Combined assessment of EGFR-related molecules to predict outcome of 1st-line cetuximab-containing chemotherapy for metastatic colorectal cancer

ABSTRACT

Several studies have reported that epidermal growth factor receptor (EGFR)-related molecules may serve as predictors of cetuximab treatment for metastatic colorectal cancer (mCRC), such as EGFR gene copy number (GCN), expression of 2 ligands of EGFR, amphiregulin (AREG) and epiregulin (EREG), and EGFR CA simple sequence repeat 1 (CA-SSR1) polymorphism; however, these biomarkers still remain not useful in clinical practice since they have been evaluated using cohorts with patients treated in various settings of chemotherapy. We therefore analyzed associations of mRNA expression of AREG and EREG, EGFR GCN, and CA-SSR1 polymorphism [short (S;≤ 19) / long (L; ≥ 20)] with clinical outcomes in 77 Japanese patients with KRAS exon 2 wild-type mCRC enrolled in phase II trials of FOLFOX (n = 28/57, UMIN000004197) or SOX (n = 49/67, UMIN000007022) plus cetuximab as first-line therapy. High AREG expression correlated with significantly better progression-free survival (median 11.6 vs. 66 months, HR 0.52, P = 0.037); moreover, it remained statistically significant in multivariate analysis (HR: 0.48, P = 0.027). S/S genotype of CA-SSR1 predicted severe skin toxicity (P = 0.040). Patients with both AREG-low and EGFR low-GCN had significantly shorter overall survival than the others (median 22.2 vs. 42.8 months, HR 2.34, P = 0.042). The multivariate analysis showed that molecular status with both AREG-low and EGFR low-GCN was a predictor of worse survival (P = 0.006). In conclusion, AREG mRNA expression and EGFR CA-SSR1 polymorphism predict survival and skin toxicity, respectively, of initial chemotherapy with cetuximab. Our results also suggest potential prognostic value of the combined assessment of AREG and EGFR GCN for first-line cetuximab treatment.

KEYWORDS:

• Amphiregulin
• cetuximab
• colorectal cancer
• EGFR
• EGFR gene copy number

Abbreviations

EGFR	=	epidermal growth factor receptor
mCRC	=	metastatic colorectal cancer
AREG	=	amphiregulin
EREG	=	epiregulin
CA-SSR1	=	CA simple sequence repeat 1
GCN	=	gene copy number
PFS	=	progression-free survival
OS	=	overall survival
ECOG	=	Eastern Cooperative Oncology Group
CT	=	computed tomography
ETS	=	early tumor shrinkage
FFPE	=	formalin-fixed paraffin-embedded

Introduction

Cetuximab, an IgG1 monoclonal antibody, binds to ligand-binding domain of epidermal growth factor receptor (EGFR) and thus exerts an inhibitory effect on tumor proliferative signaling in the downstream of EGFR pathway.¹ In metastatic colorectal cancer (mCRC) patients with tumors harboring KRAS exon 2 wild-type, this drug became a standard component of first-line treatment according to several evidences with significant survival benefit from clinical trials.^2-7 Then, retrospective analyses of large prospective studies to further understand the molecular determinants of responsiveness to anti-EGFR monoclonal antibodies revealed that an expanded RAS analysis can identify more sensitive responders to the antibodies,^8,9 proposing the use of cetuximab in only RAS wild-type patients. However, the selection for the responders to cetuximab still remains insufficient since a subgroup analysis of SWOG/CALGB80405 trial showed that survival time in RAS wild-type patients treated with cetuximab-containing regimen was almost same as that in RAS wild-type patients treated with bevacizumab-containing regimen.¹⁰ However, which biologic agent should be used up-front as the best companion for cytotoxic chemotherapy in RAS wild-type patients remains inconclusive. Further predictive markers beyond RAS are needed to refine the selection of patients who benefit from cetuximab-based chemotherapy and to prolong survival time in addition to improve cost effectiveness.

Several studies have reported that EGFR-related molecules may serve as predictors of cetuximab treatment, such as copy number of EGFR gene and the intratumoral mRNA expression of 2 EGFR ligands, amphiregulin (AREG) and epiregulin (EREG).^11-18 EGFR CA simple sequence repeat 1 (CA-SSR1) polymorphism has been shown to predict skin toxicity (Table 1). However, these biomarkers still remain not useful in clinical practice since they have been evaluated in various settings of chemotherapy or in population not restricted by KRAS status. Thence, it should be critical to accumulate evidences on if the molecules will serve as good predictors of cetuximab. In addition, there are few evidences in Japanese patients with regard to association of EGFR-related molecular markers with clinical outcome of chemotherapy with cetuximab. The aim of the present study was to clarify whether previously-reported EGFR-related biomarkers, AREG, EREG, EGFR gene copy number (GCN), and EGFR CA-SSR1 polymorphism, predict clinical outcomes in patients with mCRC harboring KRAS exon 2 wild-type tumors treated with first-line cetuximab-containing chemotherapy.

Table 1. Previously reported biomarkers and clinical outcome in patients receiving cetuximab.

Author (year)	Number of patients	KRAS status	Setting	Chemo regimen	Results
AREG/EREG
Khambata-Ford et al (2007)	103		Advanced line	Cet alone	AREG and EREG high associated with longer PFS
Jacobs et al (2009)	121	Wild-type	Advanced line	Cet alone or in combination with IRI	AREG and EREG high associated with better outcomes
Kuramochi et al (2012)	65	Wild type	none	none	EREG low associated with better prognosis
Stintzing et al (2012)	99	Wild type	First line	Combined with CAPOX or CAPIRI	AREG and EREG high associated with better outcomes
Pentheroudakis et al (2013)	199		First or advanced line	Cet plus chemo	AREG high associated with better survival time in KRAS wild-type EREG high associated with better outcomes irrespective of KRAS status
EREG
Jonker et al (2014)	385		Advanced line	Cet alone vs. BSC	Cet better in EREG high
Author (year)	Number of patients	EGFR cutoff	Setting	Chemo regimen	RR	PFS	OS	Skin toxicity
EGFR GCN
Moroni et al (2005)	31	≥ 3, < 3	Advanced line	Mono or combined with IRI	✓
Personeni et al (2008)	58	≥ 2.83, < 2.83	Advanced line	Mono or combined with IRI	✓	×	✓
Cappuzzo et al (2008)	85	≥ 2.92, < 2.92	Advanced line	Different regimens	✓	✓	×
Laurent-Puig et al (2009)	116	≥ 2, < 2	Advanced line	Mono or combined with IRI-based	✓	×	✓
Scartozzi et al (2009)	44	≥ 2.6, < 2.6	Advanced line	Combined with IRI-based	✓	×	✓
Tol et al (2010)	155	≥ 3, < 3	First line	Combined with CAPOX		×	×
Stintzing et al (2013)	98	>3.82, ≤ 3.82	First line	Combined with CAPOX or CAPIRI		×	×	×
Author (year)	Number of patients	CA repeat score	Setting	Chemo regimen	RR	PFS	OS	Skin toxicity
EGFR CA-SSR1 polymorphism
Graziano et al (2008)	110	S < 17, L ≥ 17	Advanced line	Combined with IRI-based	✓	✓	✓	✓^†
Lurje et al (2008)	117	S < 20, L ≥ 20	Advanced line	Monotherapy	×	×	×	×
Pander et al (2010)	123	S<20, L ≥ 20	First line	Combined with CAPOX		✓		×
Park et al (2011)	69	S<37, L ≥ 37	First and advanced line	Different regimens		×	×
Dahan et al (2011)	56	S ≤ 35, L > 35	Third line	Mono or combined with IRI	×	×	×	✓^†*
Stintzing et al (2013)	118	S ≤ 16, L>16	First line	Combined with CAPOX or CAPIRI		×	×	×

Cet; cetuximab, IRI; irinotecan, BSC; best supportive care, RR; response rate, PFS; progression-free survival, OS; overall survival

✓;significant, X; not significant ^†; S/S correlates with more ST, * P = 0.058

Results

The baseline characteristics of the patient cohort enrolled in this study are summarized in Table 2. In the population with a median age of 63 (range 39–79) years and follow-up time of 31.4 months, response rate, median progression-free survival (PFS), and overall survival (OS) were 73 % (95% CI 61–82%), 10.0 months (95% CI 8.8–11.8), and 33.9 months (95% CI 26.5-Not reached), respectively.

Table 2. Patients' characteristics (N = 77).

Characteristics	N	%
Gender
Male	44	57
Female	33	43
Age (year)
Median (range)	63 (39–79)
< 65	44	57
≥ 65	33	43
Performance Status
ECOG 0	69	90
ECOG 1	8	10
Diagnosis
Metachronous	18	23
Synchronous	59	77
Number of metastatic sites
1	33	43
≥ 2	44	57
Metastatic Sites
Liver	49	64
Lung	26	34
Para-aortic Lymph node	22	29
Peritoneum	17	22
Primary site of the tumor
Right	11	14
Left	64	83
NA	2	3
Previous adjuvant chemotherapy
Yes	6	8
No	71	92
Type of regimen
modified FOLFOX6	28	36
SOX	49	64

Abbreviations; ECOG, Eastern Cooperative Oncology Group; NA, not assessable.

AREG and EREG mRNA expression and clinical outcomes

Measuring mRNA expression of AREG and EREG were successful in 84 % and 81 %, respectively. In failed cases, the measurement was not successful because of limited quantity and/or poor quality of isolated total RNA. Patients with AREG high-expression (AREG mRNA level > 1.59) had a significantly longer PFS than those with AREG low-expression (median 11.6 months vs. 6.6 months, HR 0.52, 95% CI 0.28–0.98, P = 0.037) (Fig. 1); furthermore, OS was longer in patients with overexpression of AREG although there was no statistically significance (median 42.8 months vs. 26.5 months, HR 0.50, 95% CI 0.22–1.13, P = 0.084). No association of EREG mRNA expression with outcomes was observed (Table 3).

Figure 1. Probability of progression-free survival by AREG gene expression.

Table 3. Candidate predictors and clinical outcomes.

		Tumor response, RECIST		Early tumor shrinkage		Skin rash until 8 weeks		Progression-free survival		Overall survival
	N	CR, PR	SD, PD	Yes	No	Gr0-1	Gr2-3	Median (95%CI), months	HR (95%CI)	Median (95%CI), months	HR (95%CI)
AREG
≤1.59	13	7 (54%)	6 (46%)	10 (77%)	3 (23%)	7 (64%)	4 (36%)	6.6 (3.0, 13.8)	1	26.5 (5.7, 33.9)	1
>1.59	54	43 (83%)	9 (17%)	43 (83%)	9 (17%)	34 (71%)	14 (29%)	11.6 (9.4, 14.9)	0.52 (0.28, 0.98)	42.8 (27.1, 42.9+)	0.50 (0.22, 1.13)
P*			0.30		0.69		0.72		0.037		0.084
EREG
≤3.21	41	30 (75%)	10 (25%)	30 (75%)	10 (25%)	29 (78%)	8 (22%)	11.0 (8.8, 14.7)	1	33.9 (21.5, 41.1+)	1
>3.21	23	19 (86%)	3 (14%)	20 (91%)	2 (9%)	11 (58%)	8 (42%)	11.2 (7.0, 15.2)	0.90 (0.50, 1.60)	42.8 (27.1, 42.9+)	0.66 (0.29, 1.51)
P^*			0.98		0.18		0.13		0.70		0.29
EGFR GCN
Low	47	33 (75%)	11 (25%)	36 (82%)	8 (18%)	25 (61%)	16 (39%)	9.1 (5.8, 11.7)	1	27.1 (18.3, 42.8)	1
High	30	23 (79%)	6 (21%)	22 (76%)	7 (24%)	20 (77%)	6 (23%)	11.6 (9.2, 15.2)	0.63 (0.38, 1.07)	41.4+ (23.2, 41.4+)	0.52 (0.25, 1.08)
P^*			0.78		0.57		0.20		0.073		0.068
EGFR CA-SSR1
LL	36	24 (73%)	9 (27%)	25 (76%)	8 (24%)	20 (67%)	10 (33%)	11.1 (8.9, 13.6)	1	27.1 (19.0, 42.8)	1
SL	29	23 (79%)	6 (21%)	23 (79%)	6 (21%)	21 (81%)	5 (19%)	9.1 (6.6, 14.5)	0.88 (0.51, 1.52)	42.9+ (20.0, 42.9+)	0.62 (0.30, 1.30)
SS	12	9 (82%)	2 (18%)	10 (91%)	1 (9%)	4 (36%)	7 (64%)	15.2 (3.0, 17.4)	0.75 (0.34, 1.63)	39.8+ (12.6, 39.8+)	0.56 (0.19, 1.64)
P^*			0.75		0.63		0.040		0.74		0.32

* P values were based on Fisher's exact for tumor response, early tumor shrinkage, and skin rash until 8 weeks except for AREG and EREG based on maximal chi-square method for tumor response, log-rank test for PFS and OS in the univariate analysis.

Bold characters; significant, +; Estimates were not reached.

EGFR GCN and CA-SSR1 polymorphism and clinical outcomes

All enrolled patients were assessable for EGFR GCN and CA-SSR1 polymorphism. An increased GCN of EGFR was observed in 30 (39%) of 77 patients. EGFR GCN was not statistically significantly associated with clinical outcomes although EGFR high-GCN had trends toward longer PFS and OS (P = 0.073 and P = 0.068, respectively). Frequency of L/L, L/S, and S/S genotypes of EGFR CA-SSR1 were 47%, 38%, and 15%, respectively. The CA-SSR1 polymorphism did not predict survival of cetuximab-containing chemotherapy in all patients enrolled in this study. Then, we addressed at investigating association of the CA-SSR1 polymorphism with cetuximab-induced skin toxicity. We found that prevalence of grade 2 or 3 skin toxicity at 8 weeks after administration of treatment was significantly higher in patients harboring S/S genotype of the polymorphism than the other ones (P = 0.040) (Table 3).

Combined assessment of both AREG expression and EGFR GCN to predict clinical outcomes of cetuximab-containing chemotherapy

Additionally, we performed an exploratory sub-group analysis according to molecular status of both AREG mRNA expression and EGFR GCN, which are variables regarding EGFR ligand and receptor, respectively. PFS was significantly shorter in patients with either tumors harboring AREG low-expression or EGFR low-GCN compared to those with both AREG-high and EGFR high-GCN (median 9.4 months vs. 11.8 months, HR 1.75, 95% CI 1.00–3.08, P = 0.038). Patients with both AREG-low and EGFR low-GCN had significantly worse response rate and OS than the others (50% vs. 82%, P = 0.043; median 22.2 months vs. 42.8 months, HR 2.34, 95% CI 0.99–5.50, P = 0.042, respectively) (Table 4 and Fig. 2). No difference was observed in AREG expression level by EGFR GCN (P = 0.61, Wilcoxon 2-sample test).

Figure 2. Probability of overall survival by combined assessment with AREG gene expression and EGFR gene copy number (GCN), (A) Kaplan-Meier curves of 3 groups according to AREG gene expression and EGFR GCN, (B) Kaplan-Meier curves of groups with either AREG-high or EGFR GCN-high and both AREG-low and EGFR GCN-low.

Table 4. Association between AREG mRNA levels and EGFR gene copy number and clinical outcomes.

		Tumor response, RECIST		Early tumor shrinkage		Progression-free survival		Overall survival
	N	CR, PR	SD, PD	Yes	No	Median (95%CI), months	HR (95%CI)	Median (95%CI), months	HR (95%CI)
HH	26	20 (80%)	5 (20%)	19 (76%)	6 (24%)	11.8 (9.5, 16.3)	1	41.4+ (27.5, 41.4+)	1
HL	31	25 (83%)	5 (17%)	26 (87%)	4 (13%)	9.7 (6.7, 13.8)	1.63 (0.89, 2.99)	30.5 (18.4, 42.9+)	1.69 (0.71, 4.02)
LL	10	5 (50%)	5 (50%)	8 (80%)	2 (20%)	6.2 (3.0, 15.2)	2.15 (1.00, 4.64)	22.2 (4.4, 33.9)	3.18 (1.15, 8.79)
P*			0.11		0.63		0.082		0.067
HL, LL	41	30 (75%)	10 (25%)	34 (85%)	6 (15%)	9.4 (6.3, 13.6)	1.75 (1.00, 3.08)	30.5 (18.3, 42.9+)	1.98 (0.88, 4.47)
P*			0.77		0.51		0.038		0.086
HH, HL	57	45 (82%)	10 (18%)	45 (82%)	10 (18%)	11.2 (9.2, 14.7)	1	42.8 (27.1, 42.9+)	1
LL	10	5 (50%)	5 (50%)	8 (80%)	2 (20%)	6.2 (3.0, 15.2)	1.69 (0.85, 3.37)	22.2 (4.4, 33.9)	2.34 (0.99, 5.50)
P*			0.043		1.00		0.13		0.042

HH, AREG-high and EGFR high-GCN; HL, either AREG-high or EGFR high-GCN; LL, AREG-low and EGFR low-GCN

* P values were based on Fisher's exact for tumor response and early tumor shrinkage, log-rank test for PFS and OS in the univariate analysis.

Bold characters; significant, +; Estimates were not reached.

Factors with prognostic and/or predictive significance at multivariate analysis

A multivariate analysis adjusted for Eastern Cooperative Oncology Group (ECOG) performance status, regimen, and primary tumor site was performed for AREG, EREG, EGFR GCN, and CA-SSR1 polymorphism in all patients. High AREG mRNA levels was independently associated with better PFS and OS compared to patients with AREG-low levels (P = 0.027, P = 0.040, respectively). Molecular status with both AREG-low and EGFR low-GCN was a predictor of worse survival in patients treated with first-line cetuximab and oxaliplatin-based chemotherapy, irrespective of other factors (Table 5).

Table 5. Prognostic/predictive factors at multivariate analysis.

	Progression-free survival				Overall survival
	HR	95%CI	P-value^*	HR	95%CI	P-value^*
AREG
High vs. Low	0.48	0.25–0.92	0.027	0.40	0.17–0.96	0.040
EREG
High vs. Low	0.71	0.39–1.31	0.28	0.60	0.26–1.42	0.25
EGFR GCN
High vs. Low	0.73	0.41–1.30	0.29	0.49	0.21–1.11	0.087
EGFR CA-SSR1
LL vs. SL, LL	1.10	0.66–1.85	0.71	1.28	0.64–2.57	0.49
LL, SL vs. SS	1.50	0.69–3.26	0.31	1.33	0.45–3.91	0.60
AREG/EGFR GCN
HL, LL vs. HH	1.61	0.87–2.98	0.13	2.06	0.82–5.15	0.12
LL vs. HH, HL	1.77	0.86–3.65	0.12	3.73	1.47–9.46	0.006

* P values were based on Wald test for PFS and OS in the multivariable Cox regression model adjusted for regimen (modified-FOLFOX6 vs SOX), ECOG performance status (0 vs 1), and primary tumor site (right vs left).

Discussion

Our study demonstrates that AREG mRNA expression and EGFR CA-SSR1 polymorphism predict survival and skin toxicity, respectively, in Japanese patients with KRAS exon 2 wild-type mCRC treated with initial chemotherapy with cetuximab, potentially supporting previous data on the expression of AREG mRNA as a predictor for cetuximab. These results also suggest potential prognostic value of the combined assessment of AREG expression and EGFR GCN for cetuximab treatment.

Eleven ligands have been identified in the ErbB family in humans, 2 of which are AREG and EREG. ^19,20 Ligands binding to receptors induce the formation of receptor homodimers and heterodimers and activation of the intrinsic kinase domain, resulting in phosphorylation of specific tyrosine residues, which triggers intracellular signaling through the RAS/RAF/MAPK and PI3K/AKT pathways that subsequently modulates cell proliferation, adhesion, angiogenesis, migration, and survival. ^20,21 Several studies have shown that increased intratumoral mRNA expression of AREG or EREG gene is associated with clinical outcome in patients with KRAS exon 2 wild-type mCRC treated with cetuximab-containing therapy.^13,15,16,22 These studies enrolled cohorts with heterogeneous patient characteristics, type of therapy (single or combination), treatment in various lines and mixed KRAS status, while our study included a more homogeneous patient cohort that consisted of only KRAS exon 2 wild-type patients and treated with combination therapy with oxaliplatin as first-line treatment, probably indicating our findings become more reliable evidence. Recently, a sub-analysis of a big randomized clinical trial comparing irinotecan plus panitumumab with irinotecan as second-line treatment has been reported. ²³ The translational study results indicated that high expression of either AREG or EREG is a predictive marker for panitumumab therapy benefit on PFS in RAS wild-type patients, supporting results of our study using a homogeneous patient cohort, which could show significantly univariate and multivariate associations of AREG overexpression with better PFS as well as OS. Our finding supports that AREG expression may serve as a predictor of favorable outcomes of cetuximab treatment. On the other hand, we failed to find significant associations of EREG with clinical outcomes in the current study. Stahler A, et al has revealed the positive prognostic effect of high EREG expression but not AREG expression in patients treated with first-line irinotecan-based chemotherapy. ²⁴ EGFR ligands may have different prognostic effects between types of chemotherapy; therefore, there would be of interest to investigate the difference in future translational researches.

Cetuximab inhibited proliferation of colorectal-cancer cells with increased EGFR copy number, but that the same dose did not affect cells with unamplified EGFR.^25,26 EGFR GCN has been shown to correlate with better response to anti-EGFR antibodies; ^11,27-31 however, some studies have reported inconsistent results.^32-34 We failed to indicate statistically significant association of EGFR GCN with outcomes of cetuximab treatment. One of possible reasons for these divergent results is heterogeneity of copy number changes in colorectal cancer,^35,36 resulting in not only that cut-off value is different among previous studies but also that this molecular marker would be difficult to be included in clinical practice. In our study the increased GCN of EGFR was defined by the cut-off value, 2.92, based on a classification by receiver operating characteristic (ROC) analysis as previously described.²⁹ In addition, the patient number included in previous studies was relatively small. The association of EGFR GCN warrants to be validated in larger-size studies; however, EGFR GCN is less likely to become a clinically useful biomarker due to the difficulties in determining copy number by polymerase chain reaction in samples containing a mixture of somatic and tumor DNA and in standardizing the cut-off value.³¹

We found that short allele of EGFR CA-SSR1 predicts severe cetuximab-induced skin toxicity in Japanese patients with mCRC. No association with survival time was observed in our study. Some studies have previously reported significant association of long length of the EGFR CA repeat with worse outcomes of cetuximab treatment,^37,38 while the other studies have indicated no association,^39-41 leading to that it still remains controversial. Our results were consistent with previous reports that the length of the EGFR CA repeat inversely correlates with severe cetuximab-induced skin toxicity.³⁷ There has been shown to be interethnic differences in the repeat number of EGFR CA-SSR1 between Caucasians and Asians.⁴² In our study, frequency of the short allele was lower than that of the long allele, which is minor allele for Caucasians. To the best of our knowledge, this is the first report regarding distribution of the allele in Japanese mCRC patients.

Our exploratory sub-analysis revealed that patients harboring tumors with both AREG low-expression and EGFR low-GCN had statistically significantly worse survival than the other patients when treated with first-line cetuximab-containing chemotherapy. The association was observed in both univariate and multivariate analyses, suggesting potential prognostic value of the combined assessment of AREG mRNA and EGFR GCN for cetuximab treatment in mCRC. In the current study, overexpression of AREG affected outcome in term of OS, with a median OS of 42.8 months for AREG-high compared to 26.5 months for AREG-low (adjusted P = 0.040). The adjusted p-value of combined variables with AREG and EGFR GCN for OS was 0.006. Moreover, no association between AREG expression and EGFR GCN was observed (P = 0.61). In a previous study evaluating EGFR-related biomarkers, the combined assessment of EGFR pathway activations, which included receptor and downstream pathways such as RAS/MAPK, was not associated with clinical outcome.⁴³ In our study the exploratory analysis involved receptor-related marker and ligand-related marker, suggesting that combined factors of both the receptor and ligand may strongly influence survival in patients treated with cetuximab. Our findings also suggest that although each factor does not play a key role in decision-making for treatment the combined assessment may become a novel tool to distinguish patients who can receive more benefit from cetuximab.

Our study has some limitations. We demonstrated significant results for EGFR-related biomarkers as previously reported, and appeared to support their predictive value; however, sample size of our study was small, leading to some possibility that the patient number has no adequate ability to assess the association between the biomarkers and clinical outcomes. Although all patients were enrolled in a prospective trial, a selection bias cannot be excluded because the samples were collected retrospectively. The cut-off value of EGFR GCN and CA repeat still remains controversial, and the used cut-off value in our study is not validated. Additionally, we did not approach an analysis when restricted to RAS wild-type patients. Cetuximab should be administrated to RAS wild-type patients rather than KRAS exon 2 wild-type patients according several sub-group analyses of randomized clinical trials. ^9,44 Therefore, our findings should be confirmed in studies using larger cohorts with RAS wild-type patients.

In conclusion, our study demonstrates that AREG mRNA expression and EGFR CA-SSR1 polymorphism predict survival and skin toxicity, respectively, in KRAS exon 2 wild-type mCRC patients treated with cetuximab-containing chemotherapy as first-line treatment. Furthermore, our results suggest that the combined assessment of AREG expression and EGFR GCN may be a predictor of survival in mCRC patients treated with cetuximab.

Materials and methods

Patient population

We investigated mCRC patients with tumor expressing EGFR and harboring KRAS exon 2 wild-type, who participated in either Japanese phase II trial evaluating efficacy of first-line cetuximab in combination with oxaliplatin-based regimen, modified FOLFOX6 (JACCRO CC-05; n = 57, UMIN000004197) or SOX (S-1 plus oxaliplatin) (JACCRO CC-06; n = 67, UMIN000007022). A total of 77 patients with tumor tissue available from the 2 phase II trials were enrolled in this study. Tumor response was evaluated by computed tomography (CT) scan every 8 weeks until disease progression. External reviewers classified objective tumor response according to Response Evaluation Criteria in Solid Tumors (RECIST) v1.1. Patients with complete or partial response were categorized as responders, while those with stable or progressive disease were non-responders. Early tumor shrinkage (ETS) that was evaluated at the first CT scan at 8 weeks after starting treatment was recorded as yes when 20% and more decrease in the sum of diameters of target lesions, otherwise no. PFS was calculated from entry date to either progression of disease or death from any cause. OS was defined as the period from entry date until death. If events were not observed, the endpoints were censored at the last time of contact or follow-up. This study was conducted adhering to the REporting recommendations for tumor MARKer prognostic studies (REMARK).^45,46 The tissue analyses presented in this study were carried out at SRL, Inc. (Tokyo, Japan) and Response Genetics, Inc. (RGI; Los Angeles, CA, United States), following approval by the Institutional Review Board in each institute which participated in JACCRO CC-05 or CC-06 trial. Informed consent was obtained from all individual participants included in the study.

DNA and RNA isolation

Formalin-fixed paraffin-embedded (FFPE) tumor specimens were cut into sections with a thickness of 3 or 10 μm. In a preparation for macrodissection, one 3μm slide was stained with H&E and then evaluated for tumor content and marked for areas with dominant tumor foci by a pathologist. Macrodissection by scratching the marked areas was carried out using a blade to ensure that tumor cells as many as possible were dissected. The dissected particles of tissue were transferred to reaction tubes for isolation of genomic DNA and RNA. Genomic DNA was extracted from FFPE tissue derived from tumor samples using QIAamp DNA FFPE Tissue Kit (QIAGEN KK) according to the manufacturer's protocol. RNA isolation from macrodissected FFPE samples was performed using miRNeasy FFPE Kit (QIAGEN KK) according to the manufacturer's instructions. From the total RNA yielded, cDNA was converted using miScript II RT Kit (QIAGEN KK).

Quantitative reverse transcriptase polymerase chain reaction analysis for gene expression

Quantitation of gene mRNA expression levels of AREG, EREG, and an internal reference (β-actin) cDNA was done using a fluorescence-based Real Time PCR. Briefly, isolated RNA was reverse-transcribed to cDNA using random hexamers, followed by Real Time-PCR using specific primers and probes. For each sample, parallel reactions were carried out for each gene of interest and the β-actin reference gene to normalize for input cDNA. Real Time PCR was performed using the ABI PRISM 7900HT Sequence detection System (TaqMan; Perkin-Elmer Applied Biosystems). The obtained ratio between the values provided relative gene expression levels for the gene expression tested.

EGFR gene copy number and CA simple sequence repeat 1 polymorphism

One section with a thickness of 3 μm from FFPE tissue was used for a FISH assay to analyze EGFR GCN. The EGFR FISH assay was carried out with Vysis® LSI® EGFR SpectrumOrange / CEP® 7 SpectrumGreen Probe (Abbott Japan Co., Ltd.) according to methods of described PathVysion® HER-2 DNA Probe Kit (Abbott Japan Co., Ltd.). EGFR gene, which is located on the short arm of chromosome 7 was visualized as orange and green signal is α satellite of chromosome 7 and blue signal is nuclei respectively, with each filters through Monochrome CCD Camera Excel M (Dage-MTI, Inc.) using an automated fluorescence microscopy scanning system: Accord/SOLO (BioView Ltd.). The largest possible area of distant tumor areas were selected guided by the H&E-stained slide and the EGFR signal was counted in at least 30 nuclei per tumor area at X1000 magnification. Mean of ≥ 2.92 gene signals per nucleus was scored as EGFR FISH positive based on a classification by receiver operating characteristic analysis as previously described by Cappuzzo et al.²⁹ EGFR CA-repeat number was determined using PCR, followed by separation with capillary electrophoresis on ABI 3130xl Genetic Analyzer (Thermo Fisher Scientific K.K.). The assay with forward and reverse primers and sequences for analyzing the genetic variant were performed as described previously.⁴⁷ Based on the trimodal distribution of the EGFR CA repeat alleles in Asians,⁴² EGFR less than 20 CA repeats and ≥ 20 CA repeats were defined short (S) or long (L) alleles, respectively.

Statistical analysis

PFS was the primary endpoint of the current study. Tumor response (responder, non-responder), ETS (yes, no), and OS were the secondary endpoints. The associations between each categorical marker and tumor response or ETS were examined using Fisher's exact test. The maximal chi-square approach^48,49 was used to test associations between gene expression level of AREG or EREG and tumor response. A cut-off value was identified to separate patients into 2 groups in terms of likelihood of tumor response and P value was adjusted for multiple testing using 2000 bootstrap-like simulations.⁵⁰ The same cut-off value was used for assessing the associations with other endpoints. The associations between each marker and PFS or OS were assessed using Kaplan-Meir curves and log-rank test in the univariate analyses. Multivariable Cox regression model was performed to evaluate the independent effect of a marker on PFS or OS adjusting for regimen (modified FOLFOX6 vs. SOX), ECOG performance status (0 vs. 1), and primary tumor site (right vs. left).

Sixty-two of 77 patients had progressed or died when receiving the cetuximab-based first-line therapy. The minimum detectable hazard ratios ranged from 2.06 to 2.52 using a 2-sided log-rank test at the significance level of 0.05 with 80% power.

All tests were conducted using the SAS 9.4. (SAS Institute) at a significance level of 0.05. All P values were 2-sided and not adjusted for multiple hypothesis testing due to the nature of the current study.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1975 Helsinki declaration and its later amendments or comparable ethical standards.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank the patients, their families, and the investigators who participated in the JACCRO CC-05/06AR trial. We also thank Atsushi Kakimoto and Nahoko Hirabayashi (SRL, Inc., Tokyo, Japan) for genetic testing and Sachika Koyama for editorial assistance.

Funding

This study was partly funded by Dhont Family Foundation, San Pedro Peninsula Cancer Guild, and the Japan Clinical Cancer Research Organization (JACCRO).