Article Text

The stem cell marker CD133 is highly expressed in sessile serrated adenoma and its borderline variant compared with hyperplastic polyp
  1. Mahin Mohammadi1,3,
  2. Michael Bzorek1,
  3. Jesper Hansen Bonde2,3,
  4. Hans J Nielsen4,
  5. Susanne Holck3
  1. 1Department of Pathology, Hospital South, Naestved, Denmark
  2. 2Department of Clinical Research Center, Copenhagen University Hospital, Hvidovre, Denmark
  3. 3Department of Pathology, Copenhagen University Hospital, Hvidovre, Denmark
  4. 4Department of Surgical Gastroenterology, Copenhagen University Hospital, Hvidovre, Denmark
  1. Correspondence to Dr Mahin Mohammadi, Department of Pathology, Copenhagen University Hospital, Kettegård Alle 30, Hvidovre 2650, Denmark; Mahin.mohammadi{at}


Non-dysplastic serrated polyps (ND-SP) represent a heterogeneous group of colorectal lesions that comprise hyperplastic polyp (HP) and the non-dysplastic subset of sessile serrated adenoma/polyp/lesion (SSA/P/L) and its borderline variant (BSSA/P/L). Given the observer variation in their histological typing, the identification of reliable markers that assist in the characterisation is warranted. Most important is the identification of polyp qualities that may reflect the patients’ risk of developing colorectal cancer. To address these issues, CD133 may represent a potential adjunct. Here we studied the discriminatory value of CD133 expression in the classification of ND-SPs and its distribution pattern in relation to synchronous colorectal carcinoma (SCRC). 39 SSA/P/Ls, 27 BSSA/P/Ls and 21 matched HPs were immunostained for CD133. The data were further correlated to the presence of SCRC and to polyp site and size. Ignoring SCRC status, CD133 was expressed more prominently in SSA/P/Ls than in HPs. The values for BSSA/P/Ls fell in between, yet closer to the SSA/P/L scorings. This observation was retained in the context of SCRC and for SSA/P/Ls not associated with SCRC. Right-sidedness and large size of the polyps more commonly associated with increased CD133 expression. CD133 expression was not a significant discriminator as to the SCRC status. BSSA/P/Ls are more closely aligned to SSA/P/L and further that SSA/P/L and BSSA/P/Ls fundamentally differ from HP by their CD133 immunoprofile, a notion that can be exploited in the diagnostic routine practice. Recorded data further indirectly support the idea that SSA/P/Ls are more prone to neoplastic progression than are HPs.

Statistics from


Non-dysplastic serrated polyps (ND-SPs) represent a heterogeneous group of polyps that comprise hyperplastic polyp (HP) and non-dysplastic sessile serrated adenomas,1 ,2 also known as sessile serrated polyps,3 ,4 and sessile serrated lesions5 (here referred to as sessile serrated adenoma/polyp/lesion, SSA/P/L).6 Although extensively studied,7 ,8 our knowledge on the differential diagnosis of ND-SPs and their proper classification is still incomplete. HPs and SSA/P/Ls share some histological and molecular features; yet, their cancer risk is probably different9 underlining the importance of reliable diagnostic criteria and tools of clinical relevance for patient management. The morphological characterisation is still based solely on general histological attributes, specifically structural qualities, which are often a daunting challenge for practicing pathologists,3 ,10 though structural aberration of SSA/P/L, to a lesser extent of BSSA/P/L, is more conspicuous than that of HPs.6 Even in research settings with the participation of experienced gastrointestinal pathologists, inter-observer variability is common.11 Our recent demonstration of a large group of ND-SPs6 that fulfilled neither the criteria for SSA/P/L1 ,2 nor those of traditional HP can at least in part explain the reported discrepancies. We proposed the preliminary name borderline sessile serrated adenoma/polyp/lesion (BSSA/P/L) for these equivocal lesions,6 which have been the focus of attention in only one other study.3

On this background a need for adjunctive diagnostic modalities arises. To this end, several antigens, including CK20, gastric mucin glycoproteins (MUCs), MLH1, Ki-67, ß-catenin, CDX2 and E-cadherin have previously been analysed, but with inconsistent results.1 ,4 ,12–17 Notably, this applies also to aberrant MUC6 expression, which in earlier reports has been considered a promising marker of SSA/P/L with 100% specificity,12 ,14 a notion most recently challenged by another study.18

Thus, a biomarker with a reliable discriminatory power in this context is currently lacking. Given the proposed role of CD133 (=human prominin-1) as an important potential marker of various stem cell populations,19–23 including colorectal tissue,24 combined with the perception that the capacity of such cells include characteristics of stemness such as driving the crypt-/cell renewal and contributing to the initiation and sustaining of cancer,20 ,24–26 we have here directed focus on this polypeptide. Fuelling this approach was our prior demonstration of CD133 expression in conventional colorectal adenomas.27 Notably is also the correlation, recently demonstrated, between CD133 expression and microsatellite instability-high status in colorectal carcinomas (CRC s),25 and the putative restriction of the CD133 epitope to adult cells that de-differentiate in the neoplastic transformation process according to most,23 though not all, observers28 as well as the established correlation between diverse SPs and MSI.29 Specifically, molecular alterations of SSA/P/L shared with those of MSI-H CRCs have been demonstrated.30 Additionally, an abundance of CD133 in tumour-initiating stem cells has been demonstrated.31 All these observations in combination provide an added rationale for studying CD133 in relation to ND-SPs.

Thus, the main objective of this study was to evaluate the utility of the CD133 immunoprofile in the differential diagnosis of SSA/P/L, BSSA/P/L and HP. Specifically, we hypothesised that CD133 immunostaining would be more prominent in SSA/P/Ls than in HPs and that BSSA/P/Ls occupied an intermediate position. We further hypothesised that those ND-SPs that were associated with synchronous colorectal carcinoma (SCRC) expressed CD133 more prominently than did the counterparts without this association.

Materials and methods

Study design

The base-line material, which comprised 219 cases of SSA/P/Ls, 206 cases of BSSA/P/Ls and 170 matched HPs, has been previously characterised.6 In brief, we carried out a review of 8342 consecutive cases of colorectal polyps, coded as HP, whereby samples fulfilling criteria of SSA/P/L (a convincing presence of at least two of the following four structural abnormalities: dilatation of base of crypts, serration of base of crypts, branched crypts and horizontally oriented crypts) and BSSA/P/L (only one of the above structural alterations was convincingly present or at least two of these qualities were considered equivocally present) were identified. Based on surgical pathology files and patients’ records, cases with SCRC (27 SSA/P/Ls, 17 BSSA/P/Ls and 11 HPs) were identified.32 Polyp site was considered right-sided if located proximally to the hepatic flexure.33 Polyp size was based on the gross pathology reports. Samples with cautery artefact of the entire biopsy and sections that did not represent optimally oriented specimens with the full length of at least 15 crypts could not be confidently assessed, thus contributing to the reduction of the number of study cases. Thereby, three SSA/P/Ls and one HP were excluded. In addition, 35 matched ND-SPs (15 SSA/P/L, 10 BSSA/P/L and 10 HPs), un-associated with SCRC, were included in the study.

Study groups

With selection criteria thus defined, the study material was reduced to 87 cases, categorised as SSA/P/L (n=39), BSSA/P/L (n=27) and HP (n=21).

The demographic characteristics of the patients and details on the index polyps, available in all cases, are given in table 1. Additionally, normal colorectal mucosa, sourced from biopsies of 11 subjects without clinicopathological evidence of inflammatory bowel disease or intestinal neoplasm was included as controls in the immunohistochemical studies.

Table 1

Demographic characteristics of patients and details on their ND-SPs


Immunohistochemical studies were performed on paraffin sections using antibody against CD133/1 antigen (clone AC133, Miltenyi Biotec cat. no. 130-090-422). The reactions were visualised using standard polymer technique. For negative control the primary antibody was omitted.

After dewaxing, antigen retrieval was performed by immersing slides in 10.0/0.5 mM Tris-EGTA buffer pH 9, and microwaved in a domestic microwave (four jars each 24 slides) oven for maximum power (900 W) for 10 min, followed by 20 min heating at medium power (400 W). Slides were subsequently kept in heating buffer (10.0/0.5 mM Tris-EGTA buffer pH 9) for 10 min to cool. After pretreatment, slides were incubated with the CD133 antibody (at 1 : 10 dilution), reactions were detected with Super Sensitive polymer-HRP reagent (Biogenex, cat. no. QD440-XAK) and visualised with DAB ImmPACT Substrate (Vector cat. No.SK-4105) according to the manufacturer's instructions. Finally, sections were counterstained with haematoxylin and mounted with Pertex.

Evaluation of CD133 immunoreactivity

The evaluation of the CD133 immunostained slides included a recording of the distribution pattern of the staining (ie, surface epithelium and upper crypts vs lower crypts/crypt bases and distorted crypts vs straight crypts), a recording of the subcellular localisation of the staining (ie, cell membrane vs cytoplasm), as well as a proportion scoring of the immunolabelled cells.

For scoring, both the extent and intensity of the CD133 immune positivity was semiquantitatively assessed.34 Thus, one of the following scores was assigned according to the proportion of crypts forming the polyp (ie, crypts within the bounds of the polyps, regardless of their structure) (for normal control specimens: all visualised crypts) that were lined by immunostained cells: 0, <5%; 1, 5%–25%; 2, >25%–50%; 3, >50%–75%; 4, >75%. For positive cases, the intensity of the immunoreactivity was subjectively scored on a 3-tiered scale, comprising weak (1+), moderate (2+) and strong (3+), according to the most intense labelling of the polyps.

The proportion and intensity scores were multiplied to obtain a combined score of 0–12, and accordingly the specimens of score 0–3, score 4–6 and score 8–12 were deemed negative, low score positive and high score positive, respectively.

Sections were scored by two observers: a pathologist with research interest in gastrointestinal topics (MM) and a senior lab technician (MB), with no particular knowledge of the histology of ND-SPs, both observers blinded to the demographics, polyp site and SCRC status. Specimens diversely scored, which occurred uncommonly and were of one grade, were re-evaluated jointly at a conference microscope to reach consensus. To further ensure reliability of the scoring, a limited number of sections were re-scored by one of the authors (MM). No significant difference between the readings was encountered.

Statistical analysis

For comparison purpose, analysis of categorical variables were performed in SAS V.9.2 using Fisher's exact test. All p values were based on two-sided test at a significance level of 5%.


The 87 polyps eligible for immunohistochemical analysis consistently showed negativity of the surface epithelium and of the upper crypts.

Two patterns of subcellular immunoreaction were noticed. Rare isolated cells, sited at the crypt bases, displayed a weak cytoplasmic reaction, demonstrated in the insets of figures 1 and 2. Such cells, possibly representing stem cells,35 were noticed in the study groups as well as in control specimens.

Figure 1

This non-dysplastic serrated polyp was categorised as a sessile serrated adenoma/polyp/lesion (unequivocal dilatation and serration of crypt bases as well as focal crypt branching6). The majority of the visualised crypts, structurally abnormal as well as normal, straight, are immunostained at the luminal edge of the cell membrane. Inset is a close-up of one of the crypt bases, demonstrating a dual immunopattern with weak cytoplasmic staining of a single cell (arrowed), in addition to the membrane decoration.

Figure 2

This hyperplastic polyp is immunonegative, safe for a single putative stem cell with cytoplasmic expression, visualised in the inset, which is a high power of the box area.

The other pattern featured CD133 chromogen precipitation largely confined to the luminal membrane of epithelial cells of the lower crypt compartment. In the following, reactivity only refers to this immunopattern. When present, the vast majority of the lining cells of a given crypt of SSA/P/Ls (figure 1) and BSSA/P/Ls (figure 3), including structurally abnormal crypts as well as intermingled straight crypts, were labelled. Staining generally characterised rows of cells (figures 1 and 3) rather than scattered single cells regardless of the crypt architecture. Additionally, the expression of CD133 was observed among structurally normal crypts in the immediate vicinity of immunopositive SSA/P/Ls and BSSAP/Ls, whereas crypts at a distance from such polyps were non-reactive as were biopsies of the normal colorectum. Normal crypts located near immunonegative SSA/P/Ls and BSSA/P/Ls were likewise negative.

Figure 3

This non-dysplastic serrated polyp was grouped as a borderline sessile serrated adenoma/polyp/lesion (equivocal dilatation and serration of crypt bases6). A weak to focally strong staining intensity characterises the two structurally altered crypts, as well as the straight crypt at the far left.

Ignoring the SCRC status, CD133 was expressed more prominently in SSA/P/Ls than in HPs. The values for BSSA/P/L fell in between, yet closer to the SSA/P/L scorings. Thus, whereas 20 (95%) studied HPs were categorised in the 0–3 score group, and none in the 8–12 score group, 15 (38.5%) of the SSA/P/Ls and 9 (33.3%) of the BSSA/P/Ls fell into the high score8–12 group and 11 (28.2%) and 15 (55.6%), respectively, fell into the low score category (table 2). Though the scores of SSA/P/L exceeded those of BSSA/P/L the difference did not reach significance.

Table 2

CD133 expression (0–3, 4–6, 8–12) of ND-SPs and of normal mucosa

Upon collapsing the expression score categories into binary variables to indicate CD133 negativity (score 0–3) versus CD133 positivity (score 4–12), the distribution of CD133 expression was retained (table 3), with significant difference between the values for SSA/P/L and HP as well as SSA/P/L and BSSA/P/L, regardless of the SCRC status and between the values for BSSA/P/L and HP when ignoring the SCRC status and in the context of SCRC. Although SCRC-associated ND-SPs more commonly expressed CD133 than did the corresponding polyp type not associated with SCRC, the differences fell short of significance. CD133 status of the diverse polyp subtypes in relation to site and size is given in tables 4 and 5.

Table 3

CD133 expression (score 4–12) related to ND-SP subtype and SCRC status

Table 4

CD133 expression (score 4–12) stratified by ND-SP polyp site

Table 5

CD133 expression (score 4–12) stratified by ND-SP size

As to anatomical distribution, avoiding partitioning for SCRC status, right-sided SSA/P/Ls were more commonly CD133 positive than were the left-sided counterparts (p=0.002). This association was maintained when grouping all the ND-SPs (p<0.001), and for SSA/P/L in the context of SCRC (p=0.003). A similar trend was not attained for the BSSA/P/Ls or for SSA/P/L in the absence of SCRC.

Regarding polyp size, constellations that provided significant differences included SSA/P/L in the context of SCRC (p=0.027) as well as all ND-SPs (p<0.001). A trend towards higher score of the large polyp was, however, also noticed for SSA/P/L, when ignoring the SCRC status (p=0.057). For BSSA/P/L polyp size did not influence CD133 scoring.


The diagnosis of the ND-SPs is currently based on the interpretation of the H&E stained slides, largely considering the crypt aberration, which by some is referred to as architectural dysplasia2 rather than cytological attributes. This diagnostic process is, however, known to be subjective with considerable observer variations.5 ,10 ,11 Given the clinical importance of characterising and segregating these polyps, the identification of markers that may assist in the ND-SP subtyping is warranted.

As a possible contributor in this differential diagnostic process, we here studied the distribution pattern of CD133, considered an important potential cancer stem cell marker of the colorectum by several investigators, capable of initiating and driving tumour growth.36–41

Whereas growing evidence suggests that CD133 is a marker (the molecule of the moment)42 of putative cancer stem cells of the colorectum,36 ,37 and upregulation of CD133 expression is known to play a role in tumourigenesis of CRC,43 information regarding the expression pattern of CD133 in premalignant colorectal lesions is sparse. Some of us have previously noticed consistent CD133 expression of the adenomatous epithelium and further a positive association between the CD133 staining intensity and the degree of dysplasia.27 These data on colorectal adenomas have found experimental support by the demonstration of frequent neoplastic transformation of CD133-marked mouse intestinal stem cells.38

This study is the first to address the expression of CD133 in ND-SPs, encompassing other subsets of presumed CRC precursor lesions. Here we showed that SSA/P/Ls and BSSA/P/Ls express CD133 significantly more often than do HPs. This observation corroborates our previous experience that several qualities of BSSA/P/Ls are more akin to SSA/P/Ls than to HPs.6 ,32 Specifically, we have demonstrated that the occurrence of BRAF mutation of BSSA/P/Ls did not differ from that of SSA/P/Ls and significantly exceeded that of HPs.32

The presence of CD133 in the base of altered crypts in SSA/P/Ls and BSSA/P/Ls is noteworthy, considering the alleged mature nature of these cells including goblet cells44 ,45 emphasised by several investigators.4 ,9 This issue needs some elaboration. Given the notion that CD133 is a marker of immature cells46 and cells with properties of self-renewal and of high tumourigenic potential,36 ,37 it may be expected that these basally sited cells, despite apparent maturity as judged from H&E stained sections, deviate from the fully developed, CD133-negative cells that populate the superficial compartments, in concert with the previously reported down regulation of CD133 upon differentiation of epithelial cells.23

The CD133 expression of SSA/P/Ls, to a lesser extent that of BSSA/P/Ls, as opposed to the largely negative HPs, could further support an abnormal nature of the lesional cells in the former polyps, despite their apparent cytologically non-dysplastic appearance, an immunoprofile that is further pronounced in the SSA/P/L-SCRC context. This close association of CD133 labelling with structurally altered crypt bases indicates that CD133 may be a putative marker of abnormal proliferation and/or dysmaturation.1 ,12 ,47 ,48 That being said, structurally normal crypts in close spatial relation to such altered crypts did, however, share the CD133 immunoexpression. Whether the latter observation is akin to transitional mucosa49 trapped within the polyp or rather indicate a genuine abnormality of these straight crypts, in reality forming part of the polyp per se, not identifiable in routine stains, is uncertain. Be this as it may, the overt difference in the CD133 immunoprofile of the HP and the normal mucosa versus that of SSA/P/Ls and BSSA/P/Ls is remarkable and supports a closer relationship of BSSA/P/Ls to SSA/P/Ls than to HPs, an issue that we have previously investigated.6 Intuitively, our observations may further suggest that HP is not a strong player in the serrated neoplasia pathway; in other words, the observations do not provide positive evidence that HP precedes the progression of SSA/P/L in the serrated pathway, the end product of which includes the serrated carcinoma. Accordingly, the role of HP in the serrated pathway has recently been described as debatable.8

Though CD133 positive SSA/P/Ls were more commonly SCRC-related than the non-SCRC counterparts, significant differences were not obtained, possibly a consequence of the small overall sample size.

The CD133 expression of SSA/P/L-SCRCs as a function of the subanatomical site, demonstrated here, parallels the differences previously reported in the aberrant MUC6 expression between right- and left -sided SSA/P/Ls,12 ,14 ,18 contributing to the concept of biological differences in different compartments of the colorectum.49 ,50 A correlation of CD133 expression to polyp size most prominently characterised the SSA/P/L-SCRC setting, with large-sized specimens being more likely to express CD133, though this correlation likewise applied to the studied ND-SP as a group. These observations are noticeable given the increased risk of important colorectal lesions, including CRC, in the setting of large and proximal SPs reported by several observers.51–53

The upregulation of CD133 in both SSA/P/Ls and BSSA/P/Ls, demonstrated herein and in conventional adenomas, previously recorded,27 ,54 is intriguing and may suggest some shared elements in the pathogenesis of the serrated neoplasia pathway and the traditional adenoma–carcinoma sequence. Features common of both pathways may include events important for stem cell maintenance, such as Wnt/β-catenin signalling.51 In this token, the aberrant nuclear β-catenin accumulation in colorectal adenoma as well as in SSA/P/L5 ,13 ,16 is noteworthy, as is the previously proposed role of Wnt signalling, implying disrupted β-catenin, in SSA/P/L.5 Whether these observations suffice to support the concept of the previously proposed ‘fusion pathway’8 is, however, uncertain and remains an issue for further studies.

In summary, our study shows that the CD133 immunoprofile is a useful marker for discriminating between diverse ND-SPs, conspicuous expression being highly suggestive of SSA/P/L and BSSA/P/L at variance with HP, which generally was negative or barely weakly stained.

Although CD133 staining of ND-SPs did not come out as a strong reliable predictor of synchronous colorectal cancer, its expression, preferably in conjunction with other polyp qualities (large size, proximal site), may contribute to define a subgroup of ND-SPs that may justify a closer patient surveillance. Large-scale studies on sizeable cohorts of patients are required to validate the current findings as are the additional analysis of other potential stem cell markers in this context.55

Take-home messages

  • The stem cell marker CD133 is expressed more prominently in sessile serrated adenoma/polyp/lesion (SSA/P/L) than in hyperplastic polyp.

  • CD133 expression of borderline SSA/P/L falls in between, yet closer to the SSA/P/L-values.

  • Right-sidedness and large size of polyps are more often associated with CD133 expression.

  • CD133 expression is not a reliable marker of synchronous colorectal carcinoma.


We thank the study group team, including the technicians of the Pathology Department of Hospital South, Naestved, for their collaboration in this study and Steen Ladelund for his help in statistical analysis.

View Abstract


  • Contributors MM designed the study, assessed immunostained slides, analysed data and formulated the paper, MB assessed immunostained slides, JHB and HJN approved the final paper, SH analysed data and formulated the paper.

  • Funding The research supported by the local research unit at the hospital.

  • Competing interests None.

  • Ethics approval The Regional Ethics Committee approved the project.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.