Main

Anal squamous cell carcinoma (ASCC) is a rare tumour that accounts for <5% of all lower gastrointestinal tract malignancies in Europe (Glynne-Jones et al, 2014). The incidence of ASCC has increased steadily in the past decades, particularly in women (Forman et al, 2012) and in men who have sex with men and those with HIV infection (Silverberg et al, 2012). Infection from human papilloma virus (HPV) is the main aetiologic factor in the development of ASCC and >90% of patients are HPV positive (mainly HPV16 and 18) (Frisch et al, 1997; Abramowitz et al, 2011). Recent results of high-sensitivity HPV genotyping in a large series of ASCC patients showed a positivity rate of >95%. This supports the development of multivalent HPV vaccination for prevention (Baricevic et al, 2015). Concomitant chemoradiotherapy (CRT) is the standard of care for locally advanced tumours (Flam et al, 1996; Bartelink et al, 1997; Cacheux et al, 2012). So far, no predictive factor (to CRT) has been identified, excepted p16 expression, HPV status and TP53 mutations (Gilbert et al, 2013; Koerber et al, 2014; Serup-Hansen et al, 2014; Baricevic et al, 2015; Mai et al, 2015; Meulendijks et al, 2015; Rödel et al, 2015). Salvage abdominoperineal resection (APR) is the standard treatment for local failure or recurrence after CRT, but 30 to 60% of operated patients will experience a locoregional and/or metastatic recurrence (Mullen et al, 2007; Mariani et al, 2008; Lefèvre et al, 2012; Correa et al, 2013). For these patients with an inoperable locally advanced or metastatic disease, very few treatments are available and their effectiveness is limited. New therapeutic approaches and predictive factors of outcome are required in this context. A better understanding of molecular markers involved in anal carcinogenesis might lead to the identification of new therapeutic targets as well as prognostic and predictive biomarkers. Recently, the potential effectiveness of anti-epidermal growth factor receptor (EGFR) monoclonal antibodies in advanced ASCC has been suggested by case reports (Lukan et al, 2009; Barmetter et al, 2012) that may be explained by both a high frequency of EGFR overexpression (80–90%) and the rarity of KRAS mutations in these tumours (Van Damme et al, 2010; Paliga et al, 2012, Smaglo et al, 2015). The incidence of other major gene alterations, especially those implicated in the EGFR pathway, has been rarely studied in ASCC. In the present study, we examined the mutation status of RAS (KRAS, NRAS and HRAS), BRAF, MET, FBXW7, TP53 and PIK3CA genes in a large series of 148 ASCC patients and correlated mutation status with clinicopathological characteristics and patient survival.

Materials and methods

Patient population

We retrospectively analysed tumours from ASCC patients consecutively treated from 1992 to 2015 at the Institut Curie Hospital. We included all consecutive patients for whom formalin-fixed, paraffin-embedded (FFPE) tumour tissue was available, and collected clinicopathological data and outcomes. This retrospective study was reviewed and approved by the Ethics Committee of the Institut Curie (No. A10-024). According to French regulations, patients were informed of research performed with the biological specimens obtained during their treatment and did not express opposition. Staging of the disease was based on the 7th revised edition (2010) of the AJCC Anus Cancer.

DNA extraction

Six tissue sections of 6 μm thickness were obtained from FFPE tissues and a seventh tissue section stained with HE staining. The tumour-rich areas were microdissected using a single-use blade and the samples underwent proteinase K digestion in a rotating incubator at 56 °C for 3 days. DNA was extracted with the NucleoSpin kit (Macherey-Nalgen, Hoerdt, France) according to the supplier recommendations in two separate aliquots that were analysed in parallel.

Gene mutation screening

The primer sequences used both for HRM and Sanger sequencing are shown in Supplementary Table 1. The majority of the HRM primers were designed to span the entire exons with product sizes under 200 bp. Primers were designed for KRAS (exons 2–4), HRAS (exons 2 and 3), NRAS, (exons 2 and 3), BRAF (exon 15), FBXW7 (exons 9 and 10), PIK3CA (exons 9 and 20), MET (exons 18 and 19) and TP53 genes (exons 4–8) (Supplementary Table 1). The PCR for HRM and Sanger sequencing analysis was performed on a 384-well plate in the presence of the fluorescent DNA intercalating dye, LC green (Idaho Technology, Salt Lake City, UT, USA) in a LightCycler480 (Roche Diagnosis, Meylan, France). The reaction mixture in a 15 μl final volume contained LC green, UDP Glycosylase (Roche) and Roche Master Mix (Roche). The cycling and melting conditions were as follows: with an initial cycle of 10 min at 40 °C, one cycle of 95 °C for 10 min; 50 cycles of 95 °C for 10 s, 55–65 °C for 10 s, 72 °C for 30 s; one cycle of 97 °C for 1 min and a melt from 70 °C to 95 °C rising 0.2 °C per s. Depending on the melting temperature, a touchdown approach was done for some primers. All samples were tested in duplicate. The HRM data were analysed using the Genescan software (Roche). All samples including the wild-type exons were plotted according to their melting profiles on the differential plot graph. Any difference of the horizon line based on the wild-type sample was sequenced with Sanger sequencing.

Sanger sequencing

The reaction mixture in a total of 50 μl was made using 1 μl of PCR products without first purification followed by a sequencing reaction with Big Dye Terminator v3.1 (Thermofisher, Courtaboeuf, France) according to the manufacturer’s protocol. The sequencing products were purified with a Sephadex gel (GE Healthcare, Velizy-Villacoublay, France) before running on a 3500 Genetic Analyser (Applied Biosystems, Foster City, CA, USA). The sequencing data were visualised using Finch TV (Geospiza, Inc., Seattle, WA, USA) with detection sensibility of 10% mutated cells.

HPV detection

From 1998 to 2013, all samples were analysed by PCR using specific primers to identify HPV16, 18, 33, 45, 6 and 11 types and using GP5+/GP6+ primers to detect HPV L1 DNA as previously described (Lombard et al, 1998). After 2013, real-time PCR using Sybr Green (Roche Diagnostics, Mannheim, Germany) and specific primers for HPV16, 18 and 33 and the human GAPDH gene was performed on a 7900HT Fast Real-Time PCR System (Applied Biosystems). HPV L1 amplicons from HPV16-, 18- and 33-negative samples were sequenced by Sanger method with GP6+ primer and HPV type identification was performed by alignment of the sequence with HPV sequence references, using the nucleotide blast program from NCBI (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Statistical analysis

The statistical analysis plan was predefined jointly by the authors. Overall survival (OS) was defined as the period from the first day of radiotherapy (RT) or CRT to death from any cause. Data on patients who were alive at the end of follow-up (November 2015) were regarded as censored. Progression-free survival (PFS) was defined as the period from the first day of RT or RCT to the date of first disease progression or death from any cause. Cox univariate and multivariate regression was used for survival analysis and Fisher’s exact test was used for the analysis of contingency tables. All statistical tests were two sided and P-values of <0.05 were considered statistically significant. All analyses have been implemented in R version 3.2.1 R Development Core Team, 2013. The applicable R code can be found in the Supplementary Information.

Results

Tumour and patient characteristics

A total of 148 ASCC samples from patients treated in our institution were included and analysed in our gene mutation screening as summarised in the consort diagram (Figure 1): 96 tumours were treatment naive and 52 were samples from recurrence after initial RT or CRT. In total, 142 tumours (95.9%) were HPV positive among which 131 tumours (88.5%) had HPV16 infection. Only 16 patients (10.8%) had HIV infection. In the HIV+ population (n=16), all patients had concomitant HPV infection: 11 HPV16, 2 HPV18, 2 HPV6–11 and 1 HPV33. In the HIV− population (n=132), 6 patients had no HPV infection and 126 had concomitant HPV infection: 120 HPV16, 1 HPV6–11, 1 HPV33, 2 HPV35, 1 HPV59 and 1 HPV67. Tumour characteristics according to the treatment-naive or recurrence status of samples are summarised in Table 1.

Figure 1
figure 1

Consort diagram of the study.

Full size image
Table 1 Clinicopathological features of treatment-naive and recurrence tumour samples with subsequent treatment received (n=148)
Full size table

There were 114 females and 34 males. The median age at the diagnosis was 61 years (range: 37–96 years). Twenty-five patients were treated by initial surgery: exclusive surgery (n=5) and surgery followed by RT (n=11) or RCT (n=19). Thirty-two patients were treated by initial RT and 88 by initial CRT. One was treated by chemotherapy for an initial metastatic disease and 2 were not treated after the initial diagnosis. Only 16 patients (10.8%) had HIV infection. Forty-five patients underwent APR: 40 for local recurrence after RT or RCT, 3 at diagnosis and 2 for suspicion of local recurrence with complete histological response on surgical specimens. The median follow-up of 148 patients was 3.3 years (range: 0.2–39.6 years).

Gene mutation screening

Of the 148 tumours, 3 (2.0%) showed a KRAS exon 2 mutation, 30 (20.3%) a PIK3CA mutation, 9 (6.1%) a FBXW7 mutation and 7 (4.7%) a TP53 mutation (Table 2A and Supplementary Table 2). Five tumours (3.4%) had 2 synchronous mutations concerning these previous genes (PIK3CA/FBXW7 mutations in 3 tumours, KRAS/TP53 in 1 tumour and FBXW7/TP53 in 1 tumour). All tumours were wild type for HRAS, NRAS, BRAF and MET genes. In 15 ASCC patients, we analysed several available samples obtained at different therapeutic times or in different sites. We observed a total concordance of the Sanger analysis in 12 patients but the mutational profile was different between samples for 3 patients (Table 2B).

Table 2 (A) Prevalence of identified mutations in the 148 ASCC samples and distribution among treatment-naive tumours and tumour recurrences. (B) Heterogeneity of mutational profiles in different tumour samples from the same patient (n=3)
Full size table

Correlation between gene mutations and clinicopathological features and prognostic value

The distribution of the mutations was similar between treatment-naive tumours and tumour recurrences, except for TP53 mutations (Table 2A). We found that TP53 mutations were restricted to recurrence samples: 7 of 52 (13.5%) tumour recurrences vs 0 of 96 (0%) treatment-naive tumours (Fisher’s test, P=0.0005). Moreover, we observed that TP53 mutations were more frequently associated with HPV16negative samples: 3 of 131 (2.3%) HPV16-positive tumours vs 4 of 17 (23.5%) HPV16-negative tumours (Fisher’s test, P=0.003).

As the site and therapeutic status of tumour samples were heterogeneous in this large retrospective cohort of ASCC patients, we focussed our tumour analysis on homogenous groups of patients to study the association between mutational status and clinicopathological characteristics of the patients, and the impact of these parameters on OS. We also excluded nontreated tumours, tumours with ongoing treatment and those without sufficient follow-up (<6 months) in our prognostic analysis.

We identified a first group of treatment-naive tumours from 57 ASCC patients treated by initial exclusive CRT with a median follow-up of 3.1 years (range: 0.3–14 years) (Supplementary Table 3). Overall, recurrence rate was 24.6% (n=14 of 57). All tumours were HPV positive and 52 of 57 (91,2%) had HPV type 16. Only 1 (1.7%) KRAS, 3 (5.3%) FBXW7 and 1 (1.7%) TP53 mutations were identified in this group, whereas PIK3CA mutations were identified in 10 (17.5%) of them (8 in exon 9 and 2 in exon 20). No association was found between PIK3CA mutations and clinicopathological characteristics of patients (data not shown). Moreover, no correlation was found between PI3KCA mutation and PFS or OS (Supplementary Table 4).

We also selected a second group of 40 recurrent tumour samples from ASCC patients who underwent APR for local recurrence after initial RT or CRT. We excluded 2 samples from patients who died early after APR from postoperative complications (at day 6 and 10 respectively). We obtained a final cohort of 38 ASCC samples with a median follow-up of 18.2 years (range: 0.82–39.6 years). Overall, recurrence rate was 57.9% (n=22 of 38). Clinicopathological characteristics of this group of patients are summarised in Table 3. PIK3CA, FBXW7 and TP53 mutations were identified in 11 (28.9%), 5 (13.2%) and 4 (10.5%) recurrent tumours out of 38 respectively. No association was found between PIK3CA mutations and clinicopathological characteristics (Supplementary Table 5). A significant correlation by univariate Cox regression analysis was found between OS and gender (P=0.045), HPV16 status (P=0.048) and PIK3CA mutation (P=0.037) (Table 4 and Figure 2). Multivariate Cox analysis showed that HPV16 status (P=0.004), HIV status (P=0.032) and PIK3CA mutation (P=0.025) were independent prognostic factors (Table 4).

Table 3 Clinicopathological features of the 38 tumour relapse samples from APR after initial RT/CRT
Full size table
Table 4 Overall survival according to clinicopathological and mutational characteristics of the 38 patients who underwent APR for tumour recurrence after RT/CRT
Full size table
Figure 2
figure 2

Overall survival depending on the PIK3CA mutation in the 38 relapse APR patients after initial RT/RCT (P=0.025).

Full size image

Discussion

ASCC is known to be a very well radiosensitive tumour but 20% of patients failed to CRT, and no predictive markers of response have been prospectively validated. Moreover, in case of recurrence after RT/CRT, APR is the treatment of choice without any prognostic factor identified or any adjuvant treatment recommendation, although at least 50% of patients experience recurrence after this surgery (Mullen et al, 2007; Mariani et al, 2008; Lefèvre et al, 2012; Correa et al, 2013). In this context, a better biological and molecular characterisation of anal carcinogenesis is needed to improve the medical care of ASCC patients by identifying new therapeutic targets or prognostic biomarkers.

In the present study, which is the largest retrospective cohort of ASCC samples analysed by sequencing for multiple genes with complete clinicobiological data and long-term patient outcome available, we found frequent PIK3CA mutations (20.3%), as observed in previous smaller studies identifying PIK3CA mutation in 22% (11 out of 53, by pyrosequencing) and 32.5% (28 out of 86, by next-generation sequencing) of tumours (Casadei Gardini et al, 2014; Smaglo et al, 2015). The high level of PIK3CA mutation in ASCC provides a rationale to evaluate specific inhibitors of the PIK3CA/Akt/mTor pathway as demonstrated in preclinical models (Stelzer et al, 2010; Sun et al, 2013).

We identified very few KRAS exon 2 mutations (2.3%), in line with previous studies reporting low rates (Van Damme et al, 2010; Martin et al, 2014; Smaglo et al, 2015) or the absence of KRAS mutations (Paliga et al, 2012; Gilbert et al, 2013; Casadei Gardini et al, 2014) that could explain the effectiveness of EGFR monoclonal antibodies observed in ASCC patients (Lukan et al, 2009). The TP53 mutations were also rarely described in the literature, although a high frequency of TP53 protein expression is reported (Patel et al, 2007).

In our series, we confirm the low frequency of TP53 mutations (5.3%) but we found they were restricted to recurrence samples and more frequently associated to HPV16-negative samples. As observed in our cohort, it was recently reported that TP53 mutations were correlated with HPV16-negative status, predictive of resistance to RT/CRT and correlated with a poor prognosis in ASCC patients (Meulendijks et al, 2015). The same correlation between HPV status and TP53 mutations was previously described in head and neck cancer (Westra et al, 2008).

We also focussed our gene screening on FBXW7 gene. Mutations of FBXW7 gene have been frequently reported in not only various squamous cell carcinomas (Agrawal et al, 2011; Gao et al, 2014; Ojesina et al, 2014) but also adenocarcinoma (The Cancer Genome Atlas Network, 2012; Laforest et al, 2014; The Cancer Genome Atlas Research Network, 2014) and melanoma where a protein inactivation was found (Aydin et al, 2014). For the first time, we report FBXW7 mutations in 6% of ASCC. FBWX7 is known to be a key regulator of the cell cycle involved in the maintenance of normal stem cells and cancer-initiating cells (Takeishi and Nakayama, 2014). It could act as a critical tumour suppressor gene by targeting the NOTCH1 oncoprotein and therefore be an effective biomarker for the evaluation of Notch inhibitors in ASCC (Aydin et al, 2014). Our study finally shows that HRAS, NRAS, BRAF and MET genes are not mutated in ASCC.

Of note, we observed a different gene mutational status in 3 out of 15 patients for whom several tumour samples were available. For one of them, this could be the consequence of CRT on tumour DNA as the different mutational profiles were obtained from anal treatment-naive and pretreated samples respectively. For the two remaining patients, we can make the assumption of tumour heterogeneity as the different mutational profiles were observed in samples from two different tumour sites.

Our study is the first one assessing the relation between clinicopathological characteristics and mutational status of several genes in a large series of ASCC patients with details on treatments received, allowing an exploratory assessment of gene mutation predictive and prognostic value. In the large biomarker analysis of 199 ASCCs recently reported by Smaglo et al (2015), only part of the tumours were the subject of a gene sequencing (8 to 86 according to the gene analysed) and almost no clinicopathological information was available, therefore avoiding any correlation to be performed. In another series of 103 ASCC patients, paraffin-embedded tumour tissue was sufficient to perform analysis of KRAS, BRAF and PIK3CA gene mutation in only 50 patients (Casadei Gardini et al, 2014). In our study, none of the gene mutations identified was associated with clinicopathological characteristics. Casadei Gardini et al (2014) also found no association between PIK3CA mutations (found in 22% of cases) and clinical characteristics.

Several studies have reported on potential prognostic and predictive biomarkers of response to RT/CRT in ASCC (Lampejo et al, 2010; Myklebust et al, 2012; Fraunholz et al, 2013) but none of them was a gene mutation and none has been sufficiently validated to be used in clinical practice. In our large cohort, we selected 2 homogenous groups of patients regarding the treatments received to analyse: (1) the prognostic and predictive value on response to treatment of identified mutations in treatment-naive samples of patients exclusively treated by CRT (n=57) and (2) the prognostic value of identified mutations after APR for local recurrence following RT/CRT (n=38).

In the naive tumours treated by CRT, PIK3CA mutation identified on pretreatment samples was not found prognostic or predictive of response to CRT. This result is concordant with the study of Casadei Gardini et al (2014) in which PI3KCA mutation was not associated with PFS or OS of patients treated by CRT. The predictive impact of this mutation on tumour response to CRT was not explored in this study (Casadei Gardini et al, 2014). We could not study the prognostic or predictive value of KRAS, FBXW7 and TP53 mutations given their low frequency in our study.

To our knowledge, this is the first study assessing gene mutations as potential prognostic biomarkers in ASCC patients who underwent APR for local recurrence after RT or RCT. After multivariate Cox analysis we identified three independent factors associated with worse survival: a negative HPV16 status (P=0.004) and a positive HIV infection (P=0.032), which has already been reported (Wexler et al, 2008; Yhim et al, 2011), and also the presence of PIK3CA gene mutation (P=0.025) that is identified for the first time as a new independent prognostic marker in this setting. Of course, the prognosis value of PIK3CA mutations we report need to be validated in an independent and larger prospective cohort of ASCC, considering the relatively small sample size of our series. These PIK3CA mutations have been previously reported to be associated with poor prognostic in colorectal cancer (Barault et al, 2008; Ogino et al, 2009) but data in cervical squamous cell carcinoma are more divergent with both an association with better OS in early tumour stages (McIntyre et al, 2013) and a poor response following standard CRT in more advanced stages (de la Rochefordiere et al, 2015). Finally, gynecological cancer patients with PIK3CA mutations are more responsive to PI3K/Akt/mTor inhibitors than nonmutated patients (Husseinzadeh and Husseinzadeh, 2014). These results, together with our findings, suggest that PIK3CA mutations might play a major role in HPV-related squamous cell carcinoma, including anal carcinogenesis, especially in mechanisms of resistance to RT or CRT. They provide a rationale for the use of PI3K/Akt/mTor pathway inhibitors in radioresistant tumours, particularly in adjuvant setting after APR. Aspirin therapy, recently shown to be of particular efficacy in adjuvant treatment of PIK3CA-mutated colorectal cancer, could be another therapeutic option in this setting (Liao et al, 2012). In addition, there are recent data suggesting that the host immune reaction mediates response (Gilbert et al, 2016) via tumour-infiltrating lymphocytes. These data suggest the evaluation of immunotherapy in anal cancer, whose efficiency might be enhanced by cyclooxygenase inhibitors such as aspirin (Zelaney et al, 2015).