SEARCH

SEARCH BY CITATION

Keywords:

  • anaplastic lymphoma kinase;
  • fluorescence in situ hybridization;
  • copy number;
  • crizotinib

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. FUNDING SOURCES
  6. REFERENCES

BACKGROUND:

Fluorescence in situ hybridization (FISH), using break-apart red (3′) and green (5′) ALK (anaplastic lymphoma kinase) probes, consistently shows rearrangements in <100% of tumor cells in ALK-positive (ALK+) nonsmall cell lung cancer (NSCLC). Increased copy numbers of fused and rearranged signals also occur. Here, correlations are explored between the percentage of ALK+ cells and signal copy number and their association with response to ALK inhibition.

METHODS:

Ninety ALK+ NSCLC cases were evaluated. The percentage of positive cells, pattern of positivity (split, single red, or both), and copy number of fused, isolated red and green signals were recorded. Thirty patients had received crizotinib.

RESULTS:

Increased isolated red signal copy number (contributing to both single red and split patterns of positivity) correlated with a higher percentage of ALK+ cells (r = 0.743, P < .0001). Mean fused copy number was negatively associated with isolated red signal copy number (r = −0.409, P < .0001). Neither percentage of positive cells (r = 0.192, P = .3), nor copy number of isolated red signal (r = 0.274, P = .195) correlated with maximal tumor shrinkage with crizotinib.

CONCLUSIONS:

The strong association between increased copy number of key ALK signals and percentage of positive cells suggests that the <100% rate of cellular positivity in ALK+ tumors is due to technical factors, not biological factors. In ALK+ tumors, neither the percentage of positive cells nor signal copy number appear to be informative variables for predicting benefit from ALK inhibition. The inverse relationship between fused and isolated red copy number suggests ALK+ may be a distinct “near-diploid” subtype of NSCLC that develops before significant chromosomal aneusomy occurs. Cancer 2012. © 2012 American Cancer Society.

Activation of anaplastic lymphoma kinase (ALK) has been described as a primary oncogenic event in a number of different human cancers.1 In most cases, activation is through a chromosomal rearrangement that places one of several different 5′ fusion partners and their associated promoter upstream of the 3′ kinase domain of ALK.2 Rearrangements of the ALK gene occur in 3% to 5% of nonsmall cell lung cancers (NSCLCs), most notably in adenocarcinomas.1 The most common 5′ fusion partner in NSCLC is EML4 (echinoderm microtubule-associated protein-like 4), but other, rarer 5′ fusion partners, notably KIF5B (kinesin family member 5B) and TFG (TRK-fused gene), have also been described.3-5 There are various potential methods for detecting ALK positivity in solid tumors, including use of immunohistochemistry to look for aberrant expression of the kinase domain, using reverse transcription polymerase chain reaction with primers directed against specific fusion pairings and use of fluorescence in situ hybridization (FISH) with break-apart probes that bind 5′ and 3′ of the common breakpoint in ALK to look directly for a gene rearrangement.1, 6

NSCLC patients proven to have an ALK gene rearrangement, with FISH as the screening technique, commonly manifest rapid, dramatic, and prolonged responses to crizotinib, a small-molecule inhibitor of ALK and MET.7, 8 Within all of the crizotinib studies, ALK positivity as determined by FISH has been defined as >15% of tumor nuclei demonstrating either split and/or single 3′ (red in the most commonly used probe-color pairing) fluorescence patterns.7, 8 To count as positive for a rearrangement with a split pattern, splitting of the 5′ (green) and 3′ (red) signals must occur by more than 2 signal diameters. Single 5′ (green) patterns and increased copy numbers of native (fused) and rearranged signals have also been reported, but their biological significance, beyond simply reflecting the chromosomal instability of many cancers, remains uncertain.9, 10

We have previously shown that the proportion of positive cells within an overall ALK-positive (ALK+) tumor covers a broad range greater than 15% but below 100% (eg, 22%-87%).9 Although some have interpreted such findings as evidence against ALK gene rearrangements being both a primary and/or necessary oncogenic event in NSCLC, other explanations exist.11 In conjunction with the presence of positive signals in up to 11% of nontumor and negative tumor cells, we hypothesized that most of the variation above zero in the negative and nontumor areas and the variation below 100% in the positive tumor areas reflected technique rather than biology.9 Specifically, individual cells may be counted as positive but not possess a true rearrangement (false positives) or may be counted as negative when they do in fact possess a rearrangement (false negatives).

If cellular heterogeneity is related to technique and not to biology, we hypothesized that the following would be true. First, because a single red pattern may be easier to detect than a split pattern, the overall proportion of ALK+ cells should be higher in tumors expressing the single red pattern of positivity than the split pattern. Second, it should be harder to miss a true positive pattern when the number of signal copies is increased, given that only one split red/green or single red pattern per cell is required to call a cell positive, and therefore there should be a positive correlation between the number of 3′ (red) signals (contributing to both single red and split patterns of positivity) and the percentage of ALK+ cells in an ALK+ tumor. In contrast, fused signal copy number should not be associated with the proportion of ALK+ cells. Because the 5′ (green) signal contributes to both positive (split) and negative (single green) patterns, correlation with ALK+ may not be apparent for the overall 5′ signal copy number, but should be obvious when only those cases with a split pattern of positivity are analyzed. Finally, there should be no correlation between the percentage of ALK+ cells present in any ALK+ tumor and the maximal percentage of tumor shrinkage, as assessed by Response Evaluation Criteria In Solid Tumors (RECIST), seen in that tumor after therapy with crizotinib. To test these hypotheses, as well as the potential biological significance of ALK copy number alterations in ALK+ tumors, we explored correlations between the percentage of ALK+ cells, patterns of ALK positivity, and signal copy number and whether these variables are associated with response to ALK inhibition in ALK+ NSCLC.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. FUNDING SOURCES
  6. REFERENCES

The University of Colorado Thoracic Oncology Program performs routine molecular screening for multiple molecular abnormalities including EGFR (epidermal growth factor receptor) and KRAS (v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog) mutations and ALK gene rearrangements on all NSCLC cases with available tumor tissue. Testing is performed in-house within the Clinical Laboratory Improvement Amendments (CLIA)-certified Colorado Molecular Correlates (CMOCO) laboratory for the purposes of directing patients toward the most appropriate targeted therapy. Screening for ALK gene rearrangements by FISH began in late 2008 with the goal of identifying patients for entry into defined molecular cohorts within the phase 1 study of crizotinib (PF-02341066).7 Initially, only patients with metastatic disease were screened, but later, beyond the crizotinib study, all patients with NSCLC were screened regardless of stage from the first point of contact with the program. Testing for ALK with FISH, using the LSI ALK Dual Color, Break-Apart Rearrangement Probe (Abbott Molecular), designed to detect rearrangements in chromosome 2p23 encompassing the ALK gene, was conducted as described.7, 9 This probe set included a 250-kilobase (kb) DNA fragment telomeric to ALK (3′ end) labeled in SpectrumOrange and a 300-kb fragment centromeric to ALK (5′ end) labeled in SpectrumGreen. Signals were enumerated in at least 50 tumor nuclei in standard 4- to 5-μm thick sections, using an epifluorescence microscope with single interference filters sets for green (fluorescein isothiocyanate), red (Texas red), and blue (4′,6-diamidino-2-phenylindole) as well as dual (red/green) and triple (blue, red, green) band-pass filters. Tumor areas were marked by a lung pathologist in a hematoxylin-and-eosin–stained slide prior to FISH analysis to facilitate the identification of tumor-rich regions by the FISH technologist. Because the 3′ ALK orange probe is detected by an interference filter with emission/excitation within the red wavelength, its fluorescence signal is seen as red and identified as a “red” signal in this study.

On FISH testing, the occurrence of an ALK gene rearrangement was concluded if >15% of tumor cells showed a split red and green and/or single red (residual 3′) pattern, otherwise the specimen was classified as ALK FISH–negative. Tumor specimens from the first 90 ALK+ patients with NSCLC were examined. The percentage of tumor cells positive for a rearrangement, the pattern of positivity (split, single red or both), and the copy number of the fused, isolated red and isolated green signals contributing to these patterns were recorded (Fig. 1A,B). Of note, throughout this article, the term “signal” is used to refer to the individual spot representing homology to a specific probe sequence, whereas “pattern” is used to refer to how the different signals present in a given tumor cell. In addition, the term “single” is reserved for the description of a specific pattern, whereas “isolated” is used in reference to a specific probe signal. For example, an isolated red signal may be part of a split pattern or a single red pattern of positivity, depending on whether an isolated green signal either does or does not coexist in the same cell.

thumbnail image

Figure 1. (A) Schematic diagram is shown of the different cellular positive (split or single red) and negative (fused or single green) patterns for ALK (anaplastic lymphoma kinase) status testing with fluorescence in situ hybridization testing. (B) Schematic diagram exhibits how individual fused, isolated red and isolated green signals are counted and contribute to the different positive and negative patterns seen.

Download figure to PowerPoint

Of the 90 ALK+ patients, 30 received crizotinib (PF-02341066) within the phase 1 study and had available response outcomes on therapy.7, 8 Best objective responses of target lesions per RECIST, version 1.0, were assessed on patients with measurable disease following crizotinib treatment. Radiographic disease assessments were performed at baseline and continued after every other 28-day cycle, or if clinically indicated, until disease progression as determined by the clinic physician. The lowest sum of the diameter of target lesions at any assessment while on treatment was calculated against baseline measurements to determine maximum tumor response (percent tumor shrinkage). RECIST response data were then correlated with the individual's ALK profile (percent ALK+ cells, signal pattern, and signal copy number).

A protocol approved by the institutional review board at the University of Colorado permitted clinical correlates to be made on all in-house patients in whom molecular analyses have been conducted within the CMOCO laboratory. In addition, patients within the crizotinib phase 1 study signed informed consent for treatment and data capture. The protocol for the crizotinib phase 1 study was approved by the institutional review board at each institution.

Linear associations between the percentage of ALK+ cells and signal copy number or response to therapy were summarized by Spearman correlations using 2-sided P values. A 2-sided Student t test was performed to compare the average percentage of ALK+ cells between different patterns of positivity (split vs single red).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. FUNDING SOURCES
  6. REFERENCES

Percentage of ALK+ Cells According to Patterns of Positivity and Correlation With Signal Copy Number

The mean percentage of ALK+ cells was 56% (range: 18%-100%) in the dataset of 90 ALK+ patients and 59% (24%-94%) in the subset of 30 crizotinib-treated patients. In the overall dataset, 49 patients (54%) demonstrated a split signal pattern of positivity, 33 patients (37%) demonstrated a single red pattern of positivity, and 8 patients (9%) demonstrated aspects of both patterns (Fig. 1). In the crizotinib-treated dataset, 18 patients (60%) demonstrated a split signal pattern of positivity, 10 patients (33%) demonstrated a single red pattern of positivity, and 2 patients (7%) demonstrated aspects of both patterns.

The mean and range of cells positive by pattern of positivity is shown in Table 1. Tumors with a single red pattern of positivity had a significantly higher mean percentage of ALK+ cells than tumors with a split pattern of positivity (74% [26%-100%] vs 48% (18%-82%), respectively; P < .0001). However, both patterns demonstrated a similarly wide range of percentage positive cells.

Table 1. Correlation Between Percent of ALK+ Cells and Mean Copy Number per Cell of Fused (FCN), Isolated Red (IRCN), and Isolated Green (IGCN) Signals
Pattern of ALK Rearrangement PositivityNVariable% ALK+ vs Mean FCN
   MeanMinimumMaximumrP
All ALK+90%ALK+0.560.181−0.466<.0001
FCN1.770.565.25
ALK+ with split pattern49%ALK+0.480.180.82−0.440.002
FCN1.770.565.24
ALK+ with single red pattern33%ALK+0.740.261−0.042.016
FCN1.570.683.63
ALK+ with mixed pattern8%ALK+0.370.250.6−0.619.102
FCN2.5415.25
Pattern of ALK Rearrangement PositivityNVariable% ALK+ vs Mean IRCN
   MeanMinimumMaximumrP
All ALK+90%ALK+0.560.1810.743<.0001
IRCN0.660.012.32
ALK+ with split pattern49%ALK+0.480.180.820.558<.0001
IRCN0.510.011.86
ALK+ with single red pattern33%ALK+0.740.2610.884<.0001
IRCN0.950.022.32
ALK+ with mixed pattern8%ALK+0.370.250.60.898.002
IRCN0.410.280.66
Pattern of ALK Rearrangement PositivityNVariable% ALK+ vs Mean IGCN
   MeanMinimumMaximumrP
  1. ALK+ indicates positivity for anaplastic lymphoma kinase; FCN, fused copy number; IGCN, isolated green signal copy number; IRCN, isolated red signal copy number.

All ALK+90%ALK+0.560.181−0.264.012
IGCN0.3601.92
ALK+ with split pattern49%ALK+0.480.180.820.910<.0001
IGCN0.580.161.92
ALK+ with single red pattern33%ALK+0.740.261−0.492.004
IGCN0.0600.26
ALK+ with mixed pattern8%ALK+0.370.250.60.762.028
  IGCN0.260.080.4  

The correlations between the percentage of ALK+ cells, overall and by different patterns of positivity (split, single red, and mixed), and the mean fused, isolated red and isolated green signal copy number are shown in Table 1. Increased copy number of isolated red signal (contributing to both single red and split patterns of positivity) strongly correlated with a higher percentage of cells positive for a rearrangement (r = 0.743, P < .0001) (Fig. 2). The correlation with isolated red signal copy number gain was stronger for the single red pattern than for the split pattern of positivity (r = 0.884, P < .0001 vs r = 0.558, P < .0001, respectively).

thumbnail image

Figure 2. Correlation between the percentage of cells positive for anaplastic lymphoma kinase (ALK+) and the isolated red signal copy number in ALK+ tumors. The ellipse represents the 95% bivariate normal prediction area for a new observation.

Download figure to PowerPoint

Isolated green signal copy number (contributing to the split pattern of positivity and the single green negative pattern) was loosely associated with a lower percentage of cells positive for a rearrangement (negative association with higher percentage) (r = −0.264, P = .012). However, when only those tumors with a split pattern of positivity were assessed, the correlation between percentage of positive cells and isolated green signal copy number was strongly reversed (r = 0.91, P < .0001).

Fused signal copy number was negatively associated with the percentage of cells positive for a rearrangement (r = −0.466, P < .0001), particularly if the tumor was positive with a split as opposed to a single red pattern of positivity (r = −0.44, P = .002 vs r = −0.042, P = .016).

Correlations Between Copy Number of Different Signals (Fused, Isolated Red and Isolated Green)

The correlations between the mean copy number per cell of fused signals and isolated red and isolated green signals are shown in Table 2. Although there was no correlation between mean fused and isolated green signal copy number (all patterns), mean fused copy number was negatively associated with isolated red signal copy number (r = −0.409, P < .0001). Because the isolated green signal is not specifically associated with ALK positivity per se, contributing as it does to both positive (split) and negative (single green) patterns, we reanalyzed the correlation between mean fused and isolated green signal copy number looking only in tumors with split as opposed to single green patterns of positivity. A similar inverse relationship to that seen with isolated red signal copy number was noted between the isolated green (split pattern) copy number and the fused signal copy number; however, it did not reach statistical significance (r = −0.425, P = .13).

Table 2. Correlation Between Mean Copy Number per Cell of Fused Signals Versus Isolated Red and Isolated Green Signals
Variable (Copy Number per Cell)NMeanMinimumMaximumCorrelation to Mean Fused Signal Copy Number
Fused signal901.770.565.25rP
Isolated red signal0.660.012.32−0.409<.0001
Isolated green signal0.3601.92−0.007.951

Correlations Between Percentage of Positive Cells, Signal Copy Number, and Maximal Percentage Tumor Shrinkage Upon Crizotinib Treatment

Among the 30 crizotinib-treated patients, the mean maximal tumor shrinkage per RECIST was 58% (range: 0%-100%). Given the small numbers in the subgroups, there appeared to be little significant difference in the extent of tumor shrinkage seen by pattern of positivity: split (n = 18; mean = 58%; 0%-100%), single red (n = 10; mean = 65%; 31%-100%), and mixed (n = 2; mean = 24%; 16%-32%).

Neither percentage of positive cells (r = 0.192, P = .3) (Fig. 3) nor isolated red (r = 0.274, P = .195) (Fig. 4) nor isolated green (split pattern) (r = 0.438, P = .117) nor fused signal (r = −0.247, P = .187) copy number correlated with the maximal percentage tumor shrinkage as assessed per RECIST with crizotinib treatment.

thumbnail image

Figure 3. Correlation between the percentage of cells positive for anaplastic lymphoma kinase (ALK+) and the maximal percentage tumor shrinkage as assessed by Response Evaluation Criteria In Solid Tumors (RECIST) with crizotinib in ALK+ tumors. The ellipse represents the 95% bivariate normal prediction area for a new observation.

Download figure to PowerPoint

thumbnail image

Figure 4. Correlation between the isolated red signal copy number and the maximal percentage tumor shrinkage as assessed by Response Evaluation Criteria In Solid Tumors (RECIST) with crizotinib in tumors positive for anaplastic lymphoma kinase (ALK). The ellipse represents the 95% bivariate normal prediction area for a new observation.

Download figure to PowerPoint

Conclusions

ALK gene rearrangements in NSCLC are highly predictive of clinical benefit from treatment with crizotinib, a small-molecule inhibitor of ALK and MET.7 In most cases, treatment with crizotinib results in tumor shrinkage. The objective response rate in the first-in-human phase 1 study of crizotinib was 61%, with a median duration of response of 48 weeks.8 In addition, when the waterfall plot of RECIST-defined changes in tumor size on treatment is examined, although not everyone reached the criteria for a RECIST-defined objective response, nearly 90% of patients manifested some degree of tumor shrinkage from the drug.8 Within the phase 1 and other studies of crizotinib in ALK+ NSCLC, ALK positivity is determined using FISH with the Abbott Molecular break-apart probe set as the screening technique.

We have previously shown that there is a broad range of percentage of positive cells within tumors meeting the >15% criterion for positivity.9 Using separate FISH probes for both EML4 and ALK, Martelli et al similarly noted EML4-ALK rearrangements in 50% to 100% of cells of their ALK+ tumors.11 Because the presence of EML4-ALK alone is sufficient to act as a transforming event, the fact that not all cells in the tumor appear to demonstrate a gene rearrangement initially seems confusing.3 Based on their findings, Martelli et al concluded that ALK rearrangements may represent a late, rather than the primary, oncogenic event, resulting in the coproliferation of different ALK+ and ALK-negative clones within established tumors.11 The alternative explanation why ALK+ tumors do not show positivity in all cells is that there is a consistent cell-by-cell false negative rate. Consistent with this is the observation that FISH analysis of cell pellets derived from ALK+ cell lines, such as NCI-H3122 and NCI-H2228, also do not show rearrangements in every cell (data not shown). In theory, such false negatives could occur due to compression or folding of the DNA that randomly affects the distance between the sequences homologous to the red and green probes, nuclear sectioning causing loss of the 3′ (red) probe binding site, aberrant probe hybridization, or observer error. Observer error may be particularly problematic in lung cancer, where the predominant 5′ fusion partner is EML4 that resides only 12 megabases centromeric from ALK on chromosome 2p. The paracentric inversion that generates the EML4-ALK fusion causes only minimal separation of the 3′ and 5′ ALK probes and therefore may be harder to spot than rearrangements involving genes mapped on different chromosomes seen in other ALK+ cancers.9, 12 We have similarly shown that there is also the potential for a cell-by-cell false positive rate with up to 11% of nontumor cells or tumor cells within protocol-defined ALK-negative tumors appearing to demonstrate rearrangement patterns.9

Just as with false negatives, there are also numerous technical factors that could lead to false positive patterns. Among them are the stretching of the DNA leading to artificial separation of the red and green probe binding sites from each other, nuclear sectioning causing loss of the 5′ (green) probe binding site, aberrant probe hybridization, or observer error. Whether tumors with any percentage of positive cells will respond to crizotinib or whether these are indeed false positive cells (as the detection of positive patterns in nontumor tissue suggests can sometimes be the case) is unknown. Further exploration of the cell-by-cell false positivity rate would require second screening techniques such as immunohistochemistry and potentially wider access to specific inhibitors in the clinic to see if borderline tumors respond. Exploration of the potential for false cellular negatives in ALK testing by FISH is, however, already possible via retrospective analyses set up to address a series of specific hypotheses (Table 3). We have previously shown that the heterogeneity in cellular positivity appears to be a diffuse and not a focal phenomenon within positive tumors, suggesting that clonal gain or loss of rearrangements is not the dominant basis of the heterogeneity.9 To further explore whether technique rather than biology underlies this heterogeneity, we analyzed in detail the association between patterns of positivity (split vs single red patterns) and percentage of positive cells and association between signal copy number (of fused, isolated red and isolated green signals) and percentage of positive cells (Fig. 1). The higher percentage of positive cells seen in tumors with a single red pattern of positivity (74%), compared to the percentage seen in tumors manifesting a split pattern (48%) (P < .0001), suggest that difficulties in detecting the narrow probe separation in the dominant EML4-ALK paracentric inversion may indeed contribute toward “missed” true positive cells (Tables 1 and 3).

Table 3. Hypotheses Relating to ALK FISH Heterogeneity
 Technique-RelatedBiology-RelatedResolution
Distribution of positive cells within a positive tumorExpected to be diffuseExpected to be focalSliding windows analysis looking for areas where only 1 of 4 high-power field “windows” met positivity criteria failed to demonstrate a single focal event out of 17 analyzed tumors.9
Effect of pattern of positivity (single red vs split) on percentage cells positiveSingle red pattern should be easier to spot and be associated with a higher percentage of cells being positiveNo differences expectedTumors with a single red pattern of positivity had a higher mean percentage of positive cells compared to those with a split pattern of positivity (74% vs 48%, respectively; P < .0001) (Table 1).
Effect of signal copy number (fused, isolated red and isolated green signals) on percentage cells positiveIncreased isolated red signal copy number (contributing to both split and single red patterns of positivity) and increased isolated green signal copy number (in split pattern positive tumors) should increase the chances of a true positive pattern being seen in any given cell and be associated with a higher percentage of cells being called positiveNo correlation expectedIsolated red signal copy number strongly correlated with a higher percentage of cells positive (r = 0.743, P < .0001) (Table 1, Figure 2). Isolated green signal copy number (split pattern of positivity) was strongly correlated with a higher percentage of cells positive (r = 0.909, P < .0001) (Table 1). Fused signal copy number was negatively associated with the percentage of cells positive if the tumor had a split pattern of positivity (r = −0.44, P = .0016 vs r = −0.042, P = .016), consistent with fused and split patterns being the hardest to distinguish.
Correlation between percentage cells positive and maximal percentage tumor shrinkage with crizotinibNo correlation expectedCorrelation expectedPercentage cells positive did not correlate with the maximal percentage tumor shrinkage per RECIST with crizotinib (r = 0.192, P = .3) (Fig. 3).

Given that only one split or single red pattern per cell is required to call a cell positive, if cells are falsely called negative because a positive pattern is missed in individual cells, we hypothesized that increases in the signal copy number associated with positive patterns should be correlated with a higher percentage of cells being called positive within any tumor. The strong positive association between increases in isolated red signal copy number and percentage of ALK FISH+ cells (r = 0.743, P < .0001) and between isolated green signal copy number (when part of a split pattern of positivity) and percentage of ALK+ cells (r = 0.91, P < .0001) appears to support this hypothesis (Tables 1 and 3; Fig. 2). The fact that the correlation is even stronger for the isolated red signals when only tumors with a single red pattern of positivity were considered, compared with those having a split pattern, also supports the idea that single red patterns may be easier to spot than split patterns (r = 0.884, P < .0001 vs r = 0.558, P < .0001, respectively). As a corollary, the hypothesis is also supported by the observation that overall isolated green signal copy number (contributing to both the split pattern of positivity and the single green negative pattern) did not show a strong association with the percentage of cells positive for a rearrangement (r = −0.264, P = .012). Similarly, fused signal copy number was negatively associated with the percentage of cells positive for a rearrangement (r = −0.466, P < .0001), particularly if the tumor was positive with a split as opposed to a single red pattern of positivity (r = −0.44, P = .002 vs r = −0.042, P = .016), consistent with fused and split patterns being the hardest to distinguish.

If all the variation in percentage cells positive within a positive tumor only reflects technical aspects of ALK FISH testing rather than true biological heterogeneity, the proportion of positive cells should not influence therapeutic outcomes. Specifically, we hypothesized that there should be no correlation between the percentage of positive cells present in any positive tumor and the maximal percentage tumor shrinkage per RECIST seen in that tumor after crizotinib treatment; this was indeed the case (r = 0.192, P = .3) (Table 3; Fig. 3).

Copy number alterations in ALK-related signals have recently been described.9, 10 Although we have exploited these alterations to address questions about false cellular negatives in ALK+ tumors, whether copy number alteration reflected biological variation that was directly clinically relevant was unknown. Here, we have shown that neither isolated red (r = 0.274, P = .195) (Fig. 4) nor isolated green (split pattern) (r = 0.438, P = .117) nor fused signal (r = −0.247, P = .187) copy number correlated with the maximal percentage tumor shrinkage per RECIST after crizotinib treatment.

Potential flaws in the logic underlying these response-related data include the fact that the percentage of positive cells and the signal copy number mostly have been assessed on archival material, whereas the tumor itself may have changed biologically prior to commencing crizotinib. In addition, the estimate of tumor shrinkage per RECIST, although expressed as a percentage, is based on proportional change in a very specific measurement calculation (the sum of the longest diameter of a number of prespecified target lesions) in RECIST, version 1.0. Therefore, the proportional shrinkage may not reflect the exact magnitude of the biological response. For example, it may not be as direct a measure as assessing changes in tumor volume, and may underestimate the clinical activity of crizotinib in the correlations. However, based on our data, the lack of correlation between either percentage of positive cells or fused/rearranged signal copy number, and crizotinib response does suggest that these are not relevant biological variables to follow with respect to benefit from ALK inhibition.

Finally, although there was no correlation between mean copy number of fused and isolated green signal copy number, mean fused copy number was negatively associated with isolated red signal copy number (r = −0.409, P < .0001) (Table 2). The inverse relationship between fused and isolated red signal copy number suggests that ALK gene rearrangements may not be a manifestation of simple chromosomal instability in these cancer cells. Specifically, this raises the possibility that although there may be some selection pressure for increases in ALK signal copy number, or that the region around the ALK gene may be particularly fragile, overall ALK gene rearrangements tend to originate in a distinct “near-diploid” state. Similar near-diploid states have been described, notably in the microsatellite instability subtype of colorectal cancer.13 This suggests that ALK rearrangements may occur early in the cancer development process, before instability changes leading to significant chromosomal aneusomy occur.

In conclusion, our detailed analyses based on correlations between the percentage of ALK+ cells, patterns of ALK positivity, and mean signal copy number per tumor cell and association of these variables with response to ALK inhibition in ALK+ NSCLC, support the hypothesis that heterogeneity in the percentage of cells positive within ALK+ tumors reflects technical rather than biological variation (Table 3). The lack of correlation between either percentage of cells positive or signal copy number and response to crizotinib suggests that these are not relevant biological variables to capture with respect to predicting benefit from ALK inhibition in ALK+ tumors. In addition, the negative association revealed between copy numbers of fused signals and isolated red signals suggests the novel hypothesis that ALK gene rearrangements occur early in cancer development, preceding significant chromosomal aneusomy. Additional studies are warranted to investigate further whether ALK+ NSCLC is indeed a “near-diploid” subtype of NSCLC.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. FUNDING SOURCES
  6. REFERENCES

This study was funded by the Colorado Lung Specialized Program of Research Excellence (P50CA58187).

CONFLICT OF INTEREST DISCLOSURE

The authors made no disclosure.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. FUNDING SOURCES
  6. REFERENCES