Patients with anaplastic lymphoma kinase (ALK)-positive non–small cell lung cancer (NSCLC) respond to ALK inhibitors. Clinically, the presence of ≥15% cells with rearrangements identified on break-apart fluorescence in situ hybridization (FISH) classifies tumors as positive. Increases in native and rearranged ALK copy number also occur.
In total, 1426 NSCLC clinical specimens (174 ALK-positive specimens and 1252 ALK-negative specimens) and 24 ALK-negative NSCLC cell lines were investigated. ALK copy number and genomic status were assessed by FISH.
Clinical specimens with 0% to 9%, 10% to 15%, 16% to 30%, 31% to 50%, and >50% ALK-positive cells were identified in 79.3%, 8.5%, 1.4%, 2.7%, and 8.1%, respectively. An increased native ALK copy number (≥3 copies per cell in ≥40% of cells) was detected in 19% of ALK-positive tumors and in 62% of ALK-negative tumors. In ALK-negative tumors, abundant, focal amplification of native ALK was rare (0.8%). Other atypical patterns occurred in approximately 6% of tumors. The mean native ALK copy number ranged from 2.1 to 6.9 copies in cell lines and was not correlated with crizotinib sensitivity (50% inhibitory concentration, 0.34-2.8 μM; r = 0.279; P = .1764). Neither native or rearranged ALK copy number nor the percentage of positive cells correlated with extra-central nervous system progression-free survival in ALK-positive patients who were receiving crizotinib.
Anaplastic lymphoma kinase (ALK) gene rearrangements represent the primary oncogenic driver in a subset of lymphomas and solid tumors. In all of these rearrangements, the promoter and a region encoding a 5′ dimerization motif from another gene is fused to the region encoding the 3′ kinase domain of ALK, leading to constitutively active expression of ALK within the resulting fusion protein. In ALK-rearranged (ALK-positive) non-small cell lung cancer (NSCLC), the most common 5′ fusion partner is echinoderm microtubule-associated protein-like 4 (EML4), although other, less common 5′ partners have also been described.
ALK-positive NSCLC treated with crizotinib, a multitargeted tyrosine kinase inhibitor with activity against ALK as well as ROS1 and MET (hepatocyte growth factor receptor), is associated with a high response rate and prolonged progression-free survival (PFS).[2-4] Within the initial and ongoing crizotinib clinical trials in NSCLC, ALK status was determined by fluorescence in situ hybridization (FISH) using the Vysis break-apart probe set (Abbott Molecular, Des Plaines, Ill), which is the only companion diagnostic for crizotinib licensed by the US Food and Drug Administration (FDA) to date.[5, 6] The break-apart FISH testing involves DNA probes binding 5′ (green-labeled) and 3′ (orange-labeled, usually appearing as red in most microscope settings) of the common fusion breakpoint in ALK. When a rearrangement occurs, either a split or single red signal is noted in a cell (Fig. 1). Only a single rearranged signal is required to call a cell positive; however, not every cell in an ALK-positive tumor will reveal a detectable rearrangement by FISH, presumably in part because of false cellular negatives.[5, 7] Moreover, a proportion of cells even in normal tissues can appear to have a rearrangement by FISH because of presumed false cellular positives. The ALK FISH assay was original developed for the detection of ALK rearrangements in lymphomas, which are associated with a range of different 5′ fusion partners, primarily reflecting chromosomal translocations. Conversely, EML4, the most common fusion partner in NSCLC, resides on the same chromosome 2p arm as ALK, and the rearrangement is caused by a paracentric inversion. Consequently, lymphomas are associated with subtly different cytogenetic patterns of positivity than those observed in NSCLC. When the break-apart FISH assay was first modified for use in NSCLC, a natural gap in the continuum of the percentage of ALK-positive cells in lung tumors was identified that appeared to reliably distinguish between those assumed to be true-positive tumors and those in which it was assumed that the low positive cell counts reflected only the background noise of the assay. In the initial studies of crizotinib in NSCLC, >15% of tumor cells were required to indicate a rearrangement and to classify a tumor as ALK-positive.[2, 5, 7] Later, with the approval of the assay as a companion diagnostic by the FDA, ≥15% of tumor cells was adopted as the accepted cutoff point. In addition to the break-apart FISH assay, several other ALK diagnostic techniques also have been developed, including the use of immunohistochemistry (IHC) to search for the aberrant re-expression of the ALK protein and reverse transcriptase-polymerase chain reaction analysis to search for the presence of the abnormal fusion transcripts. Case reports have been published of tumors that were designated as ALK-negative by FISH but were determined to be ALK-positive using these other techniques that responded to ALK inhibitors, raising the possibility that the established FISH assay may miss an unquantified proportion of true-positive cases.[9-11] Because of the rarity of ALK-positive NSCLC, the identification of the FISH cutoff point to reliably distinguish true-positive tumors from true-negative tumors was inevitably based on a relatively small initial data set. With far more NSCLC specimens currently tested by ALK FISH, there is now the potential to more accurately re-explore whether the threshold value chosen still defines a true gap in the continuum of the assay and whether it should remain the sole determinant of ALK positivity in NSCLC.
In addition to the percentage of cells manifesting rearrangements, the break-apart FISH assay also provides information on the copy number per cell in both the native and rearranged ALK genes. The later development of rearranged ALK copy number gain (CNG) compared with baseline precrizotinib levels is one of several different identified mechanisms of acquired resistance to crizotinib.[12, 13] Yet, increases in copy number of both rearranged and native ALK relative to the diploid state in inhibitor-naive specimens also occur.[5, 14] In vitro, increases in the copy number of native ALK in NSCLC cell lines have been associated with crizotinib sensitivity in the 1 to 3 μM range. However, the clinical significance of baseline native/rearranged ALK copy number to crizotinib sensitivity at physiologic exposures remains unclear.
We previously demonstrated that neither the positive cell count, the baseline native copy number, nor the baseline rearranged copy number had any significant association with the maximal percentage shrinkage according to Response Evaluation Criteria in Solid Tumors (RECIST) (version 1.0) in ALK-positive tumors treated with crizotinib. However, not all of the clinical benefit from a drug may manifest as tumor shrinkage, and correlations between the different cytogenetic features of ALK positivity (cell count or copy number of native or rearranged signals) and PFS endpoints have not been reported previously in ALK FISH-positive tumors treated with crizotinib.
In this report, we explore whether the rearrangement of 15% of tumor cells reflects a clear biologic distinction in the frequency of ALK-positive cells in more than 1400 tested NSCLC tumors. We also explore the details of variation in native ALK copy number in tumors with and without ALK rearrangements, the significance of baseline native and rearranged copy number on PFS outcomes for treatment with crizotinib in ALK-positive patients, and the significance of native copy number on crizotinib outcomes in ALK-negative cell lines at 50% inhibitory concentrations (IC50s) compatible with clinically achievable exposures.
MATERIALS AND METHODS
Clinical NSCLC specimens that were tested for ALK between October 2008 and December 2010 at the Colorado Molecular Correlates (CMOCO) laboratory were identified and investigated. ALK copy number and genomic status (rearranged/native) were assessed using Vysis break-apart FISH, as previously described.[5, 7] Because of an institutional review board-approved protocol at the University of Colorado, we were permitted to correlate clinical data on patients who had molecular analyses conducted at the CMOCO laboratory.
Clinical Response and Progression-Free Survival Data
A subset of the tested patients received treatment at the University of Colorado and had clinical data on crizotinib outcomes available (n = 33). Patients received crizotinib treatment both within reported clinical studies and as standard of care after US Food and Drug Administration (FDA) licensing.[3, 4, 17] The best objective responses of target lesions according to RECIST version 1.0 were assessed in patients who had measurable disease after crizotinib treatment. Radiographic disease assessments were performed at baseline before patients commenced treatment with the inhibitor and continued after every other 21-day or 28-day cycle (depending on the clinical trial involved or physician preference if patients were treated off-study). PFS was calculated from the first dose of the crizotinib until the patient had either documented radiographic or clinical progression according to RECIST or a change in systemic therapy (other than discontinuation). Extra-central nervous system (eCNS) PFS was then calculated excluding CNS-only events. We used eCNS PFS as the primary PFS variable in the analyses because of the high CNS failure rate with crizotinib in ALK-positive NSCLC, which may represent low drug exposure more than biologic evolution.[4, 18, 19]
In Vitro Cell Line Sensitivity Data
Twenty-four ALK and ROS1 rearrangement-negative cell lines (A459, Calu-3, H1975, H2009, H2122, H322C, H3255, H358, HCC2279, HCC4006, HCC4011, HCC44, PC9, H125, H1703, H1299, H1334, H460, H661, H157, H226, H520, HCC95, and NE18) were analyzed by ALK FISH, and the mean copy number of the native ALK gene per cell was determined. Two ALK rearrangement-positive cell lines (H2228 and H3122) were used as positive controls. The IC50 of crizotinib was determined for each cell line using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) growth assay. Briefly, 1000 to 2000 viable cells were plated into 100 μL of growth medium in 96-well plates (Corning, Ithaca, NY) and incubated overnight at 37°C. Crizotinib 0 to 10 μM was added, and the plates were incubated for 5 days; then, tetrazolium salt was added to a final concentration of 0.4 mg/mL to each well. The medium was aspirated after 4 hours, and the reduced MTT product was solubilized by adding 100 μL of 0.2 N HCl in 75% isopropanol and 23% Milli-Qwater (Millipore Corporation, Billerica, Mass) to each well. Thorough mixing was done using a Finnpipette (Thermo Fisher Scientific, Waltham, Mass). The absorbency of each well was measured at 490 nm using an automated plate reader (Molecular Devices, Sunnyvale, Calif).
The difference between ALK-negative and ALK-positive cases in mean native ALK copy number was analyzed using a 2-sample t test. The frequency of native ALK copy number gain between ALK-negative and ALK-positive cases was analyzed using the Fisher exact test. Correlation between mean native ALK copy number and crizotinib sensitivity was analyzed using Spearman correlation. In the survival analysis with eCNS PFS as the outcome, native and rearranged ALK copy number and the percentage positive cell count were categorized with their respective quartiles; Cox proportional hazards regression was then used to analyze these associations. All statistical analyses were performed using SAS/STAT software version 9.3 of the SAS System for Windows (SAS Institute, Inc., Cary, NC).
In total, 1426 NSCLC clinical specimens that were tested for ALK rearrangements within the CMOCO laboratory (174 ALK-positive specimens and 1252 ALK-negative specimens according to standard criteria) were identified. In ALK-positive, crizotinib-treated patients, eCNS PFS data were available on 33 patients.
Clinical specimens with 0% to 9%, 10% to 15%, 16% to 30%, 31% to 50%, and >50% ALK-positive cells were observed in 79.3%, 8.5%, 1.4%, 2.7%, and 8.1% of NSCLC cases, respectively (Fig. 2). The proportions, in detail, among the ALK-negative cases are listed in Table 1.
Table 1. Proportion of ALK-Negative Patients With Different Percentages of Positive Cells Below a >15% Cutoff Point for Tumor Positivity
Cells With FISH-Positive Patterns
No. of ALK-Negative Patients (%)
Abbreviations: ALK, anaplastic lymphoma kinase; FISH, fluorescence in situ hybridization.
The mean native ALK copy number per cell was significantly higher in the ALK-negative group than in the ALK-positive group (mean ± standard deviation [SD], 2.8 ± 0.93 copies per cell [range 1.2-11.4 copies per cell] vs 1.8 ± 0.79 copies per cell [range, 0.6-5.3 copies per cell]; P < .0001) (Fig. 3). The frequency of native ALK copy number gain (defined as ≥3 copies per cell in ≥40% of cells) was significantly higher in ALK-negative tumors (62%) than in ALK-positive tumors (19%; P < .0001).
In ALK-negative tumors, although the native ALK copy number was elevated, the distribution of signals within the nucleus was largely diffuse, suggesting the occurrence of polysomy of chromosome 2 or partial polysomy of 2p23 (Fig. 4A). Abundant, focal amplification of native ALK was rare (0.8%) (Fig. 4B), whereas scanty amplification (<10% of tumor cells) occurred in 1.1% of tumors, and duplication of the entire native ALK or of the 3′ ALK end occurred in 3.5% (2.9% and 0.6%, respectively) (Fig. 4C,D). Single green signals (extra copies of the 5′ ALK with loss of 3′ ALK) were observed in 1.2% of these specimens.
Among ALK-negative lung cancer cell lines, the mean native ALK copy number per cell ranged from 2.1 to 6.9 and was not correlated with in vitro crizotinib sensitivity (IC50, 0.34-2.8 μM; r = 0.279; P = .1764) (Fig. 5A). In a 2-sided, unpaired t test, we determined that the IC50 for crizotinib was no different for cells that had native ALK copy number gain (≥3 copies per cell of native ALK in ≥40% of cells) compared with all other cell lines, defined as those with balanced or loss of native ALK copy number (P = .21) (Fig. 5B).
In ALK-positive patients who received treatment with crizotinib, neither native ALK copy number (P = .2859), nor rearranged ALK copy number (P = .4115), nor the percentage of positive cells (P = .0748) correlated with eCNS PFS on crizotinib.
Screening patients who have advanced NSCLC for an ALK rearrangement is now considered part of standard care based on the dramatic and prolonged objective responses observed when these patients are directed toward an ALK inhibitor like crizotinib.[2, 4, 17] Consequently, accurately identifying the patients who will derive most benefit from these inhibitors is imperative. Here, we describe our exploration of several different aspects of the ALK FISH break-apart assay, which is the only ALK diagnostic assay that is currently FDA approved.
Initial data suggested that a cutoff point of ≥15% or >15% of cells exhibiting an ALK rearrangement by break-apart FISH for classifying a tumor as ALK-positive would generate few borderline cases. For example, our own series had previously suggested that cell counts in positive tumors began at 22% and, in negative tumors/normal tissue, stopped at 11%. Specifically, the ≥15% cutoff point used in the initial crizotinib studies and the >15% cutoff point adopted by the FDA appeared to develop in a naturally occurring gap in the continuum of the assay. However, after analyzing a much larger series, we now perceive that considerable proportions of “negative” cases closely approach the established cutoff points (Fig. 2, Table 1). Overall, 8.5% of cases occur in the 10% to 15% range (9.7% of ALK-negative tumors). Although 15% would be considered positive in some studies and negative in others, the number of tumors with exactly 15% was very low; therefore, the overall conclusion is not changed depending on whether 15% or >15% is chosen as the criterion for positivity. Consequently, a second ALK diagnostic technique (eg, IHC) is recommended for such borderline cases to more accurately identify the true genomic status of ALK.
Because FISH analyses can also offer information about the copy number of both the native and rearranged ALK gene, several studies have already explored the importance of ALK copy number with regard to outcomes with crizotinib. It is known that baseline increases in copy number of both rearranged and native ALK relative to the diploid state occur, and associations between native copy number and sensitivity to high levels (1-3 μM) of crizotinib have previously been reported in vitro.[5, 14, 15] In our current analyses, we did confirm the frequent increase in native ALK copy number reported by others. It is interesting to note that the native ALK mean copy number was significantly higher in ALK-negative specimens than in ALK-positive specimens (mean ± SD: 2.8 ± 0.93 copies per cell [range, 1.2-11.4 copies per cell] vs 1.8 ± 0.79 copies per cell [range, 0.6 to 5.2 copies per cell]; P < .01) (Fig. 3). An increased native ALK copy number (≥3 copies per cell in ≥40% of cells) was detected in 19% of ALK-positive tumors and in 62% of ALK-negative tumors (P < .001). The lower native ALK copy number in ALK-positive NSCLC suggests that ALK rearrangement occurs early in tumorigenesis, preceding chromosomal instability, consistent with our previous studies. Because ALK positivity primarily occurs in individuals who have little or no smoking history, it is perhaps not surprising that these data suggest that ALK-positive lung cancer may reflect a “near-diploid” or genetically “simple” cancer.[5, 7]
We previously demonstrated no relation between either native/rearranged copy number or the percentage of cells positive for a rearrangement with the maximal tumor shrinkage according to RECIST observed in ALK-positive patients who were receiving crizotinib. Because not all clinical benefit may be manifest through the methods for determining tumor shrinkage used by RECIST, we attempted to correlate these same parameters with lack of tumor progression as assessed through the eCNS PFS of ALK-positive patients who received crizotinib. However, again, we were unable to demonstrate any significant correlation between eCNS PFS and either the native or rearranged copy number or the percentage of cells positive for a rearrangement in the tumor by FISH. With regard to rearranged signals, we previously demonstrated a strong correlation between the copy number of the rearranged gene and the percentage of cells counted as positive in the tumor, consistent with differences in cell counts in positive tumors largely reflecting missed positive cells and not true biologic negatives. Consequently, beyond the importance of the cell count in determining whether a tumor is positive or not in the first place, it is perhaps not surprising that PFS has no more correlation with these variables than tumor response. With regard to native copy number, additional discussion is required. Unlike the previous in vitro data obtained in experiments conducted at supraphysiologic exposures, our own data suggest no association between native copy number and sensitivity to crizotinib in ALK-negative cell lines treated at lower exposures that were more comparable to those used when studying ALK-positive cell lines (Fig. 5).[12, 15, 16]
When investigating the pattern of native copy number gain in ALK-negative patients, there is little suggestion of specific selection of this region of the genome by the cancers. Abundant, focal amplification of native ALK was rare in ALK-negative tumors (0.8%); scanty amplification (<10% tumor cells) occurred in only 1.1%, and most CNG occurred without significant amplification. However, duplication of the entire native ALK or of the 3′ and 5′ ALK occurred in 3.5% of ALK-negative tumors. Instead of simply reflecting the overall aneusomy of the cancer, such patterns may represent direct or inverse duplications or more complex rearrangements specifically involving the region of interest. In all such cases, it may be possible that the rearrangement still leads to the fusion of ALK with an activating partner. For example, Lipson et al described the in-frame fusion C2orf44:ALK in colorectal cancer, resulting from a 5-Mb tandem duplication at 2p23; and Peled et al reported an EML4:ALK fusion detected by next-generation sequencing and IHC in a specimen that exhibited a FISH pattern that was negative by standard criteria but that had complex, atypical features.[9, 20] Consequently, whereas native ALK copy number alone does not appear to be a useful predictor of ALK inhibitor sensitivity, the possibility that complex or atypical rearrangements that exist within the rare cases of focal native ALK duplication or amplification has to be considered.
Extra copies of 5′ ALK with loss of 3′ ALK are not expected to represent activation of the ALK gene by fusion, because the 5′ ALK signal represents exons 1 to 19 and does not include the tyrosine kinase domain. However, there are reports of NSCLC with extra copies of 5′ ALK and loss of 3′ ALK (appearing as single green signals with the break-apart FISH probes) in which it was determined that the tumors carried the EML4-ALK fusion. A possible explanation is that those tumors also have complex rearrangements in the 2p23-21 region, and the genomic region left in the area encompassed by the 3′ ALK probe only covers the ALK tyrosine kinase domain and is too small to be visually detectable in formalin-fixed, paraffin-embedded sections. Therefore, like the borderline-negative tumors according to the percentage cell count, “atypical” negatives (ie, tumors that do not meet the current break-apart positivity criteria but have evidence of potential complex rearrangements within the ALK locus, such as 3′ ALK doublets or single 5′ ALK) also should be considered for additional interrogation with a second ALK diagnostic technique. By capturing clinical outcomes from atypical negative and borderline negative FISH cases in which the second diagnostic suggests the true presence of an activating rearrangement, it should be possible to minimize missed therapeutic opportunities among the true inhibitor-sensitive, ALK-positive population.
This work was supported by the University of Colorado Lung Cancer Specialized Programs of Research Excellence (SPORE) grant (P50CA058187; Paul A. Bunn, Jr., principal investigator) and by the University of Colorado Cancer Center Shared Resources (Molecular Pathology) (CCSG P30CA046934). Dr. Baron was supported by a grant from the National Cancer Institute.
CONFLICT OF INTEREST DISCLOSURES
Dr. Camidge has received honoraria from Pfizer for participation in ad hoc advisory boards. Dr. Kiatsimkul is employed by The Physician Network as a hospitalist at the Good Samaritan Hospital (Kearney, Neb). Dr. Barón has received grants from the National Institutes of Health (NIH) as well as compensation from the NIH for study section participation. Dr. Aisner has received compensation as a consultant from Pfizer, GlaxoSmithKline, and Boehringer Ingelheim and has received lecture fees for an Association for Molecular Pathology webinar sponsored by Abbott Molecular. Dr. Doebele has received honoraria and research grants from Pfizer, consulting fees from Abbott Molecular and Boehringer Ingelheim, institutional grants from Eli Lilly and ImClone, and compensation for travel expenses from ImClone. Dr. Varella-Garcia has received honoraria for consulting and a research grant from Abbott Molecular, an institutional grant for an investigator-initiated study on colorectal cancer and for developing educational presentations from Abbot Molecular, and received lecture fees from Abbott Molecular.