Biomarkers for the identification of precursor polyps of colorectal serrated adenocarcinomas

Abstract

Background

In contrast to conventional colorectal carcinomas (CCs), which develop through a so-called chromosome instability or suppressor phenotype pathway, the sequence of events leading from precursor polyps/adenomas to serrated adenocarcinomas (SACs), which are more aggressive and exhibit a poorer survival than CCs, is as yet not clearly defined. Here, we aimed at detecting protein and DNA biomarkers for SAC in a series of primary colorectal polyps.

Methods

In total 303 colorectal polyps were included: 121 serrated polyps (33 hyperplastic polyps, 37 sessile serrated adenomas (SSA), 51 traditional serrated adenomas (TSA)), 143 conventional polyps (72 tubular polyps, 34 tubulovillous polyps, 37 villious adenomas), and 39 bi-phenotypic serrated-conventional polyps. The protein biomarkers tested were deduced from previously published SAC and CC expression profiling studies. A representative subset of 106 polyps was selected for DNA biomarker analyses, i.e., proto-oncogene mutation and microsatellite instability (MSI) status. In order to confer proper weight to each biomarker, a multivariate logistic regression model was employed.

Results

We found that serrated and conventional polyps differed in most of the SAC biomarkers tested. Of these biomarkers, FSCN1 showed the largest difference in expression (p = 0.0001). Despite sharing a serrated morphology, we found that SSAs and TSAs differed considerably with respect to anatomical location, expression of EPHB2 and PTCH1, presence of the V600E BRAF mutation and MSI status. Logistic regression analysis revealed that SSA was the polyp type that shared most biomarkers with SAC.

Conclusion

Based on the shared presence of protein and molecular biomarkers, especially FSCN1 expression, SSA may serve as a precursor lesion of SAC. Biomarker assessment may help in discerning colorectal carcinogenic routes with distinct prognostic implications.

Electronic supplementary material

The online version of this article (doi:10.1007/s13402-016-0269-5) contains supplementary material, which is available to authorized users.

Introduction

Awareness of the sequences of events leading to different types of colorectal carcinomas (CRCs) is not only imperative for an in-depth knowledge of their biology but, even more importantly, for an appropriate clinical follow-up of patients with polyps. Approximately 80 % of the CRCs are conventional carcinomas (CCs), which develop through the so-called chromosome instability or suppressor phenotype pathway. CC precursor lesions encompass three major histological categories of conventional polyps (CPs), i.e., tubular adenomas (TAs), tubulovillous adenomas (TVAs) and villous adenomas (VAs), characterized by microsatellite stability (MSS) and activating KRAS mutations, respectively [1]. In contrast, the molecular pathways and precursor lesions characterizing the remaining 20 % of CRCs have not been fully identified. Serrated adenocarcinoma (SAC) has recently been recognized by the World Health Organization (WHO) as a subtype accounting for 7.5–10 % of CRCs. It has been reported that SAC is more aggressive and exhibits a poorer survival than CC [1–4]. Through comparing the molecular signatures of SAC and CC, two independent studies have identified FSCN1, HPCA and HIF1α as being up-regulated, and EPHB2 and PTCH1 as being down-regulated in SAC compared to CC [5, 6].

Based on its serrated morphology and the identification of remnants of serrated polyps (SP) adjacent to SAC in surgical specimens, it has been assumed that SAC may be an end-point of a pathological sequence initiated by SP, which includes hyperplasia (HP), sessile serrated adenoma (SSA) and traditional serrated adenoma (TSA) [3, 7]. As such, SAC markers identified through microarray-based expression profiling may offer the opportunity to trace down these markers along the conventional and serrated polyp pathways.

The aims of the present study were to (i) report the clinico-pathological and molecular features of an unpublished series of 303 colorectal polyps comprising SP and CP, (ii) evaluate by immunohistochemistry the expression of biomarkers that may be useful for the distinction between SP and CP and, based on the results from a logistic regression model, (iii) identify the type of polyp with the greatest similarity to SAC based on both protein and DNA biomarkers, thereby allowing these biomarkers to be used for the identification of putative SAC precursor lesions.

Methods

Study cases

In total, 303 polyps were included, comprising 121 SPs (33 HP, 37 SSA and 51 TSA cases) and 143 CPs (37 VA, 72 TA and 34 TVA cases). We also included a set of 39 serrated-conventional polyps (SCPs), or considered by others as mixed adenomas, showing a bi-phenotypic morphology, i.e., showing both serrated and conventional features in more than 20 % of the histological sections. The diagnoses of serrated polyps and SCPs were assessed following the UK guidelines for reporting these lesions [8], whereas conventional or classical polyps were diagnosed according to the World Health Organization (WHO) classification guidelines of colorectal tumors [1]. The cases included were consecutively obtained for each category, i.e., SP, CP and SCP. In order to avoid potential bias, only typical serrated and conventional polyps were selected, whereas those showing any grade of dysplasia were excluded.

The study cases were selected by two pathologists (JGS, MPG) and comprised 236 cases obtained from endoscopic resections and 67 cases synchronously present in surgical specimens taken from a previously described series of SACs and matched CCs [3]. Polyps located up to the splenic flexure were considered as proximal. The study was approved by the Hospital Ethics Committee and was carried out in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. Written informed consent was obtained from all patients. In order to measure the predictive capacity of each immunohistochemical biomarker, a previously reported series of 98 SAC and 92 CC cases [6] was stained for FSCN1, HPCA, HIF1α, PTCH1 and EPHB2. In addition, several molecular biomarkers (i.e., KRAS, BRAF and PIK3CA mutations and MSI status) were included (see below).

Immunohistochemistry

After performing immunohistochemical staining of whole tissue sections from 44 polyps, comprising 6 HP, 7 SSA, 7 TSA, 8 TA, 6 TVA, 6 VA and 4 SCP cases, to assess immunohistochemical heterogeneity, either conventional or serrated representative areas of the remaining 259 polyps were selected by one of the pathologists (JGS) for the construction of a 1.5 mm core tissue microarray (TMA) as previously described [9]. Details of the manufacturers, codes, antigen retrieval conditions, dilutions, procedures and controls are provided in the Supplementary information. FSCN1 showed a diffuse cytoplasmic staining (Fig. 1a-c), whereas HPCA showed a granular cytoplasmic staining at the luminal border (Fig. 1d-e). EPHB2, PTCH1 (Fig. 1f-g and j-l) and HIF1α (Fig. 1h-i) showed diffuse cytoplasmic staining patterns. For FSCN1, fibroblasts and endothelial cells were taken as internal positive controls. For HPCA, Purkinje cells from cerebellum sections were taken as positive controls. For both latter markers, the normal colorectal mucosa was taken as a negative control. For EPHB2 and PTCH1, expression in normal basal cells from the intestinal crypt was taken as a positive internal control, whereas for HIF1α expression in the upper third of the crypt was taken as a positive control as previously described [10, 11].

Fig. 1.

Representative micrographs of the immunohistochemical staining. (a), (b) and (c): FSCN1 staining in SSA ((a), ×10), TA ((b), note the internal control fibroblast and endothelial cell in the stroma, ×20) and in SCP ((c), note staining only in the serrated component, ×5); (d) and (e): HPCA granular cytoplasmic staining in TSA ((d), ×20) and in TA (e), ×20); (f) and (g): EPHB2 staining in SSA ((f), note more intense expression in the lower third, ×10) and in VA ((g), ×5); (h) and (i): HIF-1α staining in SSA ((h), note focal expression in nucleus and cytoplasm, ×10) and in TA ((i), intense cytoplasmic expression, ×20); (j), (k) and (l): PTCH11 staining in SSA ((j), no expression, ×10), in TSA ((k), intense expression, ×10) and in TA ((i), absence of staining, ×10)

The staining scores for these markers were calculated by multiplying the staining intensity score (0 = no staining, 1 = weak staining, 2 = moderate to strong staining) in a given tumor area by the stained area score (1 < one-third, 2 = between one- and two-thirds, 3 > two-thirds) as described elsewhere [11]. Immunohistochemical procedures and evaluations were performed as reported before [12]. The total staining score (0–6) was considered positive when ≥ 4. In the tissue microarray analyses, the highest of the two cores was chosen.

DNA extraction and mutation analysis

Based on sufficient tissue availability, a subset of 106 representative polyps, comprising 17 HP, 17 SSA, 13 TSA, 3 SCP, 12 VA, 29 TA and 15 TVA cases, was selected for gene mutation and MSI status analyses as reported before [12] and outlined in the Supplementary information.

Microsatellite instability assay

MSI was evaluated according to the manufacturer’s instructions using a MSI Analysis System kit, version 1.2 (Promega, Madison, USA) as described before [12] and outlined in the Supplementary information. The cases were categorized as high-level microsatellite instable (MSI-H), microsatellite stable (MSS) or low-level MSI (MSI-L) according to the NIH criteria [13].

Statistical analyses

Statistical analyses were performed using SPSS (Version 15.0, Chicago, IL, USA) packages. The Pearson χ2 test was used for assessing statistical significant associations between categorical variables, except for the oncogene mutation analyses for which, due to a smaller sample size, the Fisher’s exact model test was considered more appropriate. For continuous variables (age and polyp size), the Student t test was used. A multivariate analysis through logistic regression was carried out to identify independent predictors distinguishing CC from SAC. We developed a prognostic score, giving points to each independent variable according to its regression β, rounding off to the nearest integer. A risk score was assigned to each case by adding up points for each independent variable.

Results

Clinico-pathological characteristics of the polyps

The clinico-pathological features of the polyps included in this study are listed in Table 1, and those of the CRCs are listed in Table 2. We found that, in general, SPs were diagnosed at a younger age than CPs (63.7 SD ± 12.8 vs. 67.5 SD ± 11.2; p = 0.011) and more frequently occurred in females than in males (43.8 % vs. 30.8 %; p = 0.029). No significant differences were found between the proximal or distal locations of these two polyp types (25.6 % vs. 20.3 %; p = 0.302). Regarding the SP spectrum, it was found that, compared to TSAs, SSAs significantly more often occurred in the proximal colon (43.2 % vs. 17.6 %; p = 0.009), whereas this difference was not observed between HPs and TSAs (18.2 % vs. 17.6 %; p = 0.950). Among CPs, TAs occurred more frequently than VAs in the proximal colon (27.8 % vs. 10.8 %; p = 0.046). With respect to size, HPs were found to be significantly smaller than TSAs (0.8 SD ± 0.8 vs 1.7 SD ± 1.2; p < 0.0001) and likewise for CPs, TAs were found to be smaller than TVAs (1.1 SD ± 0.9 vs. 1.6 SD ± 1.4; p = 0.029) and VAs (1.1 SD ± 0.9 vs. 2.3 SD ± 1.8; p < 0.0001). HPs were generally diagnosed at a younger age and were found to be smaller than TSAs, although these two polyp types did not differ in their anatomical locations as did SSAs and TSAs.

Table 1.

Main clinico-pathological features of the polyps

		HP	SSA	TSA	All SPs	SCP	VA	TA	TVA	All CPs	Total polyps
N		33	37	51	121	39	37	72	34	143	303
Age	Mean [±SD]	59.9 [12.5]	61.6 [12.3]	67.7 [11.1]	63.7 [12.8]	64.3 [10.8]	68.6 [11.3]	67.3 [11.2]	66.7 [11.6]	67.5 [11.2]	65.6 [11.7]
Gender	Female	16 (48.5)	15 (40.5)	22 (43.1)	53 (43.8)	16 (41.0)	10 (27.0)	22 (30.6)	12 (35.3)	44 (30.8)	113 (37.3)
Location	Distal	27 (81.8)	21 (56.8)	42 (82,4)	90 (74.9)	34 (87.2)	33 (89.2)	52 (72.2)	29 (85.3)	114 (79.7)	238 (78.5)
	Proximal	6 (18.2)	16 (43.2)	9 (17.6)	31 (25.6)	5 (12.8)	4 (10.8)	20 (27.8)	5 (14.7)	29 (20.3)	65 (21.5)
Size	Mean cm [±SD]	0.8 [0.8]	0.9 [0.5]	1.7 [1.2]	1.2 [1.1]	1.6 [1.2]	2.3 [1.8]	1.1 [0.9]	1.6 [1.4]	1.5 [1.4]	14 [1.3]
FSCN1	Positive	21 (63.6)	28 (75.7)	32 (62.7)	81 (66.9)	24 (61.5)	29 (78.4)	20 (27.8)	12 (35.3)	61 (42.7)	166 (54.8)
HPCA	Positive	20 (60.6)	28 (75.7)	34 (66.7)	82 (67.8)	31 (79.5)	20 (54.1)	50 (69.4)	20 (58.8)	90 (62.9)	203 (67.0)
HIF1alpha	Positive	24 (72.7)	29 (78.4)	39 (76.5)	92 (76.0)	38 (97.4)	28 (75.7)	64 (88.9)	31 (91.2)	123 (86.0)	253 (84.5)
PTCH1	Positive	11 (33.3)	17 (45.9)	36 (70.6)	64 (52.9)	26 (66.7)	15 (40.5)	24 (33.3)	20 (58.8)	59 (41.3)	149 (49.2)
EPHB2	Positive	23 (69.7)	31 (83.8)	29 (56.9)	83 (68.6)	30 (76.9)	29 (78.4)	56 (77.8)	29 (85.3)	114 (79.7)	227 (74.9)

HP Hyperplastic polyp; SSA Sessile serrated adenoma; TSA Traditional serrated adenoma; SPs Serrated polyps; SCP serrated conventional polyp; VA Villous adenoma; TA Tubular adenoma; TVA Tubulovillous adenoma; CPs Conventional polyps; SD Standard deviation

Table 2.

Immunohistochemical and molecular markers discriminating serrated from conventional carcinomas

		Serrated carcinoma	Conventional carcinoma	p value
n		98	92
Age	Mean [±SD]	69 [10]	68.5 [12.8]	0.764
Gender	Female (%)	44 (44.9)	47 (51.1)	0.394
Location	Distal	45 (45.9)	43 (46.7)
	Proximal	53 (54.1)	49 (53.3)	0.910
TNM	T1	4 (4.1)	5 (5.4)
	T2	17 (17.3)	8 (8.7)
	T3	44 (44.9)	56 (60.9)
	T4	33 (33.7)	23 (25.0)	0.094
	N0	41 (41.8)	44 (47.8)
	N1	31 (31.6)	28 (30.4)
	N2	23 (23.5)	20 (21.7)
	Nx	3 (3.1)	0	0.350
	M1	6 (6.1)	0
	Mx	92 (93.9)	92 (100)	0.050
FSCN1	Positive	85 (86.7)	15 (16.3)	<0.0001
HPCA	Positive	72 (73.4)	33 (35.9)	<0.0001
HIF1alpha	Positive	81 (82.7)	25 (27.2)	<0.0001
PTCH1	Positive	25 (25.5)	76 (82.6)	<0.0001
EPHB2	Positive	34 (34.7)	74 (80.4)	<0.0001
KRAS	Mutated	44 (44.9)	31 (33.7)	0.076
BRAF	Mutated	23 (23.5)	2 (2.2)	<0.0001
PIK3CA	Mutated	10 (10.2)	8 (8.7)	0.458
MSI status	Instable	11 (11.2)	3 (3.3)	0.032

SD Standard deviation; MSI Microsatellite instability

Significant associations between SAC protein biomarkers and polyp types and sizes

No significant differences were observed between the protein expression patterns of the polyps evaluated on the TMA and those included as whole tissue sections, nor between the endoscopic and cancer-associated surgically removed polyps (data not shown). In addition, we found that a positive expression of FSCN1 was significantly associated with SPs when compared to CPs (66.9 % vs. 42.7 %; p = 0.0001). We also found that HIF1α positivity occurred slightly less frequent in SPs than in CPs (76.0 % vs. 86.0 %; p = 0.038), similar to that observed for EPHB2 (68.6 % vs. 79.7 %; p = 0.039). Among SPs, the major differences in positive protein expression were found between SSAs and TSAs (45.9 % vs. 70.6 %; p = 0.020) and between HPs and TSAs (33.3 % vs. 70.6 %; p = 0.0008) for PTCH1, and between SSAs and TSAs for EPHB2 (83.8 % vs. 56.9 %; p = 0.007). Significant differences were found between VAs and TAs (78.4 % vs. 27.8 %; p < 0.0001) and between TVAs and VAs for FSCN1 positivity (35.3 % vs. 78.4 %; p = 0.0002), and between TAs and TVAs for PTCH1 positivity (33.3 % vs. 58.8 %; p = 0.013). With respect to polyp size, PTCH1 positivity was found to be associated with those >1 cm vs. those <1 cm (108/183: 57.9 % vs. 45/123: 36.6 %; p = 0.0001) and with a proximal vs. distal (21/44: 32.3 % vs.132/238: 55.5 %; p = 0.0009) location.

Significant associations between SAC DNA biomarkers and polyp types and sizes

The characteristics of the polyp series in terms of oncogene mutations and MSI status are listed in Table 3. We found that the V600E BRAF mutation and the MSI-H status were significantly associated with polyps having a serrated morphology (57.4 % vs. 1.8 %; p < 0.0001 and 61.7 % vs. 1.8 %; p < 0.0001). The V600E BRAF mutation was found more frequently in SSAs than in TSAs (82.4 % vs. 30.8 %; p = 0.010). Similarly, SSAs exhibited MSI-H more frequently than TSAs (82.4 % vs. 34.5 %; p = 0.018). With respect to the various types of CP, a significant difference was observed between TAs and TVAs, i.e., the former were more often KRAS mutation-negative than the latter (89.7 % vs. 60 %; p = 0.030). Regarding polyp size, a higher percentage of KRAS mutated polyps was found to be >1 cm (18/56: 32.1 % vs. 8/50:16 %; p = 0.043), while the V600E BRAF mutation was more frequently detected in those <1 cm in size (18/50: 36 % vs. 11/56:19.6 %; p = 0.048). No additional significant associations were found through the other comparisons made.

Table 3.

Oncogene and MSI status in the series of 106 polyps

		HP	SSA	TSA	All SP	SCP	VA	TA	TVA	All CP
N		17	17	13	47	3	12	29	15	56
KRAS	Mutated	4 (23.5)	2 (11.8)	5 (38.5)	11 (23.4)	2 (66.7)	4 (33.3)	3 (10.3)	6 (40.0)	13 (23.2)
	Wild-type	13 (76.4)	15 (88.2)	8 (61.5)	36 (76.6)	1 (33.)	8 (66.7)	26 (89.7)	9 (60.0)	43 (76.8)
BRAF	Mutated (V600E)	9 (52.9)	14 (82.4)	4 (30.8)	27 (57.4)	0	0	1 (3.4)	0	1 (1.8)
	Mutated (Other)	2 (11.8)	0	1 (7.7)	3 (6.4)	0	1 (8.3)	1 (3.4)	2 (13.3)	4 (7.1)
	Wild-type	6 (35.3)	3 (17.7)	8 (61.5)	17 (36.2)	3 (100)	11 (91.7)	27 (93.1)	13 (86.7)	51 (91.1)
PIK3CA	Mutated	5 (29.4)	4 (23.5)	3 (23.1)	12 (25.5)	0	2 (16.7)	7 (24.1)	1 (6.7)	10 (17.9)
	Wild-type	12 (70.6)	13 (76.5)	10 (76.9)	35 (74.5)	3 (100)	10 (83.3)	22 (75.9)	14 (93.3)	46 (82.1)
MSI status	MSI-H	10 (58.8)	14 (82.4)	5 (38.5)	29 (61.7)	0	0	1 (3.4)	0	1 (1.8)
	MSS/MSI-L	7 (41.2)	3 (17.6)	8 (61.5)	18 (38.3)	3 (100)	12 (100)	28 (96.6)	15 (100)	55 (98.2)

HP Hyperplastic polyp; SSA Sessile serrated adenoma; TSA Traditional serrated adenoma; SP Serrated polyp; SCP polyp; VA Villous adenoma; TA Tubular adenoma; TVA Tubulovillous adenoma; CP Conventional polyp; MSI Microsatellite instability; MSI-H High level MSI; MSI-L Low level MSI; MSS Microsatellite stable

Logistic regression analysis indicates that SPs show high similarities to SACs

The results of the multivariate regression logistic analysis and the scoring for each protein biomarker according its capacity to distinguish SAC from CC based on β value are shown in Table 4. Low values indicate a higher prediction for SAC, whereas higher values point at a CC diagnosis. The protein biomarkers with scores ≤ −3 were selected as SAC and those with scores ≥ 0 as CC, whereas for the DNA biomarkers the cut-off values were ≤ 1 and ≥ 3, respectively. Based on these analyses, FSCN1 was identified as the strongest SAC marker, whereas PTCH1 was identified as such for CC (β = −4.048 and 2.044, respectively). When this score was applied to each of the SAC and CC cases, scores ≤ −3 for SAC and ≥ 0 for CC yielded good diagnostic predictions (Fig. 2). When the same score was applied to the polyps it was found that, compared to the CP scores, most SP scores were ≤ −3 (61.2 % vs. 45.4 %), whereas the percentage of SPs with ≥ 0 scores was lower than that of the CPs (19.8 % vs. 29.4 %; p = 0.024) (Table 5).

Table 4.

Multivariate analysis for each immunohistochemical marker and its independent predictive value and scoring according to ß values

Marker	ß	OR (IC95%)	p value	Points
FSCN1	−4.048	0.02 (0.004–0.07)	<0.001	−4
HIF1α	−3.031	0.05 (0.01–0.20)	<0.001	−3
HPCA	−1.166	0.31 (0.09–1.04)	0.057	−1
EPHB2	1.540	4.66 (1.17–18.53)	0.03	2
PTCH1	2.044	7.72 (2.15–27.71)	0.002	2
KRAS wild-type	0.843	2.32 (1.24–4.34)	0.008	1
BRAF wild-type	2.925	18.63 (4.16–83.55)	<0.001	3

Fig. 2.

Predictive scores obtained for serrated carcinomas and conventional carcinomas according to immunohistochemical biomarker assessments

Table 5.

Distribution of serrated, conventional and SCP polyps according to the points obtained after applying the immunohistochemical predictive score

Points	Serrated polyps n = 121	Conventional polyps n = 143	SCP n = 39
≤ − 3	74 (61.2)	65 (45.4)	24 (62 %)
−2, −1	23 (19.0)	36 (25.2)	8 (21 %)
≥0	24 (19.8)	42 (29.4)	7 (18 %)

In Table 6 the results of the analyses carried out on the different types of SP and CP are listed. The lowest score (i.e., higher marker similarity with SAC) was obtained for SSA, of which 70 % showed a score ≤ −3. Interestingly, SAA was followed by VA, HP and TSA (57 %). In contrast, TA and TVA were the two polyp types with a higher similarity to CC.

Table 6.

Distribution of the different serrated and conventional polyp types according to the points obtained after applying the immunohistochemical β predictive score

Points	HP n = 33	SSA n = 37	TSA n = 51	TA n = 72	TVA n = 34	VA n = 37
≤ − 3	19 (58 %)	26 (70 %)	29 (57 %)	28 (39 %)	13 (38 %)	24 (65 %)
−1;-2	8 (24 %)	6 (16 %)	9 (18 %)	23 (32 %)	6 (18 %)	7 (19 %)
≥0	6 (18 %)	5 (14 %)	13 (25 %)	21 (29 %)	15 (44 %)	6 (16 %)

HP Hyperplastic polyp; SSA Sessile serrated adenoma; TSA Traditional serrated adenoma; TA Tubular adenoma; TVA Tubulovillous adenoma; VA Villous adenoma

After multivariate analysis of the various DNA biomarkers tested, only the KRAS and BRAF mutation statuses were included in the final model (Table 2). In order to present the results as positive value points (similar to the immunohistochemical analysis), KRAS and BRAF wild-type statuses were considered. In agreement with the immunohistochemical analyses, SP showed a higher similarity to SAC than CP in terms of oncogene mutation status (points ≤ 1: 66.1 % vs. 8.9 %; p < 0.0001) (Supplementary Table S2). Specifically, the highest similarity was observed for SSA followed by HP (points ≤ 1: 82 % and 70.9 %, respectively), whereas those of TSA (39 %) and VA (8.3 %) were found to be considerably lower. Furthermore, as observed in the protein biomarker analysis, the polyp type sharing the least oncogene mutation biomarkers with SAC was TA (4 points: 83 %) (Supplementary Table S3).

Bi-phenotypic SCP characteristics

The bi-phenotypic SCPs studied were mostly large and located in the distal colon. We found that they expressed, to a great extent, HIF1α (97.4 %) and HPCA (79.5 %) as serrated pathway markers (Table 1). In addition, we found that the cases that were evaluated molecularly were all BRAF wild-type, microsatellite stable and, in 60 % of the cases, KRAS mutated (Table 3). The immunohistochemical predictive scores showed a high grade of similarity between SCPs and SACs (≤ −3: 62 %) (Table S2).

Discussion

Defining the malignant potential of a CP is of paramount importance for an appropriate clinical surveillance of patients after polypectomy [14, 15]. Recently, new clinical management guidelines for serrated lesions were published [16]. However, given the morphological and molecular heterogeneity of SPs and the different types of CRC that may serve as end-points of the serrated pathway [17], it is imperative to identify SAC precursor lesion biomarkers. Here, we evaluated several immunohistochemical biomarkers, which were deduced from previous microarray-based expression profiling studies aimed at comparing SAC and CC. To the best of our knowledge, this is the largest study assessing SAC biomarkers in colorectal polyps (CPs) and the first using logistic regression to identify the possible precursor lesion of SAC. Since these biomarkers are acquired during the carcinogenic process, it is anticipated that some of them may already be present in the precursor lesions and, thus, may serve as tracers of the pathological sequence of events. Here, we established the relative importance of these biomarkers based on their capacity to distinguish SAC from CC. Through our approach putative gains or losses of these markers during the progression of the neoplastic lesions cannot be excluded, which is a limitation of the current study. Another possible limitation of this study may be the fact that we also included cancer-associated polyps next to endoscopic polyps. Obviously, most SPs were found in the SAC specimens compared to the CP and CC specimens. However, no significant differences were observed between these two groups, thus suggesting no relevant bias. Statistical analyses based at considering SP and CP as two distinct groups were aimed at discerning to what extent these two groups could have their own biomarker profiles, despite the fact that each of them comprised different types of histology-based lesions. For that purpose balanced proportions of SPs and CPs were included in this study, although CPs tend to be much more frequent than SPs as endoscopic findings.

Among the SPs, TSAs were larger and diagnosed at an older age than HPs, although they did not differ in anatomical location. These findings suggest that TSAs may develop from pre-existing HPs as previously proposed by Kim et al. [15]. In contrast, SSAs and TSAs showed differences in anatomical location, as well as in BRAF and KRAS mutations and MSI status.

From the immunohistochemical analyses of the SPs and CPs, we conclude that the expression patterns of FSCN1 and EPHB2 follow the same trends as those observed in SACs and CCs, whereas the opposite was found for HIF1α, which was slightly more often expressed in CPs than in SPs. Finally, no significant associations were found between HPCA or PTCH1 positivity and histological types. However, within the SPs, PTCH1 was higher expressed in TSAs than in SSAs. Parfitt and Driman [18] measured the expression of the survivin and hedgehog proteins in 10 TSAs and 48 SSAs and found a higher expression of sonic hedgehog, a ligand of PTCH1, in the TSAs compared to the SSAs, although the expression of PTCH1, which was also measured, did not show any difference between these two polyp types. Our results are in agreement with those recently reported by Morimoto et al. [10] on a series of 105 CPs in which a higher expression of PTCH1 in the TSAs compared to the SSAs was found.

Guo et al. [19] reported a reduction in EPHB2 expression in crypt bases developing from normal mucosa to adenoma, and from primary to metastatic CRC. Among the SPs, we found that EPHB2 was significantly more often expressed in the SSAs (83.8 %), thus suggesting that EPHB2 expression loss in SSAs could be an early event in the transformation to SAC, where EPHB2 staining was seen infrequently in our series (34.7 %) as well as in that of Laiho et al. [5]. Therefore, our findings are in accordance with the two reported serrated pathways [17, 20].

FSCN1 is an actin-bundling protein of which the expression is related to a poor prognosis in CRC patients [6]. Recently, FSCN1 over-expression was reported to be associated with the development of colonic adenocarcinoma cells from non-tumorigenic human colonic adenoma cells [21, 22]. This observation supports our finding that FSCN1 serves as a protein biomarker that is associated with the progression from adenomas to carcinomas. Overall, SPs show a higher FSCN1 expression than CPs, which provides further molecular support to the rapid malignant transformation rate of SPs [23].

The oncogene mutation and MSI data obtained here support previously reported observations, i.e., a predominance of BRAF mutations among SPs, especially SSAs, and that of KRAS mutations in the conventional type [17, 24]. The associations observed between oncogene mutations and polyp sizes suggests that BRAF mutations occur during early polyp stages and that KRAS mutations occur during more advanced polyp stages. This difference might be due to the previously suggested more rapid progression of BRAF mutated polyps than the KRAS mutated ones [25], since the former appear to exhibit an early activation of the proliferation-associated MAPK pathway. Interestingly, based on our logistic regression model for the SAC protein biomarkers, SPs share more with SACs than with CPs. This finding does, however, not exclude the possibility that TSAs may serve as precursor lesions of SACs, since the expression of SAC biomarkers may occur later in TSAs, or since this type of polyp could be the precursor lesion of a small proportion of SACs. In fact, it has recently been reported that BRAF mutated TSAs may serve as precursors of a MSS subtype of CRC with BRAF mutations [26]. Intriguingly, it has been found that the CP subset VA exhibits a 67 % biomarker similarity with SAC, although this similarity was not so evident in the molecular model (see Table S3), thus suggesting that in VA FSCN1 over-expression may be associated with different genetic alterations than those found in SSA. Several published studies support the view that the serrated polyp pathway is not streamlined, and that switching between serrated and conventional morphologies may occur during adenoma development [27]. An example of the latter may be the bi-phenotypic polyp. Concordantly, it was recently reported by Hafezi-Bakhtiari et al. [28] that VAs/TVAs often contain occasional glands typical of TSA. Although the mixed polyp is still a contentious entity, we have followed recent recommendations which suggest that it makes sense to retain the term mixed polyp which includes, among different combinations of serrated patterns, polyps showing features of both serrated and classical/conventional adenomas [8]. We have arbitrarily set a minimum proportion of 20 % of either component for such a diagnosis, as this percentage has not yet been defined. Although application of the regression model to oncogene mutations confirms the trends observed in the immunohistochemical analyses, our results indicate that these molecular biomarkers are not specific for either SAC or CC.

In conclusion, we found that certain polyps express protein biomarkers typical of SAC, that SSA is the most plausible precursor of this type of CRC, and that loss of EPHB2 and PTCH expression may act as important triggers for the progression of right-sided SSAs and left-sided TSAs, respectively.

Compliance with ethical standards

Funding

This study was funded by a grant (PI081210) from Carlos III Institute (Spanish Ministry of Economy and Competitiveness) and another (08,768/PI/08) from the Séneca Foundation Education Council from Murcia Region).

Conflict of interest

The authors declare that they have no conflict of interest.