Human Pathology (2015) 46, 643–656

www.elsevier.com/locate/humpath

Original contribution

Gastric-type expression signature in serrated pathway–associated colorectal tumors☆,☆☆ Jung Ho Kim MD, PhD a , Kyung-Ju Kim MD b , Ye-Young Rhee MD b , Jeong Mo Bae MD b , Nam-Yun Cho MSc c , Hye Seung Lee MD, PhD d , Gyeong Hoon Kang MD, PhD b,c,⁎ a

Department of Pathology, SMG-SNU Boramae Medical Center, Seoul 156-707, Republic of Korea Department of Pathology, Seoul National University College of Medicine, Seoul 110-799, Republic of Korea c Laboratory of Epigenetics, Cancer Research Institute, Seoul National University College of Medicine, Seoul 110-799, Republic of Korea d Department of Pathology, Seoul National University Bundang Hospital, Seongnam 463-707, Republic of Korea b

Received 24 October 2014; revised 22 December 2014; accepted 2 January 2015

Keywords: Colorectal cancer; Serrated polyp; Serrated pathway; Gastric differentiation; Immunohistochemistry

Summary Accumulating evidence has indicated that serrated pathway–associated colorectal tumors may be associated with aberrant gastric-type differentiation. Here, we investigated the immunoexpression profiles of gastric-type markers and intestinal-type markers in colorectal tumors, focusing on their relation to serrated pathway–associated tumors. Immunohistochemistry for 7 gastric-type markers (ANXA10, VSIG1, CLDN18, CTSE, TFF2, MUC5AC, and MUC6) and 2 intestinal-type markers (CDX2 and CK20) was performed in 36 normal gastric/colorectal mucosa tissues, 163 colorectal polyps, and 175 microsatellite-unstable colorectal carcinomas (MSI-H CRCs). In normal tissues, all 7 candidate gastric-type markers showed expressional specificity for normal gastric mucosa. Among the colorectal polyps, sessile serrated adenoma/polyps demonstrated the highest positive rate of ANXA10, CLDN18, MUC5AC, and MUC6 expression (87%, 35%, 61%, and 52%, respectively). Microvesicular hyperplastic polyps showed the highest frequencies of ANXA10, VSIG1, and TFF2 positivity (87%, 87%, and 67%, respectively). ANXA10 and MUC6 expression was not detected in all conventional adenomas. In MSI-H CRCs, the expression of ANXA10, TFF2, and MUC5AC was significantly associated with sporadic tumors (P b .001, P = .01, and P b .001, respectively). Moreover, all of the 7 gastric-type markers were significantly related to preferential expression in proximal colon carcinomas among MSI-H CRCs. CDX2 and CK20 expression was retained in all colorectal polyps, whereas there were significantly high frequencies of CDX2 loss (28%) and CK20 loss (29%) in sporadic tumors among MSI-H CRCs. In conclusion, the early gain of gastric differentiation and late loss of intestinal differentiation are immunophenotypic features in the serrated pathway to colorectal carcinoma. © 2015 Elsevier Inc. All rights reserved.

☆

Competing interests: The authors declare no conflict of interest. Funding/Support: This study was supported by a grant from the Basic Science Research Program through the National Research Foundation (NRF) funded by the Ministry of Education (2013R1A1A2059080), a grant from the Korean Health Technology R&D Project, Ministry of Health and Welfare (HI13C1804), the Priority Research Centers Program through the NRF funded by the Ministry of Education, Science and Technology (2009-0093820), and the NRF grant funded by the Ministry of Science, ICT, and Future Planning (2011-0030049). ⁎ Corresponding author at: Department of Pathology, Seoul National University College of Medicine, 103 Daehak-ro, Jongno-gu, Seoul 110-799, Republic of Korea. E-mail address: [email protected] (G. H. Kang). ☆☆

http://dx.doi.org/10.1016/j.humpath.2015.01.003 0046-8177/© 2015 Elsevier Inc. All rights reserved.

644

1. Introduction Advances in the understanding of the pathologic and molecular basis of the serrated pathway to colorectal carcinoma (CRC) have provided novel insights into diagnostic and therapeutic approaches for colorectal polyps and CRCs. Over the past 2 decades, many researchers have reported important findings regarding the characteristics of serrated pathway–associated colorectal tumors. These findings included the pathologic identification of serrated polyps, which led to the establishment of serrated polyps as the precursor lesions of CRCs and to the subclassification of serrated polyps into hyperplastic polyps (HPs), sessile serrated adenoma/polyps (SSA/Ps), and traditional serrated adenomas (TSAs) [1] and the unique molecular associations of serrated pathway–associated colorectal tumors with microsatellite instability (MSI), the CpG island methylator phenotype (CIMP), and KRAS/BRAF mutations [2]. Among serrated polyps, SSA/Ps are considered the principal precursors of serrated pathway–associated CRCs [3]. It is also thought that most sporadic MSI-high (MSI-H) and/or CIMP-high (CIMP-H) CRCs may develop from SSA/Ps [2,3]. This concept is based on the clinicopathological and molecular similarities between SSA/Ps and sporadic MSI-H and/or CIMP-H CRCs. These similar features include high frequencies of proximal colonic tumor location, female predominance, MSI-H status, CIMP-H status, and BRAF V600E mutations. Despite the improved recognition of the clinicopathological significance of SSA/Ps, some problems must still be resolved, such as the difficulty in the histopathologic differential diagnosis of serrated polyp subtypes and the lack of detailed knowledge regarding the causal factors and malignant transformational drivers of SSA/Ps. In this regard, previous studies have reported notable findings regarding candidate molecular markers that are significantly up-regulated in SSA/Ps, which include annexin A10 (ANXA10), v-set and immunoglobulin domain containing 1 (VSIG1), cathepsin E (CTSE), and trefoil factor 2 (TFF2) [4,5]. In addition, previous investigations have revealed that the aberrant expression of mucin 5AC (MUC5AC) and mucin 6 (MUC6) was frequently observed in serrated polyps [6-9]. Interestingly, most of these markers are thought to have expressional specificity for normal gastric mucosal epithelium but not for normal intestinal mucosal epithelium. In fact, the gastric differentiation feature of serrated polyps was previously suggested by some researchers [10,11], and both Walsh et al [12] and Tsai et al [13] reported that serrated pathway–associated CRCs are closely related to gastric-type mucin expression. However, these investigations provide only limited information because the full landscape of differential gastric-type expression profiles of colorectal tumors, which depends on different carcinogenesis pathways and different carcinogenesis steps, was not depicted. In the present study, to investigate whether the aberrant expression of gastric-type markers is associated with the

J. H. Kim et al. specific tumorigenesis pathway in the large intestine, we analyzed the immunohistochemical expression profile of gastric-type markers, which previous studies have suggested as candidate markers for serrated polyps, in colorectal polyps, including conventional adenomas and serrated polyps, and MSI-H CRCs. In addition, to simultaneously compare the expression alteration patterns of gastric-type markers and intestinal-type markers along the serrated neoplasia pathway, the immunohistochemical expression of caudal-type homeobox 2 (CDX2) and cytokeratin 20 (CK20), which are representative intestinal-type markers [14], was evaluated in the same samples.

2. Materials and methods 2.1. Case selection We collected 18 normal gastric mucosa tissues, 18 normal colorectal mucosa tissues, 42 conventional adenomas with low-grade dysplasia (CALGs), 21 conventional adenomas with high-grade dysplasia (CAHGs), 23 goblet cell–rich HPs (GCHPs), 30 microvesicular HPs (MVHPs), 16 TSAs, 31 SSA/Ps (26 SSA/Ps without dysplasia and 5 SSA/Ps with dysplasia), and 175 MSI-H CRCs. All formalin-fixed, paraffin-embedded (FFPE) tissues of normal gastric and colorectal mucosa, colorectal polyps, and MSI-H CRCs were retrieved from the pathology archives of our institutions. To select the optimal tissues for this study, hematoxylin and eosin–stained tissue slides of normal mucosa tissues and colorectal polyps were microscopically examined by 3 gastrointestinal pathologists (J. H. K., K. J. K., and G. H. K.). A total of 18 normal gastric mucosa tissues were obtained from surgically resected specimens to remove any type of gastric neoplasms; these tissues originated from 6 subsites of the stomach (3 antrum, 3 low body, 3 mid body, 3 high body, 3 fundus, and 3 cardia). Gastric mucosal areas showing intestinal metaplasia were excluded through microscopic examination. A total of 18 normal colorectal mucosa tissues were obtained from surgically resected specimens to remove any type of colorectal neoplasms; these tissues originated from 6 subsites of the large intestine (3 cecum, 3 ascending colon, 3 transverse colon, 3 descending colon, 3 sigmoid colon, and 3 rectum). Colorectal polyps including conventional adenomas and serrated polyps were collected from endoscopic mucosal resection or polypectomy specimens originating from the procedures conducted to remove colorectal polyps at the Seoul National University Hospital or Boramae Medical Center between 2010 and 2012. Through the microscopic examination, cases that could not be definitively diagnosed as 1 of the 6 histologic subtypes or that had inadequate material for tissue microarray (TMA) construction were excluded. In the determination of the subtypes of serrated

Gastric phenotype in colorectal tumors polyps, cases showing only superficial crypts or incomplete crypt bases were excluded. After these selection steps, 163 polyps were finally included in this study. The histopathologic criteria to determine adenoma/polyp subtypes were primarily based on the latest World Health Organization classification [1]. Specifically, cases classified as CALG should show tubular, tubulovillous, or villous structure with basally located, polarized, pencillate, and hyperchromatic nuclei and retained normal glandular architecture, indicating low-grade dysplasia. Cases classified as CAHG should show tubular, tubulovillous, or villous structure with enlarged atypical nuclei, loss of polarization, loss of mucin, a high nucleus-to-cytoplasm ratio, and loss of normal glandular architecture, indicating high-grade dysplasia. Cases classified as GCHP should show elongated and straight crypts, superficial and subtle crypt serrations, narrow crypt bases, and an absence of cytologic dysplasia and be mostly composed of goblet cells. Cases classified as MVHP should show elongated and straight crypts, upper crypt serrations, narrow crypt bases, and an absence of cytologic dysplasia and be mostly composed of microvesicular cells. Cases classified as TSA should show villiform or filiform configuration, crypt hyperserrations, ectopic crypts formation, and dysplastic cellular features, including tall columnar cells, narrow pencillate nuclei, and hypereosinophilic cytoplasm. Cases classified as SSA/P should show prominent serrations of superficial to deep crypts and dilated, distorted or branching crypt bases, including L-shaped, inverted T–shaped, or anchor-shaped bases. To classify a polyp as SSA/P, there should be at least 2 contiguous crypts demonstrating the typical features of SSA/P. MSI-H CRC tissues were collected, as previously described [15]. Initially, consecutive series of 213 CRCs that had previously been determined as MSI-H were retrieved from the pathology archives of Seoul National University Hospital and Seoul National University Bundang Hospital. In this study, although we could not perform DNA analyses to identify germline mutations in DNA mismatch repair (MMR) genes, our MSI-H CRC samples were categorized into sporadic MSI-H CRCs or presumed Lynch syndrome (LS)–associated MSI-H CRCs using strict molecular criteria that were modified from our previous study [16]. Sporadic MSI-H CRCs were defined as having at least 1 of the 3 following molecular factors: CIMP-H, MLH1 promoter methylation and BRAF V600E mutation. All cases classified as sporadic MSI-H CRCs should demonstrate a loss of MLH1 expression, and tumors determined to have other DNA MMR deficiencies (MSH2/MSH6 negative, MSH6 negative only, or PMS2 negative only) were excluded from classification as sporadic MSI-H CRCs. Presumed LS-associated MSI-H CRCs were defined as having all of the following factors: CIMP-low/negative status, an absence of MLH1 methylation, an absence of BRAF mutation, non–MLH1-type MMR deficiencies (MSH2/MSH6 negative, MSH6 negative only, or PMS2 negative only), and an age at the time of CRC diagnosis of

645 younger than 50 years. After this categorization, of the 213 initially collected MSI-H CRCs, 76 cases were determined as sporadic tumors, and 99 cases were classified into LS-associated tumors. The remaining 38 cases were classified as indeterminate tumors and were excluded from this study. This study was approved by the institutional review board (no. H-1203-072-402).

2.2. Selection of gastric-type markers Through an intensive review of the literature, we pooled a total of 53 candidate markers for serrated pathway– associated colorectal tumors that had been reported to be up-regulated genes and/or overexpressed proteins specific to SSA/Ps and/or HPs in previous studies using gene expression profiling, protein expression profiling, or immunohistochemical analysis [4,5,7,17]. Of the 53 candidate markers, we finally selected 7 gastric-type markers that satisfied both of the following criteria: (1) According to the human protein expression database (the Human Protein Atlas; www.proteinatlas.org) [18], the marker should show high expression patterns in normal gastric mucosal tissues but low or negative expression patterns in normal colorectal mucosal tissues. (2) There should be a commercially available antibody suitable for immunohistochemistry (IHC) on FFPE tissues for the marker. The 7 gastric-type markers were ANXA10, claudin 18 (CLDN18), CTSE, MUC5AC, MUC6, TFF2, and VSIG1.

2.3. Immunohistochemistry TMA construction and IHC staining were conducted, as previously described [15]. For TMA construction, 1 representative adenoma/polyp area in each of the 163 colorectal adenoma/polyp case specimens was extracted as 1 tissue core (2 mm in diameter), and 3 different tumor areas in each of the 175 MSI-H CRC case specimens were extracted as 3 tissue cores (2 mm in diameter). In this study, all IHC processes were automatically conducted using a BenchMark XT immunostainer (Ventana Medical Systems, Tucson, AZ) according to the manufacturer's protocol. IHC for the 7 gastric-type markers was performed on TMA sections using an anti-ANXA10 antibody (1:300, NBP1-90156; Novus Biologicals, Littleton, CO), an anti-VSIG1 antibody (1:100, NBP1-81074; Novus Biologicals), an anti-CLDN18 antibody (1:50, NBP1-60010; Novus Biologicals), an anti-CTSE antibody (1:100, sc-30055; Santa Cruz Biotechnology, Dallas, TX), an anti-TFF2 antibody (1:300, HPA036705; Atlas Antibodies, Stockholm, Sweden), an anti-MUC5AC antibody (1:80, NCL-MUC-5 AC, Leica Biosystems, Newcastle upon Tyne, UK), and an anti-MUC6 antibody (1:80, NCL-MUC-6, Leica Biosystems). All IHC staining evaluations in this study were carried out independently by 2 pathologists (J. H. K. and K. J. K.), who were blinded to the patients' clinicopathological and molecular data. To determine positivity in colorectal polyps

646

J. H. Kim et al.

Fig. 1 The immunohistochemical expression status of candidate gastric-type markers in normal gastric mucosa versus normal colorectal mucosa. A, All 7 candidate gastric-type markers showed positive expression in normal gastric mucosa tissues. B, All candidate gastric-type markers, except CTSE, showed completely negative expression in normal colorectal mucosa tissues. CTSE demonstrated negative expression in most colorectal epithelial cells, but focally positive cells were observed occasionally. Original magnification ×50. Abbreviations: LB, low body; MB, mid body; HB, high body; A-colon, ascending colon; T-colon, transverse colon; D-colon, descending colon; S-colon, sigmoid colon.

and CRCs, the typical staining patterns of each of the gastric-type markers were compared with their staining patterns observed in normal gastric mucosa (Fig. 1A). Positive expression of each of the gastric-type markers was defined as staining in at least 10% of the adenoma/polyp areas or carcinoma cells in TMA cores and should satisfy the following subcellular staining patterns: ANXA10, nuclear; VSIG1, membranous; CLDN18, membranous; CTSE, cytoplasmic; TFF2, cytoplasmic and/or membranous;

MUC5AC, cytoplasmic; and MUC6, cytoplasmic. Cytoplasmic staining of ANXA10 and nuclear or cytoplasmic staining of CLDN18 were frequently observed in colorectal tissues and were regarded as negative (nonspecific) expression. Focal positive expression of each marker was defined by positive staining in less than 50% of the tumor areas in TMA sections, whereas diffuse positive expression was defined by positive staining in 50% or more of the tumor areas in the TMA sections. Immunostaining for CDX2,

Gastric phenotype in colorectal tumors

647

Fig. 1 (continue).

CK20 and MMR proteins was performed and assessed, as previously described [14,15]. Conflicting assessment results between the pathologists were reviewed and discussed, and a consensus was reached.

2.4. DNA analysis DNA analyses for the determination of CIMP status, MLH1 promoter methylation, and BRAF/KRAS mutations in

the 213 MSI-H CRCs were performed, as previously described [15].

2.5. Statistical analysis Statistical analyses were performed using the IBM SPSS Statistics version 20 software (Chicago, IL). Comparisons of the categorical data were conducted using the χ2 test or Fisher exact test. All P values were 2 sided, and P b .05 indicated statistical significance.

648

J. H. Kim et al.

Fig. 2 The immunohistochemical expression status of gastric-type markers in colorectal polyps. A, Representative photomicrographs of IHC for gastric-type markers (ANXA10, VSIG1, CLDN18, CTSE, TFF2, MUC5AC, and MUC6) and intestinal-type markers (CDX2 and CK20) in SSA/Ps and CALGs (×100). Note the positivity of both gastric-type markers and intestinal-type markers in SSA/Ps and the negativity of gastric-type markers and the positivity of intestinal-type markers in CALGs. B, Expression map of gastric-type markers and intestinal-type markers in 163 colorectal polyps.

Gastric phenotype in colorectal tumors

649

Fig. 2 (continued).

3. Results 3.1. Validation of tissue specificity of candidate gastric-type markers As a preliminary step in our study, IHC for the 7 candidate gastric-type markers (ANXA10, VSIG1, CLDN18, CTSE, TFF2, MUC5AC, and MUC6) was performed in 18 normal gastric mucosa tissues and 18 normal colorectal mucosa tissues. The aim of this analysis was to confirm the specificity of these IHC markers for normal gastric mucosa. Consistent with our expectation, IHC for all 7 markers showed moderate to strong positive expression in all 18 normal gastric mucosa tissues (Fig. 1A). ANXA10, VSIG1, CLDN18, and CTSE were expressed in full levels of mucosal epithelium (from surface epithelium to deep glands) throughout anatomic subsites of the stomach (Fig. 1A). TFF2 expression was also detected in full levels of gastric mucosal epithelium but was reduced in fundic glands (Fig. 1A). Characteristically, MUC5AC was expressed in surface and pit epithelium of normal gastric mucosa, whereas MUC6

expression was localized in deep glands of normal gastric mucosa, especially in pyloric glands (Fig. 1A). In contrast with normal gastric mucosa, all 18 normal colorectal mucosa tissues demonstrated completely negative staining of all of the candidate gastric-type markers except CTSE (Fig. 1B). In IHC for CTSE in colorectal tissues, although focal superficial epithelium or some scattered crypt cells were occasionally positive for CTSE, most colorectal epithelial cells (N90% of the mucosal area) were negative for CTSE.

3.2. Staining patterns of gastric-type markers in colorectal polyps IHC for the 7 gastric-type markers (ANXA10, VSIG1, CLDN18, CTSE, TFF2, MUC5AC, and MUC6) was performed and analyzed in 163 colorectal polyps (42 CALGs, 21 CAHGs, 23 GCHPs, 30 MVHPs, 16 TSAs, and 31 SSA/Ps). Subcellular staining features of each gastric marker in the epithelium of colorectal polyps were as follows: ANXA10 showed a moderate to strong nuclear staining pattern with occasional weak to moderate cytoplasmic staining in the

650

J. H. Kim et al.

epithelial cells of colorectal polyps (Fig. 2A). VSIG1-positive and CLDN18-positive epithelial cells in the polyps showed primarily weak to moderate membranous staining (Fig. 2A). CTSE and MUC5AC showed moderate to strong cytoplasmic staining in the epithelial cells of polyps (Fig. 2A). TFF2positive epithelial cells mainly showed weak membranous staining, with occasional moderate to strong granular cytoplasmic staining (Fig. 2A). MUC6 positivity was predominantly detected in the basal crypts of polyps and showed weak to moderate cytoplasmic staining (Fig. 2A). These typical staining patterns of gastric-type markers in colorectal polyps were remarkably observed in serrated polyps, whereas the staining of gastric-type markers was rarely observed in conventional adenomas (Fig. 2A).

3.3. Gastric-type expression signature in subtypes of colorectal polyps Both the distribution of the 7 gastric-type markers expression according to the subtypes of colorectal polyps and the associations of this gastric-type expression signature with tumor location and CDX2/CK20 expression statuses in 163 colorectal polyps are graphically depicted in Fig. 2B. In this landscape, we confirmed that the gastric-type expression signature occurred predominantly in serrated polyps, particularly in MVHPs and SSA/Ps. We also determined that CDX2 and CK20 expression was well retained in all colorectal polyps, regardless of the gastric-type expression status (Fig. 2B). Among the subtypes of colorectal polyps, MVHPs and SSA/Ps showed the highest frequencies of ANXA10 positivity (both 87%; Table 1). CLDN18, MUC5AC, and Table 1

MUC6 demonstrated their highest positive rate in SSA/Ps (35%, 61%, and 52%, respectively; Table 1), whereas the highest frequencies of VSIG1 and TFF2 expression were observed in MVHPs (87% and 67%, respectively; Table 1). Most of the serrated polyps, including GCHPs, TSAs, MVHPs, and SSA/Ps, showed CTSE positivity (96%-100%; Table 1), but the frequencies of CTSE expression in CALGs and CAHGs were relatively low (31% and 33%, respectively; Table 1). Interestingly, consistent with a study by Wiland et al [19], most ANXA10-positive TSAs showed focal positive staining for ANXA10 (b50% of adenoma area) (75% of ANXA10-positive TSAs; Fig. 2B), whereas ANXA10 expression in SSA/Ps dominantly exhibited a diffuse staining pattern (≥50% of adenoma area) (63% of ANXA10-positive SSA/Ps; Fig. 2B). Moreover, in contrast with SSA/Ps and MVHPs, TSAs and GCHPs showed the absence of positivity or very rare focal positivity of MUC6 expression (0% of TSAs and 4% of GCHPs; Table 1). In terms of tumor location, significantly more frequent positivity of ANXA10 and VSIG1 was observed in proximal SSA/Ps (96% for ANXA10 and 91% for VSIG1; Table 1), compared with distal SSA/Ps (63% for ANXA10 and 50% for VSIG1; Table 1) (P = .043 for ANXA10 and P = .026 for VSIG1). In MVHPs, ANXA10 and MUC6 expression was also significantly associated with proximal tumor location (100% for ANXA10 and 39% for MUC6; Table 1) (P = .018 for ANXA10 and P = .024 for MUC6). Based on the results from our TMA analysis, in terms of diagnostic utility, we further tested whether ANXA10 and/or MUC6 expression can differentiate SSA/Ps with dysplasia (SSA/P-Ds) from conventional adenomas and can differentiate SSA/Ps from TSAs. Among the 31 SSA/Ps, 26 cases

Frequencies of immunohistochemical expression of gastric-type markers in colorectal adenomas/polyps

All CALGs Proximal CALGs Distal CALGs All CAHGs Proximal CAHGs Distal CAHGs All GCHPs Proximal GCHPs Distal GCHPs All TSAs Proximal TSAs Distal TSAs All MVHPs Proximal MVHPs Distal MVHPs All SSA/Ps Proximal SSA/Ps Distal SSA/Ps

ANXA10 positivity

VSIG1 positivity

CLDN18 positivity

CTSE positivity

TFF2 positivity

MUC5AC positivity

MUC6 positivity

0/42 (0%) 0/10 (0%) 0/32 (0%) 0/21 (0%) 0/3 (0%) 0/18 (0%) 6/23 (26%) 3/6 (50%) 3/17 (18%) 12/16 (75%) 2/2 (100%) 10/14 (71%) 26/30 (87%) 18/18 (100%) 8/12 (67%) 27/31 (87%) 22/23 (96%) 5/8 (63%)

1/42 1/10 0/32 3/21 1/3 2/18 8/23 2/6 6/17 7/16 1/2 6/14 26/30 16/18 10/12 25/31 21/23 4/8

0/42 (0%) 0/10 (0%) 0/32 (0%) 1/21 (5%) 0/3 (0%) 1/18 (6%) 2/23 (9%) 1/6 (17%) 1/17 (6%) 3/16 (19%) 0/2 (0%) 3/14 (21%) 10/30 (33%) 7/18 (39%) 3/12 (25%) 11/31 (36%) 8/23 (35%) 3/8 (38%)

13/42 (31%) 2/10 (20%) 11/32 (34%) 7/21 (33%) 1/3 (33%) 6/18 (33%) 22/23 (96%) 6/6 (100%) 16/17 (94%) 16/16 (100%) 2/2 (100%) 14/14 (100%) 30/30 (100%) 18/18 (100%) 12/12 (100%) 31/31 (100%) 23/23 (100%) 8/8 (100%)

0/42 0/10 0/32 1/21 0/3 1/18 6/23 1/6 5/17 10/16 2/2 8/14 20/30 9/18 11/12 19/31 15/23 4/8

5/42 (12%) 0/10 (0%) 5/32 (16%) 2/21 (10%) 0/3 (0%) 2/18 (11%) 10/23 (44%) 3/6 (50%) 7/17 (41%) 5/16 (31%) 0/2 (0%) 5/14 (36%) 13/30 (43%) 8/18 (44%) 5/12 (42%) 19/31 (61%) 15/23 (65%) 4/8 (50%)

0/42 (0%) 0/10 (0%) 0/32 (0%) 0/21 (0%) 0/3 (0%) 0/18 (0%) 1/23 (4%) 0/6 (0%) 1/17 (6%) 0/16 (0%) 0/2 (0%) 0/14 (0%) 7/30 (23%) 7/18 (39%) 0/12 (0%) 16/31 (52%) 13/23 (57%) 3/8 (38%)

(2%) (10%) (0%) (14%) (33%) (11%) (35%) (33%) (35%) (44%) (50%) (43%) (87%) (89%) (83%) (81%) (91%) (50%)

(0%) (0%) (0%) (5%) (0%) (6%) (26%) (17%) (29%) (63%) (100%) (57%) (67%) (50%) (92%) (61%) (65%) (50%)

Abbreviations: CALG, conventional adenoma low-grade dysplasia; CAHG, conventional adenoma high-grade dysplasia; GCHP, goblet cell–rich hyperplastic polyp; TSA, traditional serrated adenoma; MVHP, microvesicular hyperplastic polyp; SSA/P, sessile serrated adenoma/polyp; NA, not applicable.

Gastric phenotype in colorectal tumors were determined to be SSA/Ps without dysplasia, and 5 cases were determined to be SSA/P-Ds. IHC for ANXA10 and MUC6 was repeatedly performed and evaluated on whole tissue sections of 5 SSA/P-Ds and 16 TSAs. Among the 5 SSA/P-Ds, ANXA10 positivity and MUC6 positivity in dysplastic areas were detected in 4 (80%) and 2 (40%) cases, respectively (Supplementary Fig. S1). It was also confirmed that MUC6 was not expressed in basal crypts of all 16 TSAs (Supplementary Fig. S1).

3.4. Gastric-type expression signature in MSI-H CRCs IHC for the 7 gastric-type markers (ANXA10, VSIG1, CLDN18, CTSE, TFF2, MUC5AC, and MUC6) and the 2 intestinal-type markers (CDX2 and CK20) was performed and analyzed in 175 MSI-H CRCs, including 76 sporadic tumors and 99 LS-associated tumors. Representative images of IHC expression of the 7 gastric-type markers and the 2 intestinaltype markers in sporadic MSI-H CRCs and LS-associated MSI-H CRCs are presented in Fig. 3A. As depicted in the expression map of 175 MSI-H CRCs (Fig. 3B), the positive expression of gastric-type markers and negative expression of intestinal-type markers were predominantly observed in sporadic MSI-H CRCs. Through a statistical analysis, of the 7 gastric-type markers, the positive expression of ANXA10, TFF2, and MUC5AC was found to be significantly associated with sporadic tumors in MSI-H CRCs (P b .001, P = .01, and P b .001, respectively; Table 2). Significantly high frequencies of CDX2 loss and CK20 loss were also observed in sporadic MSI-H CRCs (P b .001 and P = .005, respectively; Table 3). In addition, among the MSI-H CRCs, all of the 7 gastric-type markers demonstrated significantly preferential expression in proximal colon carcinomas (P b .001 for ANXA10, P = .022 for VSIG1, P = .012 for CLDN18, P = .015 for CTSE, P = .006 for TFF2, P = .005 for MUC5AC, and P = .044 for MUC6; Table 4). Finally, detailed correlations of each gastric-type marker with various clinicopathological and molecular features in MSI-H CRCs were analyzed. In MSI-H CRCs, ANXA10 positivity was significantly associated with old age (≥57 years; P = .005), female sex (P = .032), advanced stage (stage III/IV; P b .001), mucinous histology (P = .048), CDX2 loss (P = .002), CIMP-H status (P b .001), MLH1 methylation (P b .001), BRAF mutation (P = .009), and wild-type KRAS status (P = .022) (Supplementary Table S1). VSIG1 expression was correlated with female sex (P = .025), medullary histology (P = .017), CDX2 loss (P = .038), CK20 loss (P = .002), and MLH1 methylation (P = .024) (Supplementary Table S2). CLDN18 expression showed significant associations with advanced stage (P b .001), poor differentiation (P = .049), signet ring cell carcinoma (P = .001), CK20 loss (P = .048), and wild-type KRAS (P = .049) (Supplementary Table S3). CTSE positivity was only significantly related to mucinous histology (P = .017) (Supplementary Table S4). TFF2 expression was associated with old age (P = .035), advanced stage (P = .007), mucinous component (P = .001), signet ring cell component (P b .001),

651 CDX2 loss (P b .001), CK20 loss (P = .002), CIMP-H (P = .014), and MLH1 methylation (P = .005) (Supplementary Table S5). Aberrant expression of MUC5AC was correlated with old age (P = .002), advanced stage (P = .035), extracellular mucin pools (P = .001), signet ring cell feature (P = .011), CDX2 loss (P = .001), CK20 loss (P = .008), CIMP-H (P b .001), MLH1 methylation (P b .001), BRAF mutation (P = .019), and absence of KRAS mutation (P = .049) (Supplementary Table S6). MUC6 expression demonstrated significant correlations with mucinous histology (P = .022), CDX2 loss (P = .027), and MLH1 methylation (P = .039) in MSI-H CRCs (Supplementary Table S7).

4. Discussion In this study, we successfully demonstrated that the frequent and extensive expression of gastric-type markers was characteristically observed in serrated pathway–associated colorectal tumors, including serrated polyps and sporadic MSI-H CRCs. However, one of the interesting findings regarding the gastric-type expression signature in colorectal tumors revealed in this study was the lack of exclusiveness of gastric-type expression for serrated pathway–associated colorectal tumors. The positive expression of gastric-type markers was observed not only in serrated polyps and sporadic MSI-H CRCs but also in conventional adenomas and LS-associated MSI-H CRCs (Figs. 2B and 3B), although the frequencies of positivity of gastric-type markers in conventional adenomas and LSassociated CRCs were generally lower than those in serrated polyps and sporadic MSI-H CRCs, respectively (Tables 1 and 2). This characteristic could be inferred from previous investigations on the expression of ANXA10 in MSI-H CRCs [16,20,21]. According to these prior studies, it was evident that ANXA10-positive expression could be detected in a small subset of LS-associated CRCs. Another feature of gastric-type expression in colorectal tumors was its preference for proximal colonic tumors. In MVHPs, SSA/Ps, and MSI-H CRCs, the positive expression of gastric-type markers was generally more frequent in proximal tumors than in distal tumors (Tables 1 and 4). According to our preliminary data regarding the ANXA10 expression status in presumed chromosomal instability (CIN)–associated CRCs demonstrating both microsatellite-stable (MSS) and CIMP-low/ negative molecular phenotypes, ANXA10 positivity was observed in 2.4% of presumed CIN-associated CRCs (unpublished data). Moreover, the dominant location of these ANXA10-positive CIN-associated CRCs was the proximal colon (58%), although CIN-associated CRCs were predominantly located in the distal colon and rectum. Taken together, these findings indicate that the gastric-type expression signature is not exclusive to serrated pathway–associated colorectal tumors, although there is an obvious positive correlation between gastric-type expression and serrated pathway–associated colorectal tumors, and gastric-type expression may have a preference for tumors

652

J. H. Kim et al.

Gastric phenotype in colorectal tumors

653

Fig. 3 (continued). Fig. 3 The immunohistochemical expression status of gastric-type markers in MSI-H CRCs. A, Representative photomicrographs of IHC for gastric-type markers (ANXA10, VSIG1, CLDN18, CTSE, TFF2, MUC5AC, and MUC6) and intestinal-type markers in MSI-H CRCs (×200). Note the positivity of gastric-type markers and the negativity of intestinal-type markers in sporadic MSI-H CRCs and the negativity of gastric-type markers and the positivity of intestinal-type markers in LS-associated MSI-H CRCs. B, Expression map of gastric-type markers in 175 MSI-H CRCs and its associations with various pathologic and molecular features.

arising in the proximal colon, regardless of the colorectal carcinogenesis pathways. In terms of the diagnostic usefulness of the 7 gastric-type markers in colorectal tumors, the use of IHC for each gastric-type marker to diagnose serrated polyp histologic subtypes or CRC molecular subtypes may be limited in practical application because these markers demonstrated a lack of exclusiveness for serrated pathway–associated colorectal tumors, as mentioned above. However, IHC for certain gastric-type markers can be somewhat helpful in differentiating serrated polyps from conventional adenomas or discriminating SSA/Ps from TSAs. For example, in contrast with conventional adenomas, all of which were negative for

ANXA10 and MUC6 expression (Table 1 and Fig. 2B), up to 80% of SSA/P-Ds showed positivity of ANXA10 and/or MUC6 (Supplementary Fig. S1). Therefore, ANXA10 and MUC6 can be used as putative immunohistochemical markers differentiating serrated pathway–associated dysplasia from conventional pathway–associated dysplasia in colorectal polyps with dysplasia. This finding also indicates that the aberrant gastric-type expression already appears before the dysplastic change of SSA/P and is generally maintained after the acquisition of dysplasia in SSA/P. IHC for MUC6 can also be used in differential diagnosis between SSA/P and TSA. According to our data, MUC6 expression was not observed in TSAs, whereas 52% of SSA/Ps showed MUC6 positivity in

654

J. H. Kim et al.

Table 2 Frequencies of immunohistochemical expression of gastric-type markers in sporadic MSI-H CRCs versus LS-associated MSI-H CRCs Gastric-type markers ANXA10 Positive Negative VSIG1 Positive Negative CLDN18 Positive Negative CTSE Positive Negative TFF2 Positive Negative MUC5AC Positive Negative MUC6 Positive Negative

Cases

Sporadic MSI-H CRCs (n = 76)

LS-associated MSI-H CRCs (n = 99)

P

36 139

28 (37%) 48 (63%)

8 (8%) 91 (92%)

b.001

30 145

17 (22%) 59 (78%)

13 (13%) 86 (87%)

.108

62 113

32 (42%) 44 (58%)

30 (30%) 69 (70%)

.106

136 39

62 (82%) 14 (18%)

74 (75%) 25 (25%)

.282

41 134

25 (33%) 51 (67%)

16 (16%) 83 (84%)

.01

68 107

41 (54%) 35 (46%)

27 (27%) 72 (73%)

b.001

24 151

13 (17%) 63 (83%)

11 (11%) 88 (89%)

.253

Table 4 Frequencies of immunohistochemical expression of gastric-type markers in proximal-located MSI-H colon carcinomas versus distal-located MSI-H CRCs Distal-located P Gastric-type Cases Proximal-located MSI-H CRCs markers MSI-H colon carcinomas (n = 114) (n = 61) ANXA10 Positive Negative VSIG1 Positive Negative CLDN18 Positive Negative CTSE Positive Negative TFF2 Positive Negative MUC5AC Positive Negative MUC6 Positive Negative

36 139

33 (29%) 81 (71%)

3 (5%) 58 (95%)

b.001

30 145

25 (22%) 89 (78%)

5 (8%) 56 (92%)

.022

62 113

48 (42%) 66 (58%)

14 (23%) 47 (77%)

.012

136 39

95 (83%) 19 (17%)

41 (67%) 20 (33%)

.015

41 134

34 (30%) 80 (70%)

7 (11%) 54 (89%)

.006

68 107

53 (46%) 61 (54%)

15 (25%) 46 (75%)

.005

24 151

20 (18%) 94 (82%)

4 (7%) 57 (93%)

.044

Abbreviations: MSI-H, microsatellite instability–high; CRCs, colorectal cancers; LS, Lynch syndrome.

Abbreviations: MSI-H, microsatellite instability–high; CRCs, colorectal cancers; LS, Lynch syndrome.

their basal crypts (Table 1, Fig. 2B, and Supplementary Fig. S1). Therefore, MUC6 can be used as a supportive immunohistochemical marker differentiating SSA/Ps from TSAs in equivocal cases. The detailed utility of IHC for the 7 gastric-type markers in histopathologic differential diagnosis of colorectal polyps and CRCs should be further investigated. One of our study's limitations was that the CRC samples used were restricted to the MSI-H molecular phenotype. In fact, MSI-H CRC can be a good model to compare

differential features between serrated pathway–associated tumors and non–serrated pathway–associated tumors because it is thought that sporadic MSI-H CRCs are a representative subset of serrated pathway–associated CRCs. Moreover, we had intensively collected and analyzed a large series of MSI-H CRCs through our previous studies. Although we confirmed a predilection for the expression of gastric-type markers in sporadic MSI-H CRCs, there remains a need to investigate the expression profile of gastric-type markers in overall CRCs, including MSS CRCs as well as MSI-H CRCs. Through this extended investigation, more comprehensive expression profiles of gastric-type markers in CRCs, including CIMP-H/MSS CRCs and CIN/MSS CRCs, which represent serrated pathway CRCs and conventional pathway CRCs, respectively, can potentially be established. Another limitation of our study was the use of TMA to evaluate the IHC expression status of gastric-type markers in colorectal polyps. Microscopically demonstrating the whole vertical dimension of crypts in colorectal polyps may occasionally fail when TMA cores are used instead of whole tissue sections. Therefore, IHC staining of TMA sections could result in the incomplete detection of the expression patterns of the IHC markers that are dependent on the crypt's sublocation. For example, it has been known that, in serrated polyps, MUC6 expression was generally localized to the crypt base [7]; therefore, MUC6 expression in the

Table 3 Frequencies of immunohistochemical expression of intestinal-type markers in sporadic MSI-H CRCs versus LS-associated MSI-H CRCs Gastric-type markers CDX2 Positive Negative CK20 Positive Negative

Cases

Sporadic MSI-H CRCs (n = 76)

LS-associated MSI-H CRCs (n = 99)

P

151 24

55 (72%) 21 (28%)

96 (97%) 3 (3%)

b.001

141 34

54 (71%) 22 (29%)

87 (88%) 12 (12%)

.005

Abbreviations: MSI-H, microsatellite instability–high; CRCs, colorectal cancers; LS, Lynch syndrome.

Gastric phenotype in colorectal tumors

655

Fig. 4 Schematic summary of this study. The 2 immunophenotypic characteristics of the serrated pathway to CRC, gain of gastric differentiation and loss of intestinal differentiation, are shown.

crypt base could be missed in TMA sections of serrated polyps. In our study, although we made our best effort to ensure that the TMA cores included intact vertical crypts from original FFPE tissue blocks of colorectal polyps, the possibility of errors in IHC assessment due to the incomplete inclusion of crypts in TMA cores cannot be entirely excluded. To evaluate more precise patterns of IHC expression of gastric-type markers in colorectal polyps, further IHC studies using whole tissue sections of colorectal polyps should be encouraged. In summary, we identified 2 immunophenotypic features, including “gain of gastric differentiation” and “loss of intestinal differentiation,” as pathologic hallmarks of the colorectal serrated neoplasia pathway (Fig. 4). In contrast to the loss of CDX2 and/or CK20 expression in serrated pathway–associated colorectal tumors, which is found exclusively in carcinomas, but not in polyps, the aberrant expression of gastric-type markers can be observed in most early serrated precursor lesions, including MVHPs and GCHPs (Figs. 2B and 3B). Collectively, the gain of gastric differentiation is an early event in the serrated neoplasia pathway and involves a substantial number of serrated precursor lesions, whereas the loss of intestinal differentiation is a late aberration in the serrated neoplasia pathway and involves a limited subset of CRCs and may predominantly arise during the malignant transformation of a subset of serrated precursor lesions (Fig. 4).

Supplementary data Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.humpath.2015.01.003.

References [1] Bosman FT, Carneiro F, Hruban RH, Theise ND. World Health Organization. International Agency for Research on Cancer. WHO classification of tumours of the digestive system. Lyon: International Agency for Research on Cancer; 2010.

[2] Bettington M, Walker N, Clouston A, Brown I, Leggett B, Whitehall V. The serrated pathway to colorectal carcinoma: current concepts and challenges. Histopathology 2013;62:367-86. [3] Rex DK, Ahnen DJ, Baron JA, et al. Serrated lesions of the colorectum: review and recommendations from an expert panel. Am J Gastroenterol 2012;107:1315-29 [quiz 4, 30]. [4] Gonzalo DH, Lai KK, Shadrach B, et al. Gene expression profiling of serrated polyps identifies annexin A10 as a marker of a sessile serrated adenoma/polyp. J Pathol 2013;230:420-9. [5] Delker DA, McGettigan BM, Kanth P, et al. RNA sequencing of sessile serrated colon polyps identifies differentially expressed genes and immunohistochemical markers. PLoS One 2014;9:e88367. [6] Owens SR, Chiosea SI, Kuan SF. Selective expression of gastric mucin MUC6 in colonic sessile serrated adenoma but not in hyperplastic polyp aids in morphological diagnosis of serrated polyps. Mod Pathol 2008;21:660-9. [7] Bartley AN, Thompson PA, Buckmeier JA, et al. Expression of gastric pyloric mucin, MUC6, in colorectal serrated polyps. Mod Pathol 2010; 23:169-76. [8] Fujita K, Hirahashi M, Yamamoto H, et al. Mucin core protein expression in serrated polyps of the large intestine. Virchows Arch 2010;457:443-9. [9] Gibson JA, Hahn HP, Shahsafaei A, Odze RD. MUC expression in hyperplastic and serrated colonic polyps: lack of specificity of MUC6. Am J Surg Pathol 2011;35:742-9. [10] Yao T, Kouzuki T, Kajiwara M, Matsui N, Oya M, Tsuneyoshi M. “Serrated” adenoma of the colorectum, with reference to its gastric differentiation and its malignant potential. J Pathol 1999;187:511-7. [11] Koike M, Inada K, Nakanishi H, Matsuura A, Nakamura S, Tatematsu M. Cellular differentiation status of epithelial polyps of the colorectum: the gastric foveolar cell-type in hyperplastic polyps. Histopathology 2003;42: 357-64. [12] Walsh MD, Clendenning M, Williamson E, et al. Expression of MUC2, MUC5AC, MUC5B, and MUC6 mucins in colorectal cancers and their association with the CpG island methylator phenotype. Mod Pathol 2013;26:1642-56. [13] Tsai JH, Lin YL, Cheng YC, et al. Aberrant expression of annexin A10 is closely related to gastric phenotype in serrated pathway to colorectal carcinoma. Mod Pathol 2015;28:268-78. [14] Kim JH, Rhee YY, Bae JM, Cho NY, Kang GH. Loss of CDX2/CK20 expression is associated with poorly differentiated carcinoma, the CpG island methylator phenotype, and adverse prognosis in microsatellite-unstable colorectal cancer. Am J Surg Pathol 2013; 37:1532-41. [15] Kim JH, Kim KJ, Rhee YY, et al. Expression status of wild-type HSP110 correlates with HSP110 T17 deletion size and patient

656 prognosis in microsatellite-unstable colorectal cancer. Mod Pathol 2014;27:443-53. [16] Kim JH, Kang GH. Annexin A10 expression in microsatellite-unstable colorectal cancers: is it specific to sporadic tumors? Am J Surg Pathol 2014;38:1577-9. [17] Sentani K, Sakamoto N, Shimamoto F, Anami K, Oue N, Yasui W. Expression of olfactomedin 4 and claudin-18 in serrated neoplasia of the colorectum: a characteristic pattern is associated with sessile serrated lesion. Histopathology 2013;62:1018-27. [18] Ponten F, Jirstrom K, Uhlen M. The human protein atlas—a tool for pathology. J Pathol 2008;216:387-93.

J. H. Kim et al. [19] Wiland IV HO, Shadrach B, Allende D, et al. Morphologic and molecular characterization of traditional serrated adenomas of the distal colon and rectum. Am J Surg Pathol 2014;38:1290-7. [20] Kim JH, Rhee YY, Kim KJ, Cho NY, Lee HS, Kang GH. Annexin A10 expression correlates with serrated pathway features in colorectal carcinoma with microsatellite instability. APMIS 2014; 122:1187-95. [21] Pai RK, Shadrach BL, Carver P, et al. Immunohistochemistry for annexin A10 can distinguish sporadic from Lynch syndrome– associated microsatellite-unstable colorectal carcinoma. Am J Surg Pathol 2014;38:518-25.