Genetic Variants Synthesize to Produce Paneth Cell Phenotypes that Define Subtypes of Crohn’s Disease

Abstract

Background & Aims

Genetic susceptibility loci for Crohn’s disease (CD) are numerous, complex, and likely interact with undefined components of the environment. It has been a challenge to link the effects of particular loci to phenotypes of cells associated with pathogenesis of CD, such as Paneth cells. We investigated whether specific phenotypes of Paneth cells associated with particular genetic susceptibility loci can be used to define specific subtypes of CD.

Methods

We performed a retrospective analysis of 119 resection specimens collected from patients with CD at 2 separate medical centers. Paneth cell phenotypes were classified as normal or abnormal (with disordered, diminished, diffuse, or excluded granule phenotypes) based on lysozyme-positive secretory granule morphology. To uncover the molecular basis of the Paneth cell phenotypes, we developed methods to determine transcriptional profiles from whole-thickness and laser-capture microdissected, formalin-fixed, paraffin-embedded tissue sections.

Results

The proportion of abnormal Paneth cells was associated with the number of CD-associated NOD2 risk alleles. The cumulative number of NOD2 and ATG16L1 risk alleles had an additive effect on the proportion of abnormal Paneth cells. Unsupervised clustering analysis of demographic and Paneth cell data divided patients into 2 principal subgroups, defined by high and low proportions of abnormal Paneth cells. The disordered and diffuse abnormal Paneth cell phenotypes were associated with an altered transcriptional signature of immune system activation. We observed an inverse correlation between abnormal Paneth cells and the presence of granuloma. Moreover, high proportions of abnormal Paneth cells were associated with shorter time to disease recurrence after surgery.

Conclusions

Histologic analysis of Paneth cell phenotypes can be used to divide patients with CD into subgroups with distinct pathognomonic and clinical features.

INTRODUCTION

The inflammatory bowel diseases (IBD), Crohn’s disease (CD) and ulcerative colitis (UC), currently affect approximately 1.5 million people in the United States and are a significant cause of morbidity, particular among young people^{1, 2}. A major hurdle in understanding the pathogenesis of complex chronic diseases, such as IBD, is incorporating the effects of both numerous genetic susceptibility loci and poorly defined environmental factors. This challenge often precludes precise genotype-phenotype correlations and further identification of disease mechanism-based surrogate markers³. One potential mitigating factor is the accessibility of clinical tissue samples from both involved and non-involved areas for analyses as this may allow the incorporation of histopathological and other tissue characteristics into genetic analyses in order to better understand the etiopathogenesis of disease. Among the major complex genetic diseases, IBD is unique in that tissue samples are routinely obtained as part of clinical practice. Therefore, IBD serves as an ideal platform to test the hypothesis that histological changes are a more homogenous phenotype than standard clinical manifestations for testing genotype-phenotype correlations.

One CD relevant cell type is the Paneth cell, which is specialized secretory cell type located at the bases of the crypts of Lieberkühn in the small intestine⁴. These cells produce a wide repertoire of antimicrobial peptides, such as lysozyme and α-defensins, to modulate the intestinal microbiome^{5, 6}, and thus are important mediators of the host innate immune response^{4, 7}. We previously demonstrated that the packaging of anti-microbial peptides into granules and their secretion was impaired in the Paneth cells of mice with hypomorphic expression of the CD susceptibility gene Atg16l1^{8, 9}. Importantly, we observed similar Paneth cell abnormalities in a small number of CD patients homozygous for the ATG16L1 T300A CD risk allele⁸, demonstrating that this is a valid approach to link genetics and phenotypes of a disease-relevant cell type. In addition to ATG16L1, NOD2 has been identified as a CD susceptibility locus and has been predicted to disrupt Paneth cell function^10–14.

CD is remarkable for both its heterogeneous clinical course and its varied histopathological findings^15–17. The clinical variability in natural history and response to therapy is likely, in part, a consequence of the genetic heterogeneity that underlies these conditions. Major challenges to genotype-phenotype association studies are the lack of robust and reproducible criteria to define endpoints as well as sufficient numbers of genotyped patients. Recent genome-wide association studies (GWAS) have extended the number of known IBD susceptibility loci to more than 160. These studies and others have implicated multiple pathways in IBD pathogenesis, including epithelial barrier homeostasis, innate immune response, antigen presentation, autophagy, Paneth cell defects, and IL-23/T_H17 signaling^{10, 11, 18, 19}. Here, we hypothesized that linking genetics to disease-associated phenotypes in relevant cell types based on predicted disease mechanisms (i.e., Paneth cells) may be a successful method for defining more homogenous subtypes of CD. We also propose that analyzing regions of intestine free of severe active or chronic inflammation (i.e., lack of substantial pathologic hallmarks) will provide an objective endpoint that will more accurately reflect disease pathogenesis, as these areas may harbor early molecular and pathologic changes. A greater understanding of the causes of the observed clinical heterogeneity will lead to improved clinical management through a more individualized approach to disease management and, potentially, the development of new therapies.

MATERIALS AND METHODS

Description and genotyping of patient cohort

Patients were recruited at Barnes-Jewish Hospital, St Louis between 2005 and 2013 or at Cedars-Sinai Medical Center, Los Angeles between 1999 and 2013. Patient DNA samples were genotyped for ATG16L1 T300A and the CD-associated NOD2 variants^{10, 20, 21}. Patients from the Barnes-Jewish Hospital cohort were genotyped by the Digestive Disease Research Core Center (DDRCC) using matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry and by the Genome Technology Access Center (GTAC) using the Human OmniQuad SNP genotyping arrays (Illumina, San Diego, CA). Patients from the Cedars-Sinai cohort were genotyped using the Immunochip (Illumina). The study protocol was approved by the Institutional Review Boards of Washington University – St. Louis and Cedars-Sinai Medical Center. Written informed consent was obtained from all study participants.

Morphological analysis of Paneth cells

For each resection case, a hematoxylin and eosin-stained tissue section of the proximal margin (terminal ileum) was identified by pathologists (T.S.S. and T.C.L.). Cases were included for Paneth cell analysis if the section contained at least 100 well-oriented intestinal crypts and exhibited absent or minimal acute and/or chronic inflammation (Figure S1). Lysozyme distribution was quantified as previously described⁸. For each case, a pathologist (T.C.L.) who was blinded to the characteristics of the cases scored a minimum of 200 Paneth cells (range: 206–2,702) in well-oriented crypts. Paneth cells located within Peyer’s patches were excluded.

Transcriptional analysis

RNA was procured from the set of archived formalin-fixed paraffin-embedded surgical resection samples used for histological analysis. Microarrays were performed as previously described⁸. Data are deposited at ArrayExpress (http://www.ebi.ac.uk/arrayexpress/) with accession number E-MTAB-1281.

Statistical analyses

For analysis of lysozyme quantification, permutation tests were performed to determine the association between NOD2 variants and the percentage of abnormal Paneth cells using R statistical software (version 2.13.1; R Foundation for Statistical Computing). Mann-Whitney tests were used to demonstrate statistical difference between cases with 1 or 2 NOD2 risk variants and controls. Linear regression was used to analyze the cumulative number of risk variants. For correlation analyses, Pearson correlations were calculated using GENE-E²², which were then used as the distance measure for unsupervised, hierarchical clustering of the patients. A marker selection strategy based on signal-to-noise ratios²³ was used to identify clinical variables associated with patient subtypes. A Chi-Square test and a log-rank test were performed for the analysis of granuloma incidence and time to disease recurrence, respectively (Prism GraphPad software). P < 0.05 was considered to be significant.

RESULTS

Association of NOD2 CD susceptibility variants with abnormal Paneth cell phenotype

We performed a retrospective analysis of Paneth cell phenotypes in genotyped CD patients (n = 119) using resection specimens. In order to study tissue that may exhibit early pathologic and molecular changes associated with disease pathogenesis, we examined ileal tissue samples that demonstrated no evidence of active/chronic disease. Paneth cell analysis was performed using our previously developed system for robust, quantitative scoring of Paneth cell phenotypes based on high-resolution localization of lysozyme protein⁸, a highly-expressed antimicrobial protein that is normally packaged into Paneth cell secretory granules⁷. A pathologist (T.C.L.) who was blinded to the individuals’ characteristics scored a minimum of 200 Paneth cells per case as normal or as abnormal (including Paneth cells scored as disordered, diminished, diffuse or ‘excluded granule’) (Figure 1A; Figure S2). The ‘excluded granule’ abnormal Paneth cell phenotype was identified in this study and is characterized by granule shapes that have low/absent lysozyme staining but also contain diffuse cytoplasmic lysozyme staining and occasionally a few lysozyme-positive granules (Figure 1B). For most cases, Paneth cells with normal morphology were predominant. When present, Paneth cells with an abnormal phenotype were found interspersed among those with normal morphology.

Figure 1. CD-associated NOD2 risk alleles are associated with abnormal Paneth cell morphology.

(A–C) Lysozyme immunostaining (red) was performed to visualize and score Paneth cell secretory granule morphology from cases with 0 (n = 29), 1 (n = 25) or 2 (n = 5) CD-associated NOD2 risk variants. Paneth cells were scored as normal if they contained numerous small (~1 μm), lysozyme-positive apically located granules. Disordered Paneth cells contained lysozyme-positive granules of normal size and quantity, but had some basally located granules. Diminished Paneth cells contained <10 granules, with the remaining granules frequently enlarged or fused. Diffuse Paneth cells did not contain any secretory granules and had diffuse lysozyme staining throughout their cytoplasm. (A) Representative images of lysozyme immunostaining. Nuclear counterstain, DAPI (blue). Well-oriented Paneth cells are outlined in white and the scored phenotype is indicated: normal (N), disordered (Dis), diminished (Dim), diffuse (Dif) or excluded granule (Exc). Bars = 10 μm. (B) Representative images of Paneth cells with the excluded granule phenotype (arrowheads). (C) Quantification of Paneth cell phenotypes according to the number of CD-associated NOD2 risk variants. (D) Quantification of the percent of abnormal Paneth cells in cases with 0 CD-associated NOD2 susceptibility alleles (n = 29) or in cases that carried one allele of R702W (n = 10), G908R (n = 4) or L1007fsXinsC (n = 9). Data are presented as the mean ± s.e.m. of the percent of Paneth cells with the indicated phenotype out of the total number of Paneth cells counted for each case. Cases with the ATG16L1 T300A variant were excluded from these analyses. Permutation tests were performed to demonstrate the statistically significant association between the presence of CD-associated NOD2 variants and the percentage of abnormal Paneth cells for each scoring category (*, P < 0.05; **, P < 0.01). Mann-Whitney tests were used to demonstrate statistical difference between the presence of 1 or 2 NOD2 risk variants and controls (†, P < 0.05; ††, P < 0.01).

In light of the established effect of the ATG16L1 T300A CD susceptibility variant on Paneth cell phenotype⁸, we initially excluded patients with this variant. The ATG16L1 T300A CD risk allele is common, with a risk allele carriage rate greater than 80% in people of European descent²⁰ (http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=2241880). Therefore, we included cases from both the Washington University School of Medicine and the Cedars-Sinai Medical Center patient cohorts. Confounding effects due to specimen source were not apparent. In this combined cohort ATG16L1 ‘safe’ cases (cases 1–59, Table S1), we first demonstrated that the group of cases without NOD2 risk variants (n = 29) and the group of cases with one or more of the common and/or rare²¹ CD-associated NOD2 variants (n = 30) had similar demographic compositions (Table 1). In control cases without NOD2 CD risk alleles, an average of 87% of the total number of Paneth cells counted per case had a normal phenotype based on lysozyme localization. In cases with one or more NOD2 CD susceptibility alleles, the average percent of Paneth cells scored as normal was significantly lower than controls (76% with one variant allele and 69% with two variant alleles) (Figure 1C). Accordingly, cases with one or more NOD2 CD susceptibility variants had significantly increased proportions of disordered, diminished and diffuse abnormal Paneth cell phenotypes compared to the control cases (Figure 1C).

Table 1.

Baseline characteristics of CD cohort¹.

	NOD2 Variants Absent N = 29 (%)	NOD2 Variant(s) Present N = 30 (%)	Fisher’s Exact Test P
Sex (Male)	13 (45)	15 (50)	0.796
Family history of IBD	9 (31)	8 (27)	0.779
Presence of stricturing and/or penetrating disease	25 (86)	24 (80)	0.731
Presence of upper small bowel disease	7 (24)	6 (20)	0.761
Presence of perianal disease	5 (17)	11 (37)	0.143
ASA treatment	18 (67)	21 (70)	0.589
Steroids treatment	22 (81)	22 (73)	1.000
Antibiotics treatment	14 (52)	17 (57)	0.606
Anti-TNF alpha treatment	15 (56)	15 (50)	1.000
Non-smoker status	23 (79)	26 (87)	0.506

¹Patients with the ATG16L1 T300A allele were excluded from this analysis.

ASA, acetylsalicylic acid; TNFalpha, tumor necrosis factor.

The NOD2 locus contains three common CD susceptibility alleles, namely R702W, G908R and L1007fsXinsC. Several studies have suggested that each of these variants attenuate NOD2 function based on an impaired cellular response following exposure to the NOD2 ligand, bacterial peptidoglycan^24–27. We investigated whether these variants were similarly associated with abnormal Paneth cell phenotype in the ATG16L1 ‘safe’ cohort (Figure 1D). We found that the R702W and L1007fsXinsC variants were associated with increased proportions of abnormal Paneth cells. Of note, two groupings of cases were observed for each of these alleles, one with < 20% abnormal Paneth cells and the other with ≥ 20% abnormal Paneth cells. This observation suggests that the NOD2 variant alone is not sufficient to drive the abnormal Paneth cell phenotypes, but rather additional genetic variants or host/environmental factors also contribute to this phenotype. We did not observe an association between NOD2 G908R and abnormal Paneth cell phenotypes in this study. However, we were able to obtain only a small number of patients with this allele and thus cannot exclude the possibility of such an association. Similarly, potential associations of abnormal Paneth cell phenotypes and compound NOD2 variants (e.g., R702W + L1007fxXinsC), rare CD-associated NOD2 variants or homozygous variants were not interpretable due to the rarity of cases with these genotypes (Supplemental Figure S3).

Paneth cell phenotypes define CD patient subtypes

As our previous studies showed that the CD risk variant ATG16L1 T300A was associated with increased proportions of abnormal Paneth cell phenotypes, we next wanted to test for a potential interaction between this variant and the CD-associated NOD2 variants. For this analysis, we used a larger CD cohort that included both ATG16L1 T300A ‘safe’ and ‘risk’ cases (cases 1–119, Tables S1 and S2). Similar to our previous study⁸, we observed an association between ATG16L1 T300A and abnormal Paneth cells in this expanded cohort (Figure S4). We observed a significant, positive correlation between the cumulative number of ATG16L1 T300A or CD-associated NOD2 risk alleles and the proportion of abnormal Paneth cells, demonstrating an additive effect of these alleles (Figure 2A).

Figure 2. Abnormal Paneth cell phenotypes define a distinct subtype of CD cases.

(A) Total abnormal Paneth cell proportions were examined in a larger cohort (n = 119) with cumulative number of ATG16L1 T300A and CD-associated NOD2 variants (including common and rare variants) 0 (n = 29), 1 (n = 41), 2 (n = 33), 3 (n = 10) or 4 (n = 6). Linear regression analysis demonstrated a significant, positive correlation between the percent of abnormal Paneth cells per case and cumulative risk alleles. (B) Unsupervised clustering was performed for the CD cases with Paneth cell phenotype, genetic, documented environmental exposure and demographic information available. Cases are arranged in the same order along the x and y axes. The Pearson correlation was determined for each patient-patient comparison, and then patients were clustered according to similarity with the results represented as a heat map. (C) Heat map displaying the factors that significantly contribute to the identification of Subtypes 1 and 2 in (B), which were defined by low and high proportions of abnormal Paneth cells, respectively. Heat map cells are colored red and blue to indicate positive and negative correlations, respectively. PCs, Paneth cells.

As a second experiment to test the association between the CD-associated NOD2 and ATG16L1 T300A variants and Paneth cell phenotype, we performed a patient-patient comparison using the same expanded cohort. Unsupervised analysis of Paneth cell phenotypes, genetics, documented environmental exposures and demographic information was used to cluster the CD cases according to similarity (Figure 2B). Interestingly, the resulting heat map showed two principal subtypes of cases. A minor number of cases did not fit either subtype. Using a marker selection strategy²³, we identified the factors that defined the two CD patient subtypes: the abnormal Paneth cell phenotype categories of diminished, diffuse and excluded granule Paneth cells and the total percent of abnormal Paneth cells (Figure 2C). The cumulative number of NOD2 risk variants and NOD2 R702W genotype were also defining factors for the two CD patient subtypes, supporting our initial finding that CD-associated NOD2 variants are associated with abnormal Paneth cell phenotypes. We also performed the inverse experiment and looked for potential correlations between Paneth cell phenotypes and the other experimental factors (genetics, documented environmental exposures and demographic information). Significant correlations were again observed between the abnormal Paneth cell phenotype categories and NOD2 genotype, but we did not observe significant correlations between the abnormal Paneth cell phenotypes and the environmental/demographic parameters (Figure S5). Taken together, these data demonstrate that Paneth cell phenotype is strongly linked to particular genetic alleles and can more clearly define subtypes of CD patients than other demographic parameters (treatment, disease behavior, disease location, etc.).

An activated immune response gene signature is associated with diffuse Paneth cell phenotype

To investigate the molecular basis of the Paneth cell phenotypes, we developed a method whereby we could obtain RNA from the same archival formalin-fixed, paraffin-embedded tissue samples used to morphologically phenotype Paneth cells. Transcriptional profiles of the CD patient material were generated by performing microarray analysis with one set of RNAs procured by laser capture microdissection of ileal crypt bases (enriched for Paneth cells, n = 15 samples) and a second set of RNAs procured from whole, unstained ileal tissue sections scraped from glass microscope slides (n = 40). To analyze these data (Figure S6), we first compared the two sets of transcriptional profiles to identify transcripts enriched in the Paneth cells of CD patients. Next, we performed a correlation analysis between the expression of the Paneth cell-enriched genes and the quantitative Paneth cell phenotype data for the same 40 patients using the whole ileal tissue data set to identify distinct sets of Paneth cell-enriched transcripts with expression values that were significantly correlated with each Paneth cell phenotype (Table S4).

We performed gene ontology (GO) term analysis for each transcript set that was correlated with a particular Paneth cell phenotype (Figure 3 and Figures S7, S8 and S9). This analysis showed that the transcript sets that correlated with the disordered and diffuse Paneth cell phenotypes were both enriched for immune system-related biological processes. We previously observed a gene signature of cytokine stimulation in the Paneth cells of mice with hypomorphic expression of Atg16l1 (which have increased proportions of Paneth cells with the disordered and diffuse phenotypes)⁸. These data suggest that altered immune activation may be occurring in human and mouse intestinal tissue that exhibits abnormal Paneth cell phenotypes, although the underlying causes of these transcriptional profiles remains unclear. Thus, the disordered and diffuse Paneth cell phenotypes appear to be histological readouts of specific types of inflammatory responses occurring in a subset of CD patients.

Figure 3. A specific gene signature associated with an activated immune response correlates with the disordered and diffuse Paneth cell phenotypes.

The 114 and 218 transcripts with expression values that correlated with the disordered and diffuse Paneth cell phenotype, respectively, were analyzed using DAVID to identify biological process gene ontology (GO) terms significantly overrepresented in each gene set. (A, B) GO terms related to the immune response are indicated (fully labeled bar graphs are shown in Figures S8 and S9). Bars are colored according to the major GO term category: biological process (gray), cellular component (white), molecular function (black). PC, Paneth cell.

Abnormal Paneth cell phenotype inversely correlates with granuloma incidence

We next examined whether other, more classic morphological features of CD, such as the presence of granuloma, were correlated with the Paneth cell phenotypes. Granulomas are a distinguishing feature of CD that is not present in all patients²⁸. For this analysis, pathologists (T.C.L. and D.D.) reviewed the entire case for each patient, including both involved and non-involved intestinal regions (n = 107, cohort from patients 1–119 with available material, Tables S1 and S2). We found that cases with low proportions of abnormal Paneth cell phenotypes (arbitrarily defined as < 20% based on this study and our previous in vivo studies⁸) had a higher incidence of granuloma than cases with high proportions of abnormal Paneth cell phenotypes (43.8% vs. 19.4%, respectively; P = 0.0160) (Figure 4A).

Figure 4. Abnormal Paneth cell phenotypes and presence of granuloma are inversely correlated.

The incidence of granuloma was determined in our expanded CD cohort (n = 107). Correlation of granuloma incidence and cases with low (white bars) vs. high proportions (black bars) of (A) abnormal Paneth cell phenotypes (cutoff ≥ 20%; n = 63 [< 20%], n = 36 [≥ 20%]), (B) ‘excluded granule’ Paneth cell phenotypes (cutoff ≥ 5%; n = 71 [< 1%], n = 28 [≥1%]), (C) diminished granules (cutoff ≥ 10%; n = 42 [< 10%], n = 57 [≥ 10%]) and (D) diffuse granules (cutoff ≥ 5%; n = 92 [< 5%], n = 7 [≥ 5%]). Statistical significance was determined by Chi-Square test (*P ≤ 0.05).

We then examined whether the presence of granuloma was associated with a particular abnormal Paneth cell phenotype. We found that the excluded granule phenotype exhibited the strongest inverse correlation with granuloma incidence, with 39.4% of cases with low proportions of excluded granule Paneth cells (< 1%) having granuloma compared to 17.9% of cases with high proportions of excluded granules (≥ 1%) (Figure 4B; P = 0.0344). A similar finding was observed with the diminished granule Paneth cell phenotype (46.5% vs. 26.8% incidence in cases with low [<10%] and high [≥10%] proportions of diminished granule Paneth cells, respectively; P = 0.0347) (Figure 4C). Finally, cases with low proportions of diffuse granule Paneth cells (<5%) had 37.0% granuloma incidence compared to 14.3% granuloma incidence observed in cases with high proportions of diffuse granules (≥5%), although this correlation was not significant (P = 0.4159; Figure 4D). In accordance with previous reports, we did not find a significant association between NOD2 status (including individual risk alleles) and granuloma incidence in our cohort (data not shown)^28–30. In summary, we identified that abnormal Paneth cell phenotypes are inversely associated with the presence of granuloma.

Paneth cell phenotype is associated with disease prognosis

We next investigated if Paneth cell phenotypes were associated with the disease prognosis post-resection. For this analysis, we used our expanded CD cohort (cases 1–119, Tables S1 and S2) as well as an additional 59 CD cases that have not yet been genotyped but had Paneth cell scoring and disease recurrence data (cases 120–178, Table S3). In patients who received prophylactic therapy post-resection (i.e., suggestive of patients with a more aggressive clinical course prior to resection; n =102), those with high proportions of abnormal Paneth cells (≥20%) had a significantly shorter time to disease recurrence compared to those with low proportions of abnormal Paneth cells (< 20%) (P = 0.0200; Figure 5). A significant difference was not observed in patients who did not receive prophylaxis post-resection (i.e., patients with a less aggressive clinical course prior to resection) (Figure S10). Thus, we conclude that Paneth cell phenotypes can identify clinically relevant subtypes of CD patients.

Figure 5. Abnormal Paneth cell phenotypes correlate with shorter time to disease recurrence in CD patients that received prophylaxis post-surgery.

The time to disease recurrence post-resection surgery was determined in the CD cases that received prophylaxis post-surgery (n = 102). Prophylaxis included use of immunomodulators (e.g., 6-mercaptopurine, azathioprine, methotrexate) or biologics (i.e., anti-TNF therapy). Cases were assigned to the “Abnormal Paneth cells: Low” or “Abnormal Paneth cells: High” groups using a cutoff of 20% abnormal Paneth cell phenotypes. Data are presented as a survival curve with *P = 0.0200 by Log-rank test.

DISCUSSION

Here, we have provided the first evidence that a defined cellular phenotype (in Paneth cells) is linked to multiple CD genetic susceptibility loci and subdivides patients into two groups. In addition, we defined the molecular consequences of this phenotype in human Paneth cells and found an association with immune activation. We demonstrated that Paneth cell phenotypes are associated with the presence of granuloma, a classic histological finding associated with CD. We also showed that the Paneth cell phenotypes are associated with a specific clinical outcome in the current cohort, i.e., time to disease recurrence. While there clearly is a genetic basis for this phenotype, because of the potential complexity of the genetics and environmental interactions, we propose that the Paneth cell phenotypes would be a readily accessible, integrative readout of this information (rather than genetics alone) and should be further evaluated for their ability to stratify CD patients in a clinically meaningful way.

Here, we demonstrated that CD cases with one or more NOD2 susceptibility alleles had increased proportions of abnormal Paneth cells compared to those without NOD2 susceptibility alleles. In this study, we also found an additive effect of NOD2 susceptibility alleles and ATG16L1 T300A on Paneth cell abnormalities. This finding is of interest because Paneth cell phenotypes have now been demonstrated to link two IBD susceptibility genes that have been suspected to act in a shared biochemical pathway^31–33. It has been proposed that IBD genes act in pathways¹¹, and our findings provide additional evidence that supports this theory.

There is a pressing need to classify subtypes of CD based on underlying pathological mechanisms including genetics and not solely by clinical parameters. This need is highlighted by the fact that treatment with anti-TNF-α monoclonal antibodies, the most advanced biologic currently employed, does not induce remission in the majority of CD cases, and the majority of initial responders do not maintain long-term remission^{34, 35}. Furthermore, many potential therapeutics for CD have failed to induce or maintain remission of active disease³⁵. Although some of these “failed” therapeutics truly do lack efficacy, inappropriate endpoint selection and patient heterogeneity have been cited as factors contributing to treatment inefficacy in clinical trials³⁴. Thus, it is becoming increasingly apparent that novel strategies to define and stratify CD patients that are based on genotype and other molecular parameters will be needed to progress towards improved diagnostics, prognostics and therapeutics. Because the ability to test potential CD therapeutics in a particular subtype of patients may yield better outcomes, the abnormal Paneth cell phenotypes, as an objective biomarker that is both related to genetic susceptibility loci and disease pathogenesis mechanisms, should be explored as a method to identify a subtype of CD patients that has a similar disease pathogenesis mechanism.

In this study, we observed a novel and important histological correlate to the Paneth cell phenotype, an inverse relationship between the presence of granuloma and the proportion of abnormal Paneth cells. This is, to our knowledge, the first report linking defects in a CD-relevant cell type and other, more classical histologic changes of CD. The standard practice for pathologic examination for CD includes extensive sampling of the specimens (at a minimum of 1 section/10 cm), and as there might be unsampled granulomas in patients where no granulomas were found, it is perhaps best to classify the two subsets of CD patients as “granuloma-rich” and “granuloma-poor”. We showed that the total proportion of abnormal Paneth cells, and in particular, excluded and diminished granules, were strong indicators of the granuloma-poor subtype. This is important as while we showed that NOD2 status was associated with Paneth cell phenotype, we and others also showed that there was no significant correlation between NOD2 status and granuloma^28–30, and that Paneth cell phenotype is the best predictor of granuloma. Our work supports the model that cell-specific readouts that can integrate the effects of both genetic and environmental factors are more informative and clinically relevant (Figure 6). There has been controversy over the biologic and prognostic implications of granulomas in CD^{36, 37}, with some recent studies indicating that granulomas might be linked to more aggressive clinical behavior²⁸ and others that it is not^{37, 38}. An immediate clinical implication of our finding is that stratification of CD based on Paneth cell phenotypes may be a more reliable approach for clinical management and clinical trial design, as features such as granuloma could be sparse and more likely to be undersampled, especially in biopsy specimens, whereas Paneth cell phenotypes are more easily analyzed within limited samples. Prospective studies using Paneth cell phenotypes as stratification for novel CD therapeutics may yield new insights.

Figure 6. Proposed model for the use of disease-relevant phenotypes as integrative readouts for the stratification of CD patients.

Abnormal Paneth cell phenotypes can synthesize multiple genetic and environmental inputs. These phenotypes are associated with molecular and pathological features and can be used to subdivide CD patients.

In contrast to the success that has been achieved in treating single gene diseases (e.g., chronic myelogenous leukemia and imatinib (Gleevac)³⁹), the development of effective therapeutic options for complex genetic diseases, such as rheumatoid arthritis, systemic lupus erythematosus and IBD has been difficult^{34, 40, 41}. Of the complex genetic diseases, CD studies are uniquely positioned to lead the way in the development of novel therapies. First, there is a clearer understanding of the genetic susceptibility loci that are associated with CD susceptibility compared to rheumatoid arthritis or systemic lupus erythematosus. Second, relevant tissue and cell types (not restricted to Paneth cells) are relatively easy to access. In contrast, for diseases such as type I diabetes, whereas there is a good understanding of the associated genetics, there is poor tissue availability³. Thus, we believe that CD can be the prototype disease for moving forward the study of complex genetic diseases for both the development of new therapeutics and a more personalized approach to disease management.

Abbreviations

GO

gene ontology

GWAS

genome-wide association studies

TNF-α

tumor necrosis factor α