelial, tumor-reactive TILs that confer favorable prognosis in epithelial cancers.

Using a new anti-CD103 antibody that works with formalin-fixed, paraffin-embedded tissues (26), we have performed the first analysis of the intratumoral location and prognostic influence of CD103⁺ TILs in human cancer. We show herein that CD103 is widely expressed by TILs in human ovarian tumors and that it is preferentially expressed by intraepithelial versus stromal CD8⁺ TILs in both ovarian and breast tumors. In some patients, CD103 is also expressed by tumor-infiltrating NK cells. Importantly, CD103⁺ TILs are strongly associated with survival in patients with HGSC, indicating that CD103 demarcates a clinically beneficial TIL subset that warrants further analysis as a potential effector cell population for immunotherapy.

All specimens and clinical data were obtained with either informed written consent or a formal waiver of consent under protocols approved by the Research Ethics Board of the BC Cancer Agency (BCCA) and the University of British Columbia (Vancouver, British Columbia, Canada). The prospective cohort was comprised of HGSC patients admitted to the BCCA between August 2007 and August 2011. Patients were treated with primary cytoreductive surgery followed by chemotherapy with carboplatin with or without paclitaxel. Tumor tissue obtained during primary cytoreductive surgery was fixed in formalin and embedded in paraffin using standard methodologies. A tissue microarray was constructed with duplicate 1.0 mm cores from representative tumor areas using a manual tissue arrayer (Beecher Instruments). Immunohistochemical analyses were performed on either whole tumor sections or on tissue microarrays, as indicated.

The retrospective cohort used for survival analysis has been previously described (5). In brief, tumor samples were obtained from EOC cases of various histologic subtypes seen at the BC Cancer Agency from 1984 to 2000 (OvCaRe Ovarian Tumour Bank). All tumor tissues were obtained at the time of primary surgery and patients had no macroscopic residual disease following surgical cytoreduction. Table 1 describes this cohort in accordance with REMARK criteria. The breast tumor microarray utilized has also been previously described (27) and contained 24 primary breast tumors with no specific selection criteria.

To detect CD103⁺ cells, 4 μm sections of either whole tumor tissue or tissue microarrays were deparaffinized and subjected to heat-induced epitope retrieval using a decloaking chamber and Diva decloaking reagent (Biocare Medical). Anti-CD103 rabbit monoclonal antibody (Epitomics clone EPR4166(2)) was applied at a dilution of 1/1,000 for 30 minutes at room temperature. Following washing, anti-rabbit-HRP polymer reagent was applied for 30 minutes at room temperature followed by 3,3′-diaminobenzidine (DAB) reagent. The slides were counterstained with hematoxylin, dehydrated, and cover-slipped using Cytoseal 60.

For two-color immunohistochemical staining, sections were stained with rabbit anti-CD103 as described above in combination with a murine primary antibody against either CD8 (C8/144B) or CD56 (123C3.D5). Slides were incubated with Mach 2 Double Stain #1 polymer reagent (Biocare) for 30 minutes at room temperature, and the horseradish peroxidase (HRP), and alkaline phospatase (AP) enzyme tags were detected using Vina Green and Warp Red chromogens (Biocare), respectively. Slides were counterstained with hematoxylin.

For the CD103/CD8/cytokeratin triple stain, rabbit anti-CD103 and murine anti-CD8 antibodies were added for 30 minutes at room temperature followed by Mach 2 Double Stain 2 polymer for 30 minutes. CD103 was detected with DAB chromogen and CD8 was detected with Warp Red. Slides were subjected to a denaturation protocol followed by a second round of staining with pan-Cytokeratin antibody (clone PCK-26, Sigma-Aldrich). Slides were incubated with Mach 2 Mouse-AP polymer followed by Perma-Green Plus AP chromogen (Diagnostic Biosystems).

Slides were analyzed using the Nuance Multispectral Imaging System (CRi) with a BX-53 microscope (Olympus). Single-stained slides were used to create spectral profiles for each chromogen and these profiles were then used to discriminate the individual colours on multistained slides. False color images were generated using Metamorph software (Molecular Devices).

Single-cell suspensions were prepared from freshly excised tumor specimens and cryopreserved, as described (28). After thawing, cells were stained with the following fluorescently conjugated antibodies: CD3-BV510, PD-1-PECy7 (Biolegend), CD103-PECy7 (Ebioscience), CD8-BV421, CD103-PE, HLA-DR-PE, CD127-AF647, CD56-PE/APC, CD4 FITC, and Ki67-FITC (BD Biosciences). All flow cytometry was performed using a FACS Calibur that had been upgraded to 8-color capability by Cytek (DxP8). Analyses were performed using FlowJo software (Tree Star).

Scoring for epithelial localization of CD8 and CD103 markers was undertaken by an experienced histopathologist (P.H. Watson) blinded to results of other markers or case identity. Positively stained cells that had morphologic features compatible with lymphocytes were counted according to two localization patterns: (i) intraepithelial infiltrates, defined as TILs within cancer cell nests or in direct contact with cancer cells; and (ii) stromal infiltrates, defined as TILs within stromal areas that lack direct contact with epithelial cells. For prognostic analyses, absolute numbers of CD103⁺ TILs (independent of location) in cores of the retrospective cohort TMA were enumerated by visual inspection. Duplicate cores were counted independently by two different investigators and the average cell count per core was used for statistical analyses. Statistical analyses and HR calculations were performed using GraphPad software. Univariate survival analysis was performed using the Kaplan–Meier method and P values were determined using log-rank test. Multivariate survival analysis was performed using a Cox regression model. P values <0.05 were regarded as significant.

To assess the frequency and distribution of CD103⁺ TILs in primary ovarian tumor samples, we used a recently developed anti-CD103 antibody (26) to stain a previously described tissue microarray (5) containing cores from a cohort of 497 EOC solid tumor specimens. CD103⁺ TILs were detectable in tumors of all stages and grades and in all four major EOC subtypes (Fig. 1; Table 1). In all tumor subtypes, CD103⁺ TILs seemed to be preferentially localized to the epithelial regions of tumors (Fig. 1A). The density of CD103⁺ TILs within each subtype ranged from negligible to heavy infiltration (Fig. 1B) and the highest mean frequency of CD103⁺ TILs was seen in serous, followed by the endometrioid, clear cell, and mucinous subtypes, respectively (Fig. 1B; Table 2).

Focusing on the HGSC (grades 2 and 3), we then quantified the intratumoral location of CD103⁺ TILs as epithelial or stromal. Serial sections of a prospective HGSC TMA were stained with hematoxylin and eosin (H&E), anti-CD8 or anti-CD103 and localization was evaluated by an experienced pathologist. There was a statistically significant bias in the localization of CD103⁺ versus CD8⁺ TILs to tumor epithelium with an average epithelial:stromal density ratio of 2.29 for CD103⁺ TILs versus 1.36 for CD8⁺ TILs (P = 0.008 by the Wilcoxon test, n = 34; Fig. 2A). The biased intraepithelial localization was particularly evident in those tumors with denser CD103⁺ TIL infiltration.

As a complementary approach, we performed triple staining of HGSC tumors to simultaneously detect CD8, CD103, and cytokeratin, the latter being a well-established marker of tumor epithelium. Consistent with the above results, CD8⁺CD103⁺ TILs were preferentially localized to cytokeratin-positive regions of the tumor (Fig. 2B). In addition, we evaluated the localization of CD103⁺ TILs in breast cancer, where epithelial and stromal regions are typically more distinct than in HGSC. CD103⁺ TILs were present in 5 of 20 invasive breast cancer specimens analyzed, and they again showed preferential localization to tumor epithelium (Fig. 2C). Thus, as seen with normal mucosal IEL, CD103 is preferentially expressed by “intraepithelial” TILs in ovarian and breast cancer.

The only known ligand for CD103 is the epithelial surface molecule E-cadherin (6). Consistent with prior reports (29), E-cadherin expression was variable across HGSC tumors, ranging from negative to strong expression (Fig. 3). However, there was no obvious correlation between E-cadherin staining intensity and the presence of CD103⁺ TILs (P = 0.667, Pearson test, n = 60). Thus, infiltration by CD103⁺ TILs is determined by factors other than E-cadherin expression.

To determine the lineage and activation status of CD103⁺ TILs, multiparameter flow cytometry was performed on disaggregated tumor preparations from 6 HGSC solid tumor specimens (Fig. 4A). In most patients, the majority of CD103⁺ TILs were CD3⁺CD8⁺ T lymphocytes. The second most common subset of CD103⁺ TILs was comprised of CD3⁻CD56⁺ NK cells. In some tumors, we also observed small numbers of CD3⁺CD4⁺CD8- and CD3⁺CD8-CD4- T cells among CD103⁺ TILs (Fig. 4A). CD103 staining was not detected on CD20⁺ B cells, FOXP3⁺ lymphocytes or CD14⁺ monocytes; CD11c⁺ DC were too rare to assess (data not shown). The presence of tumor-infiltrating CD103⁺CD8⁺ and CD103⁺CD56⁺ lymphocytes was confirmed by multiparameter IHC using antibodies against CD103, CD8, and CD56 (Fig. 4B). Consistent with the flow cytometry results, both CD8⁺ and CD56⁺ cells expressing CD103 were readily detectable by IHC and were localized primarily to tumor epithelium.

We also compared the activation and differentiation status of CD103⁻ and CD103⁺ TILs by flow cytometry. Figure 5 shows representative plots from tumors in which the CD103⁺ TIL population was predominantly CD8⁺ T cells (Case #014) or a mixture of CD8⁺ T cells and NK cells (Case #042). Consistent with the above observation that most CD103⁺ TILs were either CD8⁺ T cells or CD56⁺ NK cells, CD103⁺ TIL cells universally expressed TIA-1, a marker of cytolytic potential. Compared with their CD103- counterparts, CD103⁺CD8⁺ TILs had markedly increased expression of the activation markers HLA-DR, Ki-67, and PD-1 and lower levels of IL-7 receptor α chain (CD127).

We next assessed the relationship between CD103⁺ TILs and patient survival, using two distinct statistical approaches, both of which have previously been used to demonstrate the prognostic significance of TILs in EOC. The first approach used an absolute value (>5 CD103⁺ TILs per 0.6 mm core) to stratify the cohort into infiltrated versus noninfiltrated subsets, similar to as has been described previously (3, 4). Using this strategy, the presence of CD103⁺ TILs correlated with increased disease-specific survival in HGSC (P = 0.0003) and mucinous EOC (P = 0.0481), but not endometrioid or clear cell cancers (Table 2; Fig. 6A). The second approach relied upon the quantitative characteristics of the cohort to determine the cutoff value. As shown in Table 2, the distribution of CD103⁺ TIL infiltration was positively skewed (i.e., the mean was significantly higher than the median) for all 4 histologic subtypes. Similar patterns of positive skewing have been observed previously for other TIL subsets and has led to the use of the lowest tertile as an approach for stratifying cohorts into infiltrated versus noninfiltrated subsets (4). When the lowest tertile approach was used, the presence of CD103⁺ TILs again correlated with increased disease-specific survival in HGSC (P < 0.0001). However, using this strategy, CD103⁺ TIL infiltration did not reach statistical significance in mucinous, endometrioid, or clear cell EOC, likely due to the disproportionately low tertile values obtained in each of these subtypes (Table 2). CD103⁺ TIL infiltration remained a significant factor in multivariate survival analysis when age, stage, and grade were used as covariants (HR = 0.509; 95% confidence interval = 0.331–0.783; P = 0.002). Importantly, while previous studies have shown that only “intraepithelial” CD8⁺ TILs are associated with survival in HGSC, herein, TILs were scored solely on the basis of CD103 expression, which obviated the need to precisely account for their localization within the tumor.

Finally, we attempted to determine whether CD103⁺ CD8 cells or CD103⁺ NK cells were primarily responsible for the observed prognostic effect in HGSC. Although it is technically feasible to perform three-color IHC with antibodies to CD8, CD56, and CD103, the resulting images are difficult to score reliably on the scale of a large TMA. Therefore, as an alternate approach, we stained serial sections of the TMA with antibodies to CD8 and CD103 and then scored tumors as positive for CD8⁺ and/or CD103⁺ TILs using a threshold of more than 5 cells per core. The majority of tumors were positive for both CD103⁺ and CD8⁺ TILs and, accordingly, were associated with favorable prognosis (Fig. 6B). In contrast, tumors that contained CD8⁺ TILs that lacked CD103 expression were associated with inferior survival, similar to those tumors lacking CD8⁺ TILs altogether, implying that it is the CD103⁺CD8⁺ TIL cells that are primarily responsible for improved prognosis. Moreover, very few tumors contained CD103⁺ TILs without CD8⁺ TILs; therefore, we were unable to precisely infer the contribution of CD103⁺ NK cells to prognosis.

CD8⁺ TILs are strongly associated with patient survival in a broad range of epithelial cancers (30); however, the mechanisms that promote their localization to tumor epithelium remain poorly understood. We examined the hypothesis that intraepithelial CD8⁺ TILs, like normal mucosal IEL and T_RM, might express the α_E/β₇ integrin subunit CD103. Indeed, we found that CD103 expression is a common feature of intraepithelial CD8⁺ TILs in the four major subtypes of EOC. CD103⁺ TILs were found to have a highly activated, cytolytic phenotype, and included both CD8⁺ T cells and CD56⁺ NK cells. Consistent with their cytolytic phenotype, CD103⁺ TILs were strongly associated with survival in HGSC as well as mucinous EOC. Although CD103⁺CD8⁺ TILs have been reported in other epithelial cancers (19–23), to the best of our knowledge this is the first demonstration of their prognostic significance in any cancer. Our data indicates that CD103 demarcates a subset of TILs that seem to mediate protective antitumor immunity in HGSC and warrant further investigation as effector cells for immunotherapy.

In our analysis of 497 EOC tumors, the density of CD103⁺ TILs ranged widely, from 0 to more than 300 CD103⁺ TILs per tumor core. Although the underlying cause of this variability is unknown, earlier, in vitro studies have demonstrated that upregulation of CD103 on the surface of CD8⁺ T cells is dependent upon the concurrent stimulation of the TCR and TGF-β receptors (19, 31). Based upon this dual signal model, variability in CD103⁺ TIL density could therefore reflect inter-patient differences in the levels of TGF-β, cognate antigens, or both. Although TGF-β is widely considered an immunosuppressive factor, it can facilitate immune function in some physiologic contexts. For example, gut mucosal tissue contains resident IELs that upregulate CD103 expression in response to stimulation with cognate microbial Ag and locally expressed TGF-β (32). This culminates in the retention of IEL through adhesive interactions between α_E(CD103)/β₇ integrin and E-cadherin on gut epithelial cells (11). Accordingly, CD103 knockout mice show substantially reduced numbers of IEL in gut mucosa (33). Furthermore, allograft rejection involves infiltration of grafted tissues by CD8⁺ T cells, which express CD103 upon exposure to TGF-β in the local microenvironment (9). Importantly, allogeneic islet cell transplants show increased in vivo persistence in CD103-deficient mice due to impaired infiltration of T cells into graft tissue (34). Ovarian cancers have long been known to express TGF-β (35), and indeed TGF-β expression has been linked to metastatic potential (36) and to the process of epithelial to mesenchymal transition (37). Moreover, we previously reported that the malignant ascites of some patients with HGSC contains CD103-expressing CD8⁺ T cells, which correlate with the amount of TGF-β in the ascites fluid (24). Thus, the density of CD103⁺ TILs in solid tumors might reflect local TGF-β levels.

The fact that T cells also require antigen stimulation to upregulate expression of CD103 implies that CD103⁺ TILs may be responding to antigens in the tumor microenvironment. Although there is currently a paucity of well-characterized tumor antigens to test this hypothesis comprehensively in EOC, we previously showed that ascites-derived CD8⁺ T cells specific for the cancer-testis antigen NY-ESO-1 were predominantly CD103⁺ (24). The notion that CD103 is a marker of antigen-specific T cells is further supported by recent studies demonstrating CD103 expression by influenza-specific CD8⁺ T cells in human lung (15) and EBV-specific cells in human tonsil (16). CD103 is also expressed by virus-specific CD8⁺ cells in mouse brain after experimental infection with vesicular stomatitis virus (VSV; refs. 14, 18). In the above examples, CD103 demarcates tissue resident memory CD8⁺ T cells (CD8⁺ T_RM) that persist at sites of antigen encounter, even long after pathogen clearance (for review see refs. 11, 38, 39). Finally, in the present study, it is noteworthy that CD103⁺ TILs express high levels of PD-1 (Fig. 5), which has been shown to be marker of tumor-reactive TILs in melanoma (40).

At present, the antigens recognized by the vast majority of TILs in ovarian cancer remain poorly defined. Based upon the evidence described above, we speculate that CD103 may specifically demarcate TILs that are tumor-specific and, as such, may serve as a useful marker for enriching tumor-reactive TILs for antigen discovery efforts. On the other hand, given that CD103 is also a marker by T_RM, it is also conceivable that CD103⁺CD8⁺ TILs are specific for pathogen antigens rather than tumor antigens. This possibility is consistent with the recent hypothesis that many “ovarian” cancers actually originate in the fallopian tubes (41), which constitute a barrier tissue that is continuously exposed to various microbial agents. Indeed, similar to other mucosal tissues, the mucosal surface of the female genital tract is known to harbor intraepithelial T cells that express CD103 (42). Thus, it is plausible that at least some of the CD103⁺ TILs in EOC may represent pathogen-specific T_RM that are remnants of prior immune response(s) in the fallopian tubes or neighboring mucosal tissues. However, this alternative hypothesis needs to be reconciled with the strong prognostic effect of CD103⁺ TILs observed herein, which is indicative of tumor-specificity rather than pathogen specificity.

In addition to its potential utility as both a prognostic marker and a tool for antigen discovery, CD103 provides a convenient cell surface marker for investigating the functional phenotype of intraepithelial TILs. Previously, CD103⁺CD8⁺ T cells have been shown to have regulatory/immunosuppressive properties, at least in some settings (43). Moreover, CD103 is often expressed on conventional regulatory CD4 T cells (Treg; ref. 44). Although Tregs are known to influence outcomes in ovarian cancer (4), we did not detect significant numbers of CD103-expressing Tregs during our analyses. In contrast, we demonstrate herein that, compared with their CD103⁻ counterparts, CD103⁺CD8⁺ TILs express higher levels of the activation markers HLA-DR, PD-1, and Ki-67 and lower levels of the naïve T-cell marker CD127. Most strikingly, virtually all CD103⁺ TILs expressed TIA-1, a marker of cytolytic potential that is strongly associated with survival in EOC (45) and breast cancer (46). The strong association between CD103⁺ TILs and patient survival, together with their universal expression of the cytolytic marker TIA-1, suggests that CD103 demarcates a subset of highly activated, cytolytic CD8⁺ TILs. However, the expression of PD-1 by CD103⁺CD8⁺ TILs could also signify functional exhaustion (47). Indeed, one can imagine that CD103⁺ TILs might be trapped within the tumor bed through interactions with E-cadherin expressed on tumor cells, and that they subsequently become exhausted due to chronic antigen stimulation. Indeed, the fact that these cells express high levels of PD-1 suggests that their antitumor activity might be further enhanced through blockade of PD-1 or other inhibitory pathways.

Our finding that CD103 is preferentially expressed by intraepithelial TILs also has practical implications for TILs scoring. In particular, a worldwide task force is being assembled to validate the prognostic significance of TILs as defined by the so-called Immunoscore (48). The ultimate goal of this effort is to bring the Immunoscore into standard clinical practice as a prognostic biomarker. Colorectal cancer is likely to be the first indication, but other tumor sites are being considered. It is often difficult to distinguish between epithelial versus stromal compartments of tumors, especially for nonexperts. Our results suggest that immunohistochemical detection of CD103⁺ TILs may provide an objective surrogate means to score intraepithelial TILs. Indeed, we showed a strong correlation between CD103⁺ TILs and survival in HGSC without the need to discriminate their epithelial versus stromal location (Fig. 6A).

Although the majority of CD103⁺ TILs observed herein were CD8⁺ T cells, the next largest subset was comprised of CD56⁺ NK cells. Like CD8⁺ T cells, NK cells express CD103 when exposed to TGF-β (49); however, it is unknown whether they require a second signal equivalent to antigen stimulation. In healthy individuals, CD103⁺CD56⁺ NK cells are found in microenvironments rich in TGF-β, including uterine decidua (50) and tonsil. Very little is known about the functional profile or prognostic significance of CD3⁻CD56⁺ NK cells in HGSC. To date, greater attention has been focused on CD56⁺CD3⁺ NKT-like cells, which correlate with poor outcome as defined by platinum resistance (51). Because CD103⁺ NK cells were almost always found together with CD103⁺CD8⁺ TILs, we were unable to precisely define their independent prognostic significance. However, given the strong association between CD103⁺ TILs and survival, further study regarding the role of CD103⁺CD56⁺ TILs in tumor immunity is clearly warranted.

In closing, the strong association between intraepithelial CD8⁺ TILs and patient survival has spawned great interest in the use of TILs as predictive and prognostic biomarkers and as effector cells for immunotherapy. We propose that CD103, by serving as a definitive marker of these cells, has the potential to dramatically increase our understanding of TIL function, antigen specificity and immunotherapeutic potential, opening new avenues in our understanding of tumor immunity in EOC and other cancers.

No potential conflicts of interest were disclosed.

Conception and design: J.R. Webb, B.H. Nelson

Development of methodology: J.R. Webb, K. Milne, R.J. deLeeuw

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.R. Webb, K. Milne, R.J. deLeeuw, B.H. Nelson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.R. Webb, K. Milne, P. Watson, R.J. deLeeuw, B.H. Nelson

Writing, review, and/or revision of the manuscript: J.R. Webb, K. Milne, P. Watson, R.J. deLeeuw, B.H. Nelson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Milne

Study supervision: B.H. Nelson

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.