Contribution of KRAS mutations and c.2369C > T (p.T790M) EGFR to acquired resistance to EGFR-TKIs in EGFR mutant NSCLC: a study on circulating tumor DNA

Introduction KRAS oncogene mutations (MUTKRAS) drive resistance to EGFR inhibition by providing alternative signaling as demonstrated in colo-rectal cancer. In non-small cell lung cancer (NSCLC), the efficacy of treatment with EGFR tyrosine kinase inhibitors (EGFR-TKIs) depends on activating EGFR mutations (MUTEGFR). However, inhibition of EGFR may select resistant cells displaying alternative signaling, i.e., KRAS, or restoration of EGFR activity due to additional MUTEGFR, i.e., the c.2369C > T (p.T790MEGFR). Aim The aim of this study was to investigate the appearance of MUTKRAS during EGFR-TKI treatment and their contribution to drug resistance. Methods This study used cell-free circulating tumor DNA (cftDNA) to evaluate the appearance of codon 12 MUTKRAS and p.T790MEGFR mutations in 33 advanced NSCLC patients progressing after an EGFR-TKI. Results p.T790MEGFR was detected in 11 (33.3%) patients, MUTKRAS at codon 12 in 3 (9.1%) while both p.T790MEGFR and MUTKRAS codon 12 were found in 13 (39.4%) patients. Six patients (18.2%) were KRAS wild-type (WTKRAS) and negative for p.T790MEGFR. In 8 subjects paired tumor re-biopsy/plasma samples were available; the percent concordance of tissue/plasma was 62.5% for p.T790MEGFR and 37.5% for MUTKRAS. The analysis of time to progression (TTP) and overall survival (OS) in WTKRAS vs. MUTKRAS were not statistically different, even if there was a better survival with WTKRAS vs. MUTKRAS, i.e., TTP 14.4 vs. 11.4 months (p = 0.97) and OS 40.2 vs. 35.0 months (p = 0.56), respectively. Conclusions MUTKRAS could be an additional mechanism of escape from EGFR-TKI inhibition and cftDNA is a feasible approach to monitor the molecular development of drug resistance.


INTRODUCTION
Activating mutations of epidermal growth factor receptor ( MUT EGFR) predict sensitivity to tyrosine kinase inhibitors (TKI) in non-small cell lung cancer (NSCLC). Despite a very high response rate (about 70%) to firstline treatment with the EGFR-TKIs (erlotinib, gefitinib or afatinib) in MUT EGFR NSCLC, tumors invariably progress after a median of 9-13 months from the beginning of treatment [1][2][3].
NSCLC heterogeneity can drive the therapeutic decisions [12]; therefore, tissue availability is increasingly recognized as a crucial issue. Unfortunately, the location of the tumor and the risk of complications are serious limitations to re-biopsies in NSCLC [13]. Alternatively, the detection of somatic mutations in cell-free tumor DNA (cftDNA) released in plasma could be instrumental for a better understanding of the genetic modifications driven by the selective pressure of drug treatments [14].
Interestingly, approximately 15-25% of patients with NSCLC have KRAS mutations ( MUT KRAS), resulting in constitutive activation of KRAS signaling pathways. MUT KRAS is a negative predictor of benefit to anti-EGFR antibodies in colo-rectal cancer, while it seems to be a negative predictor of response to EGFR-TKIs in EGFR wild type ( WT EGFR) NSCLC patients [15]. In a previous study on a large collection of NSCLC tissues from patients with acquired resistance to EGFR-TKI, NRAS or KRAS mutations were not demonstrated [10]. Despite these negative results, we employed a sensitive ddPCR-based platform to investigate the presence of MUT KRAS alleles in plasma of patients resistant to EGFR-TKIs and we were able to demonstrate a potential role of MUT KRAS in acquired resistance to EGFR-TKI, besides the p.T790M EGFR.
This finding reveals a potential new mechanism of resistance to EGFR-TKI and underscores the need of a periodic monitoring of somatic mutations of known oncogenes to deliver the best personalized treatment in a timely fashion.
The description of patients with activating MUT EGFR in their primary tumors as well as the percentages of p.T790M EGFR and MUT KRAS alleles in cftDNA at the time of EGFR-TKI progression is reported in Table 2. In 16 patients (48.5%), a codon 12 MUT KRAS was detected in cftDNA ( Figure 1). In addition to this, the p.T790M EGFR (c.2369C > T) second site mutation was present in the cftDNA of 24 patients (72.7%). Interestingly, 13 patients (39.4%) had both the MUT KRAS and p.T790M EGFR, while 3 (9.1%) and 11 (33.3%) subjects displayed only MUT KRAS or p.T790M EGFR, respectively ( Figure 1). Six subjects displayed no mutations.
The association between smoking habit and the occurrence of MUT KRAS in cftDNA was investigated. Regarding the 11 patients with smoking history, 2 (18.2%) Conclusions: MUT KRAS could be an additional mechanism of escape from EGFR-TKI inhibition and cftDNA is a feasible approach to monitor the molecular development of drug resistance.
In 8 patients, paired re-biopsies and cftDNA were available. The 8 re-biopsies were performed in a different tumor site with respect to the initial diagnosis, the choice being dependent on several factors, i.e., anatomical accessibility, new or progressing lesions. The analysis of re-biopsies by standard methods and ddPCR demonstrated p.T790M EGFR in 4 (standard) vs. 2 (ddPCR) samples and MUT KRAS in none (standard) vs. 3 (ddPCR) specimens. p.T790M EGFR and MUT KRAS were detected in 7 and 5 cftDNA specimens, respectively.
The analysis of KRAS status by ddPCR in the biopsies at diagnosis revealed the presence of MUT KRAS in 2 patients, who were WT KRAS by standard method. Therefore, the percentage of patients developing MUT KRAS as a mechanism of acquired resistance is 42% (14 patients). In terms of TTP and OS, there was no difference between these patients and the others (p = 0.19 and p = 0.13, respectively).
The concordance between tissue of re-biopsies (standard methods) and plasma (ddPCR), calculated by combining positive and negative results, was 62.5% and 37.5% for p.T790M EGFR and MUT KRAS, respectively (Table 3). Moreover, three paired samples found positive for the p.T790M EGFR on re-biopsies (standard methods) and on cftDNA (ddPCR), were negative by ddPCR on primary tissue ( Table 3).
The analysis of survival data stratified according to KRAS status in cftDNA, showed that TTP and OS were not statistically different, despite a trend towards a better survival in WT KRAS vs. MUT  BRAF and all the samples were found to be wild-type. Unfortunately, due to sample restriction, cftDNA was insufficient to perform the analysis of PI3K/Akt mutations.

DISCUSSION
The present study demonstrates the presence of MUT KRAS in the cftDNA of a significant proportion of patients progressing after EGFR-TKI treatment. In addition to this, the present study provides evidence that sensitizing MUT EGFR and MUT KRAS can coexist after the selective pressure of EGFR-TKI treatment.
p.T790M EGFR determines acquired resistance by increasing the affinity of EGFR to ATP [16,17]. p.T790M EGFR has been described in re-biopsies of 50-63% of tumors progressing under EGFR-TKI treatment [4,11] and in the cftDNA at a frequency similar to our study . Because drugs active on p.T790M EGFR, such as AZD9291 and rociletinib [21,22], are under clinical study and will be available soon in the clinical practice, the identification of this molecular marker is of utmost clinical relevance.
In our study, p.T790M EGFR was more frequent in L858R EGFR patients than in ex19del EGFR ones (80% vs. 67%); on the contrary, MUT KRAS in cftDNA was detected in 55% of patients with ex19del EGFR vs. 30% of patients with EGFR. To our knowledge, a mechanism of resistance depending on activating MUT EGFR has not been previously reported; however, this cohort is too small to draw any conclusion.
MUT EGFR and MUT KRAS are mutually exclusive in primary NSCLC and only anecdotal case reports described their coexistence [23,24]. MUT KRAS occurs in approximately 20% of NSCLC cases at diagnosis, more frequently in Caucasian population, adenocarcinomas, males and current smokers [25,26]. About 90% of KRAS mutations occur in exon 2 (codon 12 and 13), while exon 3 (codon 61) is less frequently involved [26,27]; in never-smokers with lung adenocarcinoma, MUT KRAS is more frequently a transition (G to A) compared to transversion in current smokers [25]. Colo-rectal cancer cells with MUT KRAS treated with anti-EGFR monoclonal antibodies are able to escape growth inhibition by several mechanisms, including MUT RAS [28]. While the role of MUT KRAS in primary resistance to EGFR-TKIs in molecularly unselected NSCLC is quite well established [29,30], its development and role in acquired resistance to EGFR-TKIs in MUT EGFR patients has not been explored in detail. In a previous work on a large collection of NSCLC tissues from patients with acquired EGFR-TKI resistance, MUT NRAS or MUT KRAS were not demonstrated [10]. However, comparison with the present results is not possible because detailed information were not provided neither on the timing of sampling with respect of development of TKI resistance nor on the type of tissue analysed. Therefore, we addressed this issue and a sensitive ddPCR-based platform was employed to investigate the presence of MUT KRAS alleles besides the well-known p.T790M EGFR. Due to its high sensitivity, ddPCR is able to identify small amounts of MUT KRAS and many methodological issues need to be addressed prospectically, particularly the threshold level of both MUT KRAS and p.T790M EGFR to be considered clinically relevant. However, a mechanism of drug resistance does not necessarily reflect biologic aggressiveness and the lack of difference in OS between KRAS wild-type and mutated patients it is therefore not surprising. The numeric dimension of the cell clone bearing MUT KRAS should be taken into consideration as well. Nevertheless, despite the low proportion of smokers in our cohort, the high prevalence of MUT KRAS could support its role of as mechanism of acquired resistance.
Eight patients underwent re-biopsy after tumor progression during EGFR-TKI, allowing a comparison between tissue and cftDNA. The detection of mutations in cftDNA but not in re-biopsy, using both standard methods and ddPCR, could suggest the presence of heterogeneity within metastatic sites or the lower performance of ddPCR in the presence of paraffin. Nevertheless, the detection of mutations in both plasma and tissue by ddPCR, but not by standard methods, could be due to the higher sensitivity of ddPCR analysis. Two patients, initially diagnosed WT KRAS by standard method, were re-analysed by ddPCR and were found MUT KRAS in the primary biopsy, suggesting that the MUT KRAS clone co-existed with activating MUT EGFR since the beginning, as also demonstrated in previous reports [23,24]. In these patients, MUT KRAS cannot strictly be considered a mechanisms of resistance but it could be possible that EGFR-TKI treatment may have favored the expansion of MUT KRAS-positive clones. However, conclusions cannot be drawn as pre-treatment cftDNA was not available. Our observation is a pivotal evidence of the presence of MUT KRAS in cftDNA of tumors with sensitizing MUT EGFR resistant to EGFR-TKIs. A small percentage of our patients received EGFR-TKI as third line therapy and a new biopsy was not repeated at this time. It could be possible that MUT KRAS appeared before the initiation of TKI as a mechanism of resistance to previuos therapy. This hypotesis is weakened by the evidence that patients given second or further lines of therapy showed TTP and OS similar to patients treated in first line, although a mechanism of resistance does not necessarily affect survival. The presence of MUT KRAS has been recently reported using next generation sequencing analysis of tumor re-biopsies after progression under EGFR-TKI treatment [31], similarly to colo-rectal cancer treated with EGFR antibodies [32,33]. It remains to be determined if the presence of p.T790M EGFR and MUT KRAS coexist in the same tumor cell or arise in different subclones.
Targeting MUT KRAS proteins is still a challenge [34,35]. Theoretically, combined treatment with KRAS and EGFR inhibitors can be administered to patients to prevent MUT KRAS-dependent resistance or restore sensitivity to EGFR-TKIs, as recently demonstrated co-targeting EGFR and MEK [36]. To date, p.T790M EGFR remains the most important predictor of efficacy of third generation EGFR-TKIs. Moreover, it was found that the coexistence of both activating MUT EGFR and MUT KRAS was not necessarily a negative predictor for EGFR-TKI therapy [23]. With these evidences in mind, all patients with p.T790M EGFR should receive third generation EGFR-TKI, even in the presence of MUT KRAS. Theoretically, it is possible that MUT KRAS identify a less responsive subgroup of patients but this hypothesis should be validated by monitoring patients prospectically during secondline therapy. Beside p.T790M EGFR and MUT KRAS, other mechanisms of acquired resistance not evaluated in our study have been described in tumor re-biopsies after EGFR-TKI progression, including actionable mutations of MET [7], HER2 [8], PIK3CA [9] or transformation into small cell histology (3%) [11].
In conclusion, despite the small number of patients involved, the retrospective analysis and the low rate of re-biopsies, our results confirm the importance of cftDNA analysis for the monitoring of secondary mutations associated with EGFR-TKI resistance in NSCLC and underline the role of a highly sensitive approach, i.e., ddPCR, for the detection of low-level mutations. The clinical relevance of these findings, expecially for what concerns MUT KRAS, needs to be evaluated prospectively. These observations open new perspectives in the molecular mechanisms of acquired resistance, indicating a possible role of MUT KRAS in tumor escape from pharmacologic treatment. The effect of MUT KRAS in NSCLC with activating MUT EGFR needs to be further elucidated at the molecular level and encourages the development of inhibitors MUT KRAS for optimal treatment of patients.

Study population
-NA --Mut -Mut -Mut: presence; "-": absence of mutation; NA: sample not available; std: standard sequencing approach. submitted and approved by the Ethics Committee of Pisa University Hospital and conducted in accordance to the principles of the Declaration of Helsinki; all patients gave their signed informed consent before blood collection and cftDNA analysis.

Plasma collection and cftDNA extraction
Six ml of blood were collected in EDTA and centrifuged at 4°C for 10 min at 3000 rpm within two hours after blood drawing. Plasma samples were stored at −80°C until analysis. cftDNA was extracted using a QIAmp Circulating nucleic acid Kit (Qiagen ® , Valencia, CA, USA) from 1 to 3 ml of plasma following the manufacturer's protocol and the DNA was eluted in 100 μl of buffer.

Analysis of cftDNA
The investigational part of this study included the assessment of codons 12, 13 (p.G13D) and 61 MUT KRAS and MUT NRAS, p.T790M EGFR, V600E BRAF in cftDNA. Other mutations potentially associated with EGFR-TKI resistance were not examined because of the limited amount of cftDNA available. The analysis of cftDNA was performed by digital droplet PCR (ddPCR, BioRad ® , Hercules, CA, USA) and ddPCR Mutation Assay (BioRad ® ). The analytic procedure was unable to discriminate the nature of the MUT KRAS because the analysis was performed with a ddPCR KRAS Multiplex assay.
PCR reactions were assembled into individual wells of a single-use injection molded cartridge, according to the following protocol: 20 ng of template DNA (4 μl), 1 μl of 20X target primer/probe assay (FAM), 1 μl of 20X wild type primer/probe assay (HEX), 10 μl of 2X ddPCR Super Mix and 4 μl of DNAse/RNAse-free water up to a total volume of 20 μl. Droplet generation oil (70 μl) was then loaded and the cartridge was placed into the droplet generator. Using vacuum, sample and oil were mixed, generating mono-disperse droplets. Thereafter, 40 μl of packed droplets were transferred into a 96-well PCR plate for thermal cycling amplification. The protocol was standardized for all mutations to the following conditions: 95°C × 10 min, 94°C × 30 s and 55°C × 60 s (35 cycles), 98°C × 10 min, and 4°C hold. The droplet reader (BioRad ® ) was used for fluorescence signal quantification. The concordance between MUT KRAS and p.T790M EGFR was assessed on pairwise cftDNA and tissue DNA of 8 patients who underwent re-biopsy for diagnostic purposes. DNA was extracted from formalin-fixed paraffin-embedded biopsies using the QIAmp DNA Mini Kit (Qiagen ® ) and analyzed using conventional diagnostics as reported above. As a positive control for MUT KRAS, the cftDNA from 30 patients with known MUT KRAS pancreatic cancer was used, while the DNA extracted from plasma of 43 healthy blood donors was employed as negative control for MUT KRAS and p.T790M EGFR.

Data analysis
TTP and OS were calculated following standard procedures and survival curves were generated by the SPSS statistical software (version 16).

Financial support
This study was supported in part by grants from the Italian Association for Cancer Research (AIRC, Italy), MIUR and ITT (Italy) and Fondazione Berlucchi (2014) to RD; AIRC (grant #12182), Cam-Pac FP7 (grant #602783) and FIMP-Ministero Salute (CUP_J33G13000210001) to AS.