Advertisement

Oral Cancer

, Volume 2, Issue 1–2, pp 45–55 | Cite as

CD44 expression at the invasive tumor front: effect on patterning and prognosis in head and neck cancer

  • Sindhu Govindan
  • Roshan D. Cruz
  • Nisheena Raghavan
  • Safeena Kulsum
  • Radhika M. Bavle
  • Ravindra D. Ravi
  • Vijay Pillai
  • Athira Ramakrishnan
  • Jeyaram Illiaraja
  • Balaji Ramachandran
  • Jayaprakash Aravindakshan
  • Mukund Seshadri
  • Vikram D. Kekatpure
  • Wesley Hicks
  • Moni A. Kuriakose
  • Amritha Suresh
Original Article
  • 190 Downloads
Part of the following topical collections:
  1. Oral Pathology

Abstract

Purpose

The invasive tumor front (ITF) and pattern of invasion (POI) are significant prognosticators in head and neck cancer squamous cell carcinoma (HNSCC); this study evaluates the expression of CSC marker, CD44 at the ITF and their combined prognostic potential.

Methods

POI, as scored by the invasive pattern grading score (IPGS), was correlated with CD44 expression in HNSCC cell line-derived xenografts. Immunohistochemical expression of CD44 was correlated with the ITF features in patients and their combined prognostic ability assessed by multivariate analysis, Kaplan–Meier plots, and Cox regression models.

Results

Assessment of cell line-derived in vivo models indicated that drug-resistant, CD44-enriched cell lines led to generation of xenografts with aggressive POIs (IV) at the ITF. In patients (N = 173), a near significant association of CD44 with IPGS (p = 0.05) was observed. CD44 (p = 0.02), along with differentiation (p = 0.04), and lymphoplasmocytic infiltration (LPI) (p = 0.03) were independent prognosticators of recurrence. Notably, IPGS and poorly differentiated tumors were predictors of recurrence in CD44high cohort (p < 0.05), while severe LPI was a good prognosticator only in CD44low cohort (p < 0.006). In the treatment naïve cohort, the prognostic efficacy was confined to the early stage patients (p < 0.007). A multifactorial risk model significantly stratified patients based on outcome (p = 0.029).

Conclusion

CD44+ CSCs at the ITF correlated with aggressive POI and both factors together were poor prognosticators of HNSCC.

Keywords

Pattern of invasion Head and neck squamous cell carcinoma Prognosis CD44 Cancer stem cells and risk factors 

Introduction

The invasive tumor front (ITF), along with other clinical and pathological parameters, is considered a principal prognostic factor in most solid tumors, including head and neck squamous cell carcinoma [HNSCC] [1]. Given the high rates of recurrence (40%) in HNSCC, there is a need for parameters that can prognosticate with improved accuracy. Currently, molecular identifiers are being investigated for their prognostic ability [2], either independently or in combination with clinical/histological parameters. The patterning of the ITF, its underlying molecular basis, and combined prognostic implications gain prominence in this context.

The invasive front is considered the most active part of tumor, characterized by various histopathological features such as degree of keratinisation, pattern of invasion (POI), lymphoplasmocytic infiltration (LPI), and nuclear pleomorphism [3]. Studies in oral cancer indicate that these patterns observed at the host-tumor interface may determine prognosis and thereby are better predictors [4]. However, histopathological parameters and ITF scoring systems such as Invasive Pattern Grading Score (IPGS), devised for decision making and adjuvant treatment management, have not been adopted in clinical practice due to the lack of accuracy and consensus guidelines. Efforts to identify the underlying molecular basis have revealed the role of extra cellular matrix re-organization changes in cell adhesion (E-Cadherin, β-Catenin) and protease activity (MMPs). In colorectal cancer, expression of S100A4 and EMT markers (Vimentin, N-Cadherin) at ITF was determinant of survival [5, 6]. HNSCC patients showed an increased expression of α-Catulin and MMP19 at the invasive front with a knockdown of α-Catulin abrogating the invasive/migratory capacities of the cells [7]. Notwithstanding these initial insights into the molecular pathways involved, further investigations to identify novel processes/predictors are warranted.

Cancer stem cells (CSCs) are widely accepted today as one of the key determinants of invasion, metastasis, immunosuppression [8], and therapeutic resistance in tumors [9]. CD44, ABCG2, BMI-1, and β1-Integrin are associated with loco-regional recurrence and poor outcome after radiotherapy in HNSCC [10, 11]. Although a few studies have reported the presence of CSC markers and cells with EMT phenotype at the invasive front and their correlation with poor prognosis [12, 13], an in depth correlation of CSC expression with the histopathologic features at the ITF is not known in HNSCC. We hypothesise that the presence of CSCs at ITF and expression of CSC-specific markers contribute to biologic behaviour at the tumor front and thereby prognosis of head and neck squamous cell carcinoma (HNSCC).

Materials and methods

Cell lines

Head and neck squamous cell carcinoma cell lines, CAL-27 and Hep-2, and their resistant cell lines (CAL-27 CisR and Hep-2 TPFR) developed and characterized in the lab [14, 15] were used in this study. These resistant variants were enriched in CD44+ CSCs and showed enhanced migratory potential and were hence an apt in vitro model to correlate the effect of CSCs on POI [14, 15]. All cell lines were cultured in DMEM with 10% FBS (Thermo Fisher Scientific, Massachusetts, USA). CAL-27 CisR was cultured in the presence of the cisplatin (IC 6.25) and Hep-2 TPFR in the presence of the three drugs (IC 6.25) with a drug free period of 1–2 weeks before performing all the experiments.

Xenograft assay

The proposal was approved by the Institutional Animal Ethics Committee of the collaborating institution (Syngene Pvt Ltd, Biocon, Bangalore, India). 1 × 102–1 × 105 of resistant/parental cells of CAL-27 or Hep-2 (CAL-27 P, CAL-27 Cis R, Hep-2 P, and Hep-2 TPFR) were mixed with 1× Matrigel (BD Biosciences, California, U.S.A) and injected subcutaneously into the right and left flanks of NOD-SCID mice (5–6 weeks). Tumor formation/growth was assessed weekly, animals were sacrificed at a tumor size of 1000 mm3, and tumors analysed for histology and CD44 expression.

Patient details

The study was approved by the Institutional Ethical Committee. Patients diagnosed with head and neck cancer of all sites, who have undergone curative intent surgical treatment (December 2009-January 2012) were screened for the study. Exclusion criteria were age (< 18 years) and cases of verrucous and other variants of carcinoma (basaloid, papillary, and adenosquamous). Clinico-pathological data were collected from the hospital medical records and the patients followed up for a median period of 2 years.

Histopathological evaluation

Histological evaluation was carried out by two independent pathologists in a blinded manner and the consensus diagnosis used for data analysis. The POI at the host–tumor interface was determined using established criteria [16]. Slides with maximum tumor depth were selected, wherein the deep invasive front within stromal tissue was considered as ITF. The patterns were scored for their predominant patterns, as POI I (broad pushing manner), II (pushing finger-like and/or with large tumor islands), III (presence of tumor islands of > 15 cells), and IV (presence of single cells along with the islands of < 15 cells). POI was also scored using the invasive pattern grading score (IPGS) [17] accounting for the two most prevalent patterns, the predominant, and the secondary pattern at the ITF. A cutoff of 20% prevalence was considered to include a pattern in the score, and in case of a consistent single pattern, the IPGS scores were obtained by doubling the score. Since the patterns have a score of 1–4, IPGS scores range from 2 to 8 (IPGS  ≤ 4: low invasive; IPGS > 4: highly invasive).

Differentiation of the tumors was diagnosed as well, moderate and poor using the established grading system [18]. LPI at the tumor/host interface was quantified as strong (continuous, dense rim of lymphoid tissue), moderate (patches of dense lymphoid infiltrate at the interface with a discontinuous inflammation), or mild response (limited response without lymphoid patch/lymphoid response) [19]. Peri-neural invasion (PNI) was defined as carcinoma specifically tracking along or within a nerve [20].

Expression analysis of CD44

CD44 expression at protein and mRNA levels was evaluated by immunohistochemistry (IHC) and quantitative real time PCR (qRT-PCR). IHC was performed on the formalin fixed paraffin embedded (FFPE) tumor sections representing the tumor–host interface by the standard protocols. Briefly, sections were deparaffinised, treated with 3% hydrogen peroxide, incubated with the primary antibody (Mouse anti Human CD44, Clone DF1485, #AM310-5M, BioGenex, California, U.S.A), detected by the Poly-HRP detection system as per manufacturer’s instructions (#HK330-9K, BioGenex), and imaged using NIKON ECLIPSE 50i (200×, 400×, Nikon Instruments Inc., New York, USA). Sections processed without the primary antibody were negative controls. CD44 expression was scored at the invasive margins for patterns I and II, whereas in III and IV, it was scored from the tumor islands along the invasive margin [21]. Staining was considered positive in the presence of complete membranous staining (> 10%). The expression was scored negative if < 10% of the tumor cells showed focal, irregular staining. The percentage of positive cells (0–100%) was multiplied with the intensity of staining (mild/1+, moderate/2++, strong/3+++) to obtain the score (0–300).

RNA profiling was carried out in a subset of the IHC cohort based on the method described elsewhere [14]. The fold change in the expression of CD44 mRNA levels was calculated by 2∆∆CT method and correlated with the POI in the patients. The expression in the normal oral mucosa samples served as calibrator/control for both IHC and qRT-PCR.

Multifactorial risk assignment

A risk assessment score was developed to stratify the patients of all cohorts into risk groups (low, intermediate, and high) with varying survival probability (good, moderate, and poor) based on the ITF features and CD44 expression (four variables). The score assignment was adapted from the original scoring system introduced by Gensler et al. [22] (Table S1).

Statistical methods

The statistical results were presented either as mean with SEM (Standard error of the Mean), median with range or as frequencies (STATA11.1). Association of CD44 and IPGS with all the clinico-pathological variables was evaluated by the Chi-square test. Multivariate and univariate analyses and Kaplan–Meier curves were used to correlate the factors with outcome and survival. The prognostic significance of multifactorial scoring system was assessed by Cox regression model. Student’s t test (Unpaired and two tailed) was done for all the sub-group analysis using GraphPad Prism (version 5.0) software. A p value less than 0.05 was considered statistically significant.

Results

CD44 expression correlates with POI in HNSCC cell line-derived xenograft tumors

Hep-2 TPFR and CAL-27 CisR showed a 4.3 and 12.0% enrichment of CD44+ cells, respectively, by FACS (91.87 ± 1.53 vs. 87.55 ± 3.7% and 88.0 ± 0.1 vs. 74.0 ± 3) [14]. Assessment of the corresponding mice xenografts (n = 3–5 for each cell number) indicated a squamous epithelial histology of the tumors. Correlation of the POI at the ITF with the CD44 expression in both Hep-2 and CAL-27 xenografts showed a significant correlation between the aggressive POI (IV) and increased CD44 expression levels (n = 15; IHC score 197 ± 13; p = 0.01) as compared to POI I/II (n = 2; IHC score: 100 ± 20). There was no statistically significant difference observed in the CD44 expression of xenografts showing POI III (182 ± 24) as compared to POI I/II (Fig S1a, b and Table S2).

In Hep-2-derived xenografts, POI IV was the most predominant pattern in CD44+ cell-enriched Hep-2 TPFR (6/6) as compared to POI I–II in Hep-2 P (1/6, p = 0.01; POI III: 4/6) and this pattern showed correlation with increase in CD44 levels in Hep-2 TPFR (Hep-2 P: 110 ± 18.9 and Hep-2 TPFR: 178 ± 28.6; p = 0.07) although the difference in expression was not statistically significant. Among the Cal-27-derived xenografts, 87% (14/16) of the tumors derived from both the parental and resistant variant showed high expression of CD44 (IHC score > 200), which in turn correlated with aggressive POI (16/16; POI III/IV). However, there was no statistically significant difference in the CD44 scores or the corresponding POI patterns between the parental and Cal-27 CisR-derived xenografts.

Evaluation of the histopathological features at the ITF in HNSCC patients

Clinical characteristics of patients

Out of the 247 patients diagnosed during the time period, 173 were selected based on the exclusion and inclusion criteria. Patients were categorized into treatment naïve (cohort I; n = 119) and recurrent (cohort II; n = 54). Majority of the patients were with risk habits (126/159; 79.2%) and diagnosed with cancers of the oral cavity (141/173; 81.5%) (Table S3). In cohort I, 21.8% (n = 26) were of stages I and II, while majority of the patients in cohort II were stage IV (79.6%, n = 43).

The invasive tumor front features in HNSCC patients

POI was assessed based on the most predominant POI and by the IPGS. POI III (55.5%) and IV (22.5%) were the predominant patterns observed in a majority of the patients (Table S3). Categorization into IPGS showed a similar trend with 78% of the cases in the total cohort (n = 135) showing IPGS ≥ 4. Forty-five percentage (n = 79) showed mild lymphoplasmocytic infiltration, while the rest of the cohort was equally distributed between moderate and severe (24–29%). Majority of the tumors were well differentiated (67.6%. n = 117), while peri-neural invasion was observed in 46.2% (n = 80) of patients (Table S3). Chi-square analysis indicated that IPGS was significantly associated with differentiation (p = 0.004), LPI (p = 0.02) and PNI (p ≤ 0.001) in the total cohort (Table 1).

Sub-group analysis showed that in cohort I (n = 119), 74% tumors had IPGS > 4, 37.8% had mild LPI and 27.7% showed moderately/poorly differentiated carcinoma. In cohort II (n = 54), 87.3% tumors had IPGS > 4, while 63% showed mild LPI, and 42.6% were with moderately and poorly differentiated carcinoma. This indicated a more aggressive pattern at the ITF in a majority of the patients in cohort II. Among the early stage patients of the cohort I (stage I–II; n = 26), 95% had IPGS > 4, 34.6% showed mild LPI, while 27% had moderately to poorly differentiated carcinoma.

CD44 expression at the ITF correlates with aggressive POI in HNSCC patients

Expression levels of CD44 in HNSCC patients

CD44 protein expression was scored in the normal mucosa (restricted to the basement membrane) and in the HNSCC samples (at the ITF) (Fig. 1a–d). The average IHC score in the patients was 2.1 fold (174.3 ± 7.5) higher than the normal controls (85 ± 10.9). The patients were categorized into CD44high/CD44low cohorts using the normal score as the cut-off value (≤ 100: low expression; > 100: high) A majority of the patients were CD44high (62.0%; n = 98) with 15 cases (8.7%) being negative for CD44 expression (excluded from analysis). A comparison of CD44 IHC median scores in relation with IPG scores did not show a statistically significant correlation (POI III–IV (IPGS > 4): median 150, range 10–100; POI I–II (IPGS < 4): median 100, range 10–300), although a marginal increase in CD44 was recorded in patients with aggressive POI. The distribution, when assessed in a Stage-specific manner (Stage I–II; median 125 vs. 150, range 10–300, stage III–IV: median 100 vs. 150, range 10–300) (Fig. 1e), was also not statistically significant.
Fig. 1

CD44 correlates with aggressive POI in HNSCC patients. Representative images of a negative staining of CD44, b weak basement membrane expression of CD44 in normal oral mucosa, c low CD44 expression and d strong CD44 expression in the HNSCC patients (magnification ×400). Boxplots (e, f) of median CD44 IHC scores between POI I–II and III–IV of total cohort, Stage I–II and III–IV tumors of cohort I (e) and median fold difference in the CD44 mRNA expression between POI I–II and III–IV (f)

Furthermore, to see if this association is relevant at the mRNA levels, CD44 expression profiling was done in a subset of 42 patients (cohort 1, n = 23 and cohort II, n = 19) from the IHC cohort. Subsequent analysis of CD44 mRNA levels in relation with the different patterns of invasion indicated a trend similar to the protein levels and increased median fold expression of CD44 in POI III–IV (median 10.9; range 249.7–2082) to that of POI I–II (median 7.9, range 0.52–328.4) (Fig. 1f).

Distribution of CD44+ cells at the invasive tumor front

Immunohistochemical analysis indicated that the distribution of CD44+ cells correlated with POI. Few CD44+ cells were distributed in the borders of pattern I, while in the pattern II, they infiltrated the finger-like extensions with weak staining of CD44. Majority of the tumor islands and single cells in patterns III and IV were strongly (+++) positive for CD44 (Fig S2, a). In addition, patients with well differentiated tumors and with severe LPI show decreased prevalence of CD44+ cells at the ITF (Fig S2, b).

Correlation of CD44 expression level with ITF features

Statistical analysis in the entire cohort revealed that CD44 expression pattern showed an association with IPGS (p = 0.05) in the total cohort (Table 1). In cohort I, 78.1% (50/64) of CD44high patients showed IPGS > 4 as compared to 69% (31/45) in the CD44low group, although the distribution was not statistically significant (Table S4). Comparatively in cohort II, 69.4% (34/49) of patients were CD44high, 91.2% (31/34) of them showing IPGS > 4, indicating that treatment-induced upregulation of CD44 expression can lead to aggressive patterns of invasion. In addition, 55.9% (19/34) of patients with CD44high in cohort II also had mild LPI pointing out a possible role for CD44 in patterning immune response.
Table 1

Association of IPGS with clinico-pathological features

Clinical variables

IPG ≤ 4 (n = 38) No (%)

IPG > 4 (n = 135) No (%)

p value

Gender

 Male

25 (21%)

93 (79%)

0.717

 Female

13 (24%)

42 (76%)

TNM stage

 I and II

6 (19%)

26 (81%)

0.627

 III and IV

32 (23%)

109 (77%)

Differentiation

 Well

33 (28%)

84 (72%)

0.004

 Poor and moderate

5 (9%)

51 (91%)

Lymphoplasmacytic infiltration (LPI)

 Strong

17 (33%)

34 (67%)

0.02

 Mild and moderate

21 (17%)

101 (83%)

Peri-neural invasion (PNI)

 Present

6 (8%)

74 (92%)

< 0.001

 Absent

32 (34%)

61 (66%)

Recurrence (n = 152a)

 Yes

14 (16%)

72 (84%)

0.151

 No

17 (26%)

49 (74%)

Nodal

 N0

18 (27%)

48 (73%)

0.17

 N+

20 (19%)

87 (81%)

Habits

 With habits

31 (23%)

102 (77%)

0.4

 Without habits

7 (18%)

33 (82%)

Outcome (n = 153b)

 Dead

10 (15%)

57 (85%)

0.08

 Alive

23 (27%)

63 (73%)

CD44 (n = 158*)

 High

17 (49%)

81 (67%)

0.05

 Low

18 (51%)

42 (33%)

IPG score was available for 173 cases, but outcome (b) and recurrence (a) data was available only for 153 and 152 cases, respectively

*The CD44 staining was negative for 15 patients and excluded from analysis. The significant P values are indicated in bold

Prognostic significance of CD44 expression

Assessment of a subset of patients with follow-up details (n = 153; median 24; range 6–44 months) indicated an overall survival (OS) of 56.2% (86/153) in the total patient cohort. However, cohort I patients showed a comparatively higher rate of OS (65%; 65/100) with 32.7% (32/98) of recurrences. Cohort II patients experienced a poor survival rate (39.6%, 21/53).

The analysis of CD44 protein in cohort I indicated an increased median score in recurrent patients (n = 30, 150, range 10–300) as compared to the non-recurred patients (n = 59, 120), though the difference was not statistically significant (Fig. 2a). However, CD44 was significantly associated with recurrence in early stage (Stages I–II) patients of this cohort (p = 0.007, Table S4) and these patients also (n = 8) showed a significantly higher median CD44 score compared to those who did not recur (n = 13) on follow-up (225 vs. 100, range 60–300, p = 0.04) (Fig. 2b). In contrast, the advanced stage (Stages III–IV) patients with and without recurrence showed no difference in CD44 levels (150, range 10–300) (Fig. 2c). At mRNA levels, there was no statistically significant difference in the fold level expression between the different categories.
Fig. 2

Prognostic efficacy of CD44 expression and ITF features. CD44 expression was associated with recurrence in cohort I (a) between non-recurrent (NR) and recurrent (R) groups and in a stage-specific manner (b, c); df prognostic efficacy of individual ITF features were analysed in CD44low and CD44high groups by a three-tier analysis. d overall cohort showed a trend of correlation of IPGS with recurrence in CD44high group (p = 0.06) as compared to the CD44low group). e LPI was predictive of recurrence in CD44low patients and f tumor differentiation was predictive of recurrence in CD44high patients; g KM curve of multifactorial (CD44, IPGS, LPI and differentiation) risk assessment model stratify the patients into three different risk groups according their probability of overall survival

Prognostic efficacy of ITF features in the background of CD44 expression

Prognostic efficacy of ITF-CD44 in total patient population

Assessment of the different clinical and histological parameters along with the CD44 expression for their prognostic efficacy (univariate analysis) in entire cohort indicated that differentiation (OR = 2.8; 95% CI 1.35–5.8; p = 0.01) and LPI (OR = 3.28; 95% CI 1.56–6.9; p = 0.002) were the significant predictors of recurrence. Multivariate analysis indicated that differentiation (OR = 3.36; 95% CI 1.08–10.37; p = 0.04), LPI (OR = 3.06; 95% CI 1.11–8.41; p = 0.03), and CD44 (OR = 3.34; 95% CI 1.26–8.89; p = 0.02) were associated with recurrence (Table 2).
Table 2

Factors predicting recurrence in total cohort (N = 173)

 

Univariate analysis

Multivariate analysis

Odds ratio

95% CI

p value

Odds ratio

95% CI

p value

Gender

 Female

1

     

 Male

1.66

0.83–3.3

0.15

1.70

0.68–4.28

0.26

TNM stage

 I–II

1

     

 III–IV

1.38

0.61–3.2

0.44

3.24

0.73–14.36

0.12

Differentiation

 Well

1

     

 Poor and moderate

2.80

1.35–5.8

0.01

3.36

1.08–10.37

0.04

Lymphoplasmacytic infiltration (LPI)

 Severe

1

     

 Mild and moderate

3.28

1.56–6.9

0.002

3.06

1.11–8.41

0.03

PNI

 Absent

1

     

 Present

1.61

0.84–3.1

0.15

0.67

0.24–1.85

0.44

IPG score

 ≤ 4

1

     

 > 4

1.78

0.81–4.0

0.15

0.98

0.32–2.99

0.97

Nodal

 N0

1

     

 N+

1.01

0.52–2.0

0.97

0.43

0.14–1.33

0.15

Habits

 Without

1

     

 With

1.53

0.74–3.2

0.25

0.85

0.30–2.41

0.76

CD44

 Low

1

     

 High

1.93

0.96–3.89

0.07

3.34

1.26–8.89

0.02

The table shows significant results of univariate and multivariate analysis. The cox regression analysis was done for odds ratios and p value with a 95% confidence level. Differentiation (p = 0.04), LPI (p = 0.03) and CD44 (p = 0.02) were significant in a multivariate setting to predict the recurrence

Assessment of prognostic efficacy of the ITF features in the CD44low and CD44high groups indicated an increased susceptibility towards recurrence in patients with high IPGS in the CD44high cohort (p = 0.06) (Fig. 2d and Table S5). In the CD44low group, IPGS was not predictive of recurrence. CD44 expression and LPI was inversely correlated with recurrence; the levels of LPI correlating with recurrence only in the CD44low group (p = 0.006) (Fig. 2e). Similarly, poor differentiation correlated with recurrence in the CD44high patients (p = 0.001), while this correlation was not significant in CD44low group (p = 0.27) (Fig. 2f, Table S5). There was no correlation observed between LPI, IPGS and tumor differentiation, CD44 expression, and overall survival (data not shown).

The three-way analysis was not carried out using mRNA expression levels due to low sample numbers. However, a comparison of the overall survival advantage of the patients with high CD44mRNA levels and high IPG score to that of low CD44 mRNA and low IPG score indicated that the CD44low/IPGlow patients had 1.8 fold (n = 4, 75% OS) survival benefit compared to the CD44high/IPG high group of patients (n = 14, 42.8% OS); however, this association was not statistically significant due to the too low sample size (data not shown).

Prognostic efficacy of ITF in treatment naive cohort (cohort I)

Univariate analysis in the cohort I showed significant predictive power of differentiation (p = 0.023) and habits (p = 0.035) for recurrence. In the multivariate setting, differentiation (OR = 4.32; 95% CI 1.67–11.21; p = 0.003) and LPI (OR = 3.27; 95% CI 1.38–7.74; p = 0.007) showed a statistically significant odds ratio (OR) for developing recurrence. A trend towards association of CD44 (OR = 2.00; 95% CI 0.90–4.45 p = 0.089) with recurrence was observed in cohort I (Table S6).

Analysis of the prognostic effect of CD44 and ITF features indicated that CD44 was not significantly associated with other clinico-pathological variables including IPGS (Table S4). However, analysis within the CD44high and CD44low cohorts indicated that 78.1% of the IPGS > 4 sub-group patients had high expression (Table S4). A three-way analysis in the background of CD44 expression showed a statistically significant correlation of differentiation (p = 0.02) with recurrence in the CD44high groups, whereas IPGS did not show any significant correlation (Table S7).

In the early stage sub-group within cohort I, CD44 was associated with recurrence (p = 0.007) as 100% of recurred patients (8/8) had CD44 high expression, as compared to only 53.8% (7/13) in non-recurred group (Table S4). In addition, 61.5% (8/13) of patients with CD44high/IPGS > 4 developed recurrence, while none of the patients (3/3) recurred in the CD44low/IPGS ≤ 4 group (p = 0.01).

Risk stratification of patients including ITF and CD44 expression

CD44, IPGS, LPI, and differentiation were incorporated into the risk assessment model, and the patients stratified into three groups and their survival risk assessed based on their combined score (0: good survival, 1/2: moderate, and 4: poor survival) (Table S1). Regression analysis using this model showed that the high-risk group in the total cohort had two times increased risk of recurrence (HR-1.97, p = 0.002; 95% CI 1.23–3.00). A similar analysis in the treatment naive cohort also indicated poor prognosis for the high-risk group (HR of 1.79; p = 0.087; 95% CI 0.92–3.5), (Fig. 2g and Table S8).

Discussion

Risk assessment with the inclusion of pathological and molecular markers is essential in cancer patients to improve the treatment planning/management. Multi-parameter histological risk models have been validated in primary head and neck cancer as significant predictors of poor clinical outcome [23]. In parallel, studies have implicated cancer stem cells (CSCs) in the invasive properties of the tumor, recurrence, and metastasis. This study highlights the role of CSC marker, CD44 in the pattern of invasion at the ITF and prognosis of patients with head and neck squamous cell carcinoma (HNSCC).

In HNSCC, CSCs are characterized by markers such as CD44, CD133 [24], and ALDH1A1 [25], which are in turn associated with tumor initiation, drug resistance, and metastasis [26]. However, POI, the main feature at the ITF scored as IPGS, has not been previously associated with CSC expression patterns in HNSCC. In this study, xenografts generated from CSC enriched, and drug-resistant HNSCC cell lines developed aggressive POI at the tumor invasive front. Furthermore, analysis of post-treatment tumors from recurrent patients indicated higher CD44 expression accompanied with aggressive POI, mild LPI, and poor differentiation, demonstrating a possible treatment-induced, possibly CSC-driven, changes in the ITF. This was in accordance with other studies, including our previous findings, wherein CSC enrichment post-treatment, resulted in highly aggressive tumorigenic behaviour in vitro, a similar pattern of enrichment correlated with poor prognosis in patients [14].

Aggressive patterns of POI necessitate multiple processes/changes in tumor cells and its environment. Recent studies in nasopharyngeal carcinoma (NPC) and breast cancer demonstrated the presence of spindle-like cells in the invasive front that acquire EMT properties and turn migratory, cells with similar behaviour being CD44+ in breast cancer [27]. This speculation that budding CSCs can disseminate from the tumor and migrate to create islands in the stroma is a pointer towards their possible role in generating aggressive patterns. These evidences, combined with our results of CD44 expression in the tumor invasive margins and tumor islands, indicate a strong role for CD44+CSCs in patterning the tumor front of HNSCC. In the model proposed, the CSCs at the invasive front of HNSCC correlated with different types of invasion; the presence of very few/no cancer stem cells can leave the tumor front at a broad pushing pattern (POI type I, Fig. 3a). In contrast, enrichment of these cells due to individual tumor biology [28] or as a result of treatment [14] can lead to finger-like outgrowths (POI type II, Fig. 3b). A subsequent detachment from the epithelium and migration into the stroma can lead to tumor islands, which later increase in number or completely invade the stromal area (POI type III and IV, Fig. 3c). Further functional studies need to be carried out to gather evidences in support of this model.
Fig. 3

Model for CSCs in Pattern of Invasion. This model explains the possible mechanisms of CSCs in patterning the invasive front. a Presence of fewer CSCs (Red) at the invasive front, that drives the proliferating cells, leads to pushing pattern of invasion (POI I); b an increase in the number of CSCs in combination with their high migratory potential leads to finger-like pattern of invasion (POI II) and c detachment of these cells from tumor margin leads to the formation of small islands or single tumor cells (POI III and IV)

Our study indicated a strong association of CD44 levels at the tumor front with prognosis, both in terms of increased expression level in recurrent patients and as a significant predictor in multivariate analysis (p < 0.05). This observation is similar to studies wherein CSC markers reported at the invasive front of various solid tumors [12] were associated with TNM Stage, lymphatic invasion and local recurrence [29]. Further, in this study, CSCs, along with ITF features improved the prognostic efficacy of these parameters. IPGS was a better predictor of disease recurrence when evaluated in combination with CD44. Interestingly, the prognostic efficacy of CD44 in combination with ITF features was more predominant in treatment naïve early stage patients compared to the advanced cases.

Lymphoplasmocytic infiltration and differentiation at the tumor–host interface are other ITF features predictive of clinical prognosis in many cancers [22, 30]. In this study, the prognostic efficacy of LPI is inversely dependent on CD44 levels, with mild LPI being poor prognostic only in CD44low background. Studies in glioblastoma (GBM) and hepatocellular carcinoma (HCC) revealed a correlation between CSCs and TGF-β-mediated immunosuppression leading to aggressive tumors [31, 32] the negative correlation of CD44 with immune infiltration in our study needs to be further validated in this context. Degree of differentiation is determined by several molecular events such as gain/loss of adhesion molecules, secretion of proteolytic enzyme, initiation of angiogenesis, Epithelial Mesenchymal Transition (EMT) and downregulation of CSC markers [33, 34]. In our study, elevated CD44 at the tumor interface is associated with poor differentiation, both in combination, being a better predictor of prognosis. The findings of this study indicated that ITF features in combination with CSCs can be explored as a poor prognosticator in HNSCC, subjective to site- and stage-specific validation in a larger cohort of patients.

Given that tumorigenesis and progression is an interplay of multiple factors, risk assessment needs to consider these parameters in order to provide effective and accurate prognosis. In this study, the proposed risk assessment model, including the features of ITF and CD44 expression patterns, categorized the patients accurately based on their outcome (p < 0.029). These evidences do suggest that CSCs may play a decisive role at the tumor front with clinical implications as a prognosticator, subject to validation in larger cohorts of patients. Further evidences to evaluate the mechanistic details of the role of CD44 and thereby the CSCs at the tumor invasive front are also definitely warranted and are currently ongoing in our laboratory.

Notes

Acknowledgements

The authors acknowledge Indian Council of Medical Research (ICMR), for the Research Fellowship awarded to SVG and Department of Biotechnology, (DBT), Government of India for the grant to AS. Mr. Deenadayal, Department of Pathology and Mr Jais Kurian, Head and Neck Oncology, Mazumdar Shaw Medical Center are duly acknowledged for their contributions.

Compliance with ethical standards

Financial support

Senior research fellowship to Sindhu V Govindan by ICMR Extramural project No 3/2/2/162/2008/NCD-III, ICMR, India. This work was supported by RGYI grant No.BT/PR15027/GBD/27/286/2010, Department of Biotechnology, India.

Conflict of interest

The authors declare no potential conflicts of interest.

Supplementary material

41548_2018_8_MOESM1_ESM.tif (4.5 mb)
Fig S1 CD44 expression determines pattern of invasion (POI) in HNSCC cell line-derived xenograft models. Representative images for POI and CD44 are shown for Hep-2 (a-b) and CAL-27 (e-h) derived xenografts. Hep-2 (a) and CAL-27 P (e) show less aggressive POI as compared to Hep-2 TPFR (b) and CAL-27Cis R (f). Correspondingly the CD44 expression in Hep-2 TPFR (d) and CAL-27Cis R (h) was higher as compared to the parental cells (c and) (Magnification 200X). (TIFF 4596 kb)
41548_2018_8_MOESM2_ESM.tif (4.7 mb)
Fig S2 Distribution of CD44+ cells at the invasive front of HNSCC tumors. CD44 expression was compared in different patterns of POI (a), differentiation and LPI (b) (Magnification 400X). Analysis in the different POI patterns indicated a lower presence in pattern I and II (a, upper panel). The expression is restricted to the borders in POI I while it infiltrates the finger–like extensions of POI II. The aggressive patterns (III & IV) showed a high expression of CD44 (b, lower panel) with most of the cells being positive. Well differentiated tumors showed a low CD44 expression as compared to poorly differentiated tumors while a severe LPI pattern had low CD44 as compared to the mild patterns (b) (TIFF 4851 kb)
41548_2018_8_MOESM3_ESM.docx (63 kb)
Supplementary material 3 (DOCX 64 kb)

References

  1. 1.
    Tokodai K, Narimatsu H, Nishida A, Takaya K, Hara Y, Kawagishi N et al (2016) Risk factors for recurrence in Stage II/III colorectal cancer patients treated with curative surgery: the impact of postoperative tumor markers and an infiltrative growth pattern. J Surg Oncol 114:368–374.  https://doi.org/10.1002/jso.24320 CrossRefPubMedGoogle Scholar
  2. 2.
    Sinevici N, O’Sullivan J (2016) Oral cancer: deregulated molecular events and their use as biomarkers. Oral Oncol 61:12–18.  https://doi.org/10.1016/j.oraloncology.2016.07.013 CrossRefPubMedGoogle Scholar
  3. 3.
    Bankfalvi A, Piffko J (2000) Prognostic and predictive factors in oral cancer: the role of the invasive tumour front. J Oral Pathol Med 29:291–298CrossRefPubMedGoogle Scholar
  4. 4.
    Attramadal CG, Kumar S, Boysen ME, Dhakal HP, Nesland JM, Bryne M (2015) Tumor budding, EMT and cancer stem cells in T1-2/N0 oral squamous cell carcinomas. Anticancer Res 35:6111–6120PubMedGoogle Scholar
  5. 5.
    Choi JE, Bae JS, Kang MJ, Chung MJ, Jang KY, Park HS, Moon WS (2017) Expression of epithelial-mesenchymal transition and cancer stem cell markers in colorectal adenocarcinoma: clinicopathological significance. Oncol Rep 38:1695–1705.  https://doi.org/10.3892/or.2017.5790 CrossRefPubMedGoogle Scholar
  6. 6.
    Sugai T, Yamada N, Eizuka M, Sugimoto R, Uesugi N, Osakabe M et al (2017) Vascular invasion and stromal S100A4 expression at the invasive front of colorectal cancer are novel determinants and tumor prognostic markers. J Cancer 8:1552–1561.  https://doi.org/10.7150/jca.18685 CrossRefPubMedPubMedCentralGoogle Scholar
  7. 7.
    Cao C, Chen Y, Masood R, Sinha UK, Kobielak A (2012) alpha-Catulin marks the invasion front of squamous cell carcinoma and is important for tumor cell metastasis. Mol Cancer Res 10:892–903.  https://doi.org/10.1158/1541-7786.mcr-12-0169 CrossRefPubMedPubMedCentralGoogle Scholar
  8. 8.
    Codony-Servat J, Rosell R (2015) Cancer stem cells and immunoresistance: clinical implications and solutions. Transl Lung Cancer Res 4:689–703.  https://doi.org/10.3978/j.issn.2218-6751.2015.12.11 PubMedPubMedCentralGoogle Scholar
  9. 9.
    Bourguignon LY, Wong G, Shiina M (2016) Up-regulation of histone methyltransferase, DOT1L, by matrix hyaluronan promotes MicroRNA-10 expression leading to tumor cell invasion and chemoresistance in cancer stem cells from head and neck squamous cell carcinoma. J Biol Chem 291:10571–10585.  https://doi.org/10.1074/jbc.m115.700021 CrossRefPubMedPubMedCentralGoogle Scholar
  10. 10.
    Yanamoto S, Yamada S, Takahashi H, Naruse T, Matsushita Y, Ikeda H et al (2014) Expression of the cancer stem cell markers CD44v6 and ABCG2 in tongue cancer: effect of neoadjuvant chemotherapy on local recurrence. Int J Oncol 44:1153–1162.  https://doi.org/10.3892/ijo.2014.2289 CrossRefPubMedGoogle Scholar
  11. 11.
    Joshua B, Kaplan MJ, Doweck I, Pai R, Weissman IL, Prince ME, Ailles LE (2012) Frequency of cells expressing CD44, a head and neck cancer stem cell marker: correlation with tumor aggressiveness. Head Neck 34:42–49.  https://doi.org/10.1002/hed.21699 CrossRefPubMedGoogle Scholar
  12. 12.
    Luo WR, Yao KT (2014) Cancer stem cell characteristics, ALDH1 expression in the invasive front of nasopharyngeal carcinoma. Virchows Arch 464:35–43.  https://doi.org/10.1007/s00428-013-1508-z CrossRefPubMedGoogle Scholar
  13. 13.
    Kodama H, Murata S, Ishida M, Yamamoto H, Yamaguchi T, Kaida S et al (2017) Prognostic impact of CD44-positive cancer stem-like cells at the invasive front of gastric cancer. Br J Cancer 116:186–194.  https://doi.org/10.1038/bjc.2016.401 CrossRefPubMedGoogle Scholar
  14. 14.
    Valiyaveedan SG, Ramachandran B, Iliaraja J, Ravindra DR, James BL, Kulsum S et al (2015) Acquisition of cancer stem cell behaviour plays a role in drug resistance to combination chemotherapy and prognosis in head and neck cancer. J Stem Cell Res Ther.  https://doi.org/10.4172/2157-7633.100026 Google Scholar
  15. 15.
    Kulsum S, Sudheendra HV, Pandian R, Ravindra DR, Siddappa GRN, Chevour P et al (2017) Cancer stem cell mediated acquired chemoresistance in head and neck cancer can be abrogated by aldehyde dehydrogenase 1 A1 inhibition. Mol Carcinogenesis 56:694–711.  https://doi.org/10.1002/mc.22526 CrossRefGoogle Scholar
  16. 16.
    Bryne M, Koppang HS, Lilleng RAK (1992) Malignancy grading of the deep invasive margins of oral squamous cell carcinomas has high prognostic value. J Pathol 166:375–381CrossRefPubMedGoogle Scholar
  17. 17.
    Chang Yun-Ching, Nieh Shin, Chen Su-Feng, Shu-Wen Jao Yu-Lu, Lin EF (2010) Invasive pattern grading score designed as an independent prognostic indicator in oral squamous cell carcinoma. Histopathology 57:295–303.  https://doi.org/10.1111/j.1365-2559.2010.03616 CrossRefPubMedGoogle Scholar
  18. 18.
    Broders AC (1927) Carcinoma of the mouth: types and degrees of malignancy. Am J Roentgenol Red Ther Nucl Med 90–93Google Scholar
  19. 19.
    Anneroth G, Batsakis J, Luna M (1987) Review of the literature and a recommended system of malignancy grading in oral squamous cell carcinomas. Scand J Dent Res 95:229–249PubMedGoogle Scholar
  20. 20.
    Hosal AS, Unal OF, Ayhan A (1998) Possible prognostic value of histopathologic parameters in patients with carcinoma of the oral tongue. Eur Arch Otorhinolaryngol 255:216–219CrossRefPubMedGoogle Scholar
  21. 21.
    Mack B, Gires O (2008) CD44s and CD44V6 expression in head and neck epithelia. PloS One 3:e3360.  https://doi.org/10.1371/journal.pone.0003360 CrossRefPubMedPubMedCentralGoogle Scholar
  22. 22.
    Brandwein-Gensler M, Teixeira MS, Lewis CM, Lee B, Rolnitzky L, Hille JJ et al (2005) Oral squamous cell carcinoma. Histologic risk assessment, but not margin status, is strongly predictive of local disease-free and overall survival. Am J Surg Pathol 29:167–178CrossRefPubMedGoogle Scholar
  23. 23.
    Strieder L, Coutinho-Camillo CM, Costa V, da Cruz Perez DE, Kowalski LP, Kaminagakura E (2016) Comparative analysis of three histologic grading methods for squamous cell carcinoma of the lip. Oral diseases.  https://doi.org/10.1111/odi.12586 PubMedGoogle Scholar
  24. 24.
    Kaseb HO, Fohrer-Ting H, Lewis DW, Lagasse E, Gollin SM (2016) Identification, expansion and characterization of cancer cells with stem cell properties from head and neck squamous cell carcinomas. Experim Cell Res 348:75–86.  https://doi.org/10.1016/j.yexcr.2016.09.003 CrossRefGoogle Scholar
  25. 25.
    Leinung M, Ernst B, Doring C, Wagenblast J, Tahtali A, Diensthuber M et al (2015) Expression of ALDH1A1 and CD44 in primary head and neck squamous cell carcinoma and their value for carcinogenesis, tumor progression and cancer stem cell identification. Oncol Lett 10:2289–2294.  https://doi.org/10.3892/ol.2015.3542 CrossRefPubMedPubMedCentralGoogle Scholar
  26. 26.
    Grau JJ, Mesia R, de la Iglesia-Vicente M, Williams ES, Taberna M, Caballero M et al (2016) Enrichment of cells with cancer stem cell-like markers in relapses of chemoresistant patients with locally advanced head and neck squamous cell carcinoma. Oncology 90:267–272.  https://doi.org/10.1159/000445118 CrossRefPubMedGoogle Scholar
  27. 27.
    Liu S, Cong Y, Wang D, Sun Y, Deng L, Liu Y et al (2013) Breast cancer stem cells transition between epithelial and mesenchymal states reflective of their normal counterparts. Stem Cell Rep 2:78–91.  https://doi.org/10.1016/j.stemcr.2013.11.009 CrossRefGoogle Scholar
  28. 28.
    Prince ME, Sivanandan R, Kaczorowski A, Wolf GT, Kaplan MJ, Dalerba P et al (2007) Identification of a subpopulation of cells with cancer stem cell properties in head and neck squamous cell carcinoma. Proc Natl Acad Sci USA 104:973–978.  https://doi.org/10.1073/pnas.0610117104 CrossRefPubMedPubMedCentralGoogle Scholar
  29. 29.
    Luo W, Yao K (2013) Molecular characterization and clinical implications of spindle cells in nasopharyngeal carcinoma: a novel molecule-morphology model of tumor progression proposed. PloS One 8:e83135.  https://doi.org/10.1371/journal.pone.0083135 CrossRefPubMedPubMedCentralGoogle Scholar
  30. 30.
    Hayashi A, Shibahara J, Misumi K, Arita J, Sakamoto Y, Hasegawa K et al (2016) Histologic assessment of intratumoral lymphoplasmacytic infiltration is useful in predicting prognosis of patients with hepatocellular carcinoma. PloS One 11:e0155744.  https://doi.org/10.1371/journal.pone.0155744 CrossRefPubMedPubMedCentralGoogle Scholar
  31. 31.
    Beier CP, Kumar P, Meyer K, Leukel P, Bruttel V, Aschenbrenner I et al (2012) The cancer stem cell subtype determines immune infiltration of glioblastoma. Stem Cells Dev 21:2753–2761.  https://doi.org/10.1089/scd.2011.0660 CrossRefPubMedPubMedCentralGoogle Scholar
  32. 32.
    Mima K, Okabe H, Ishimoto T, Hayashi H, Nakagawa S, Kuroki H et al (2012) CD44s regulates the TGF-beta-mediated mesenchymal phenotype and is associated with poor prognosis in patients with hepatocellular carcinoma. Cancer Res 72:3414–3423.  https://doi.org/10.1158/0008-5472.can-12-0299 CrossRefPubMedGoogle Scholar
  33. 33.
    Jogi A, Vaapil M, Johansson M, Pahlman S (2012) Cancer cell differentiation heterogeneity and aggressive behavior in solid tumors. Ups J Med Sci 117:217–224.  https://doi.org/10.3109/03009734.2012.659294 CrossRefPubMedPubMedCentralGoogle Scholar
  34. 34.
    Li Z, Bao S, Wu Q, Wang H, Eyler C, Sathornsumetee S, Shi Q et al (2009) Hypoxia-inducible factors regulate tumorigenic capacity of glioma stem cells. Cancer Cell 15:501–513.  https://doi.org/10.1038/cmi.2009.105 CrossRefPubMedPubMedCentralGoogle Scholar

Copyright information

© Springer International Publishing AG, part of Springer Nature 2018

Authors and Affiliations

  • Sindhu Govindan
    • 1
  • Roshan D. Cruz
    • 2
  • Nisheena Raghavan
    • 3
  • Safeena Kulsum
    • 1
  • Radhika M. Bavle
    • 4
  • Ravindra D. Ravi
    • 1
  • Vijay Pillai
    • 2
  • Athira Ramakrishnan
    • 2
  • Jeyaram Illiaraja
    • 5
  • Balaji Ramachandran
    • 6
  • Jayaprakash Aravindakshan
    • 6
  • Mukund Seshadri
    • 7
    • 8
    • 9
  • Vikram D. Kekatpure
    • 2
  • Wesley Hicks
    • 8
    • 9
  • Moni A. Kuriakose
    • 1
    • 2
    • 9
  • Amritha Suresh
    • 1
    • 2
    • 9
  1. 1.Integrated Head and Neck Oncology Program, DSRG-5, Mazumdar Shaw Centre for Translational ResearchMazumdar Shaw Medical FoundationBangaloreIndia
  2. 2.Department of Head and Neck OncologyMazumdar Shaw Medical Centre, Narayana HealthBangaloreIndia
  3. 3.Department of PathologyMazumdar Shaw Medical Centre, Narayana HealthBangaloreIndia
  4. 4.Department of Oral and Maxillofacial PathologyKrishnadevaraya College of Dental SciencesBangaloreIndia
  5. 5.Department of Clinical ResearchMazumdar Shaw Medical Centre, Narayana HealthBangaloreIndia
  6. 6.Department of PharmacologySyngene International Pvt. LtdBangaloreIndia
  7. 7.Department of Pharmacology and TherapeuticsRoswell Park Cancer InstituteBuffaloUSA
  8. 8.Department of Head and Neck/Plastic and Reconstructive SurgeryRoswell Park Cancer InstituteBuffaloUSA
  9. 9.Mazumdar Shaw Medical Centre-Roswell Park Collaborative Head and Neck Oncology Research ProgramRoswell Park Cancer InstituteBuffaloUSA

Personalised recommendations