RAI1 Alternate Probe Identifies Additional Breast Cancer Cases as Amplified Following Equivocal HER2 Fluorescence In Situ Hybridization Testing: Experience From a National Reference Laboratory Open Access

Cases of HER2 FISH testing, performed between April 2015, when the alternate RAI1 probe was available in our laboratory, and October 2015, were identified. Only primary or metastatic breast cancer cases were included, and hematoxylin-eosin–stained slides were reviewed to confirm histologic type, morphologic features, and grade of tumors. Our laboratory uses the US Food and Drug Administration (FDA)–approved, dual-probe HER2 IQFISH (Dako, Carpinteria, California) for initial testing, with equivocal FISH cases retested with HER2 and RAI1 probes (both from Agilent Technologies, Santa Clara, California). At least 40 cells were counted manually in equivocal FISH cases and were reviewed by both a technician and a board-certified pathologist trained in anatomic pathology. Information on prior HER2 immunohistochemical testing (if performed) and FISH results with the FDA-approved probe set and the alternate probe set were retrieved from electronic files. The average copy number per cell of HER2, CEP17, HER2, and RAI1 from each probe set and their corresponding ratios were recorded. The FDA-approved probe set was scored according to the ASCO/CAP 2013 guidelines for interpreting HER2 dual-probe FISH assays (average HER2 copy number of 4.0 to 5.9 per cell with a HER2:CEP17 ratio of <2.0). The HER2 FISH assay with the RAI1 control probe was interpreted in the following manner: reported as nonamplified if the HER2:RAI1 ratio was less than 2.0, the average number of RAI1 per cell was 1.5 or greater, and there were 16 or fewer cells out of 40 with only one RAI1 signal; and as amplified if HER2:RAI1 ratio was 2.0 or more. If RAI1 was deleted (defined as a RAI1 copy number per cell of <1.5 and, of 40 cells, there were 17 or more with only one RAI1 signal per cell), HER2 status was considered unresolved/indeterminate, and additional testing was recommended. A cutoff for RAI1 gains was not used because of the assumption that cases with gains in HER2, the centromere, and RAI1 were unlikely to represent true HER2 amplification.

During the study period, 97 of 876 cases (11.1%) tested with CEP17 as the reference probe had equivocal FISH results. Thirty-three of 97 cases (34%) had prior equivocal (2+) immunohistochemical staining, with 10% to 30% having weak or moderate, circumferential membrane staining. In the remaining cases, FISH was the primary testing method in our laboratory. Tumors that had equivocal HER2 FISH requiring alternate RAI1 testing had the following morphologic features: 87 ductal not otherwise specified, 6 ductal with lobular features, 2 lobular, 1 mucinous, and 1 ductal with micropapillary features. Additional testing with the RAI1 probe identified 39.2% (38 of 97) of the cases classified as amplified with a HER2:RAI1 ratio ranging from 2.0 to 3.2 (average ratio, 2.37), 57.7% (56 of 97) were nonamplified by a HER2:RAI1 ratio (<2.0) (Figure 1, A through H), and 3.1% (3 of 97) were still unclassifiable because of a deletion of RAI1 (Table 1). Genetic heterogeneity was not detected in any of the cases tested with CEP17 and RAI1 probes. Seventy-six of 97 cases (78.4%) had a RAI1 copy number of 2 or more per cell, of which, 54 (71.1%) had a nonamplified HER2:RAI1 ratio (<2.0). Of the 97 cases, 43 (44.3%) had a RAI1 copy number of 2.6 or more per cell; of which, 41 (95.3%) had a nonamplified HER2:RAI1 ratio (<2.0). Although the indication for additional testing with the RAI1 probe was an average HER2 copy number of 4 to 5.9 (classified as an equivocal FISH result per the ASCO/CAP 2013 guidelines) and not the CEP17 number, all 97 cases had average CEP17 copy number of 2 or more and 84 (86.6%) had copy number 2.6 or more (range, 2.2–5.4; average, 3.2). Seventy-six (78.4%) of all cases had a RAI1 copy number of 2 or more, and 43 (44.3%) had copy number of 2.6 or more (range, 1.2–20.0; average, 2.9). The correlation between average CEP17 and RAI1 copy numbers for all cases is shown in Figure 2. In 3 cases, the original equivocal FISH result with FDA-approved probe set could not be resolved after testing with the alternate probe because of RAI1 deletion (Table 2).

In this study, the RAI1 alternate probe identified an additional subset of cases as amplified (39.2%) based on a HER2:RAI1 ratio 2.0 or more after an equivocal FDA-approved HER2 FISH test result.

HER2-amplified cancers show complex genetic abnormalities of chromosome 17, especially involving the long arm.¹²  The FISH probes, such as RAI1, located on band 17p11.2, and RARA, located on band 17q21.2, in addition to microarray-based comparative genomic hybridization and multiplex ligation-dependent probe amplification, have shown true polysomy of chromosome 17 to be rare.^12–14  However, copy number gains in the centromeric or pericentromeric regions of chromosome 17 are well documented in breast cancers and may lead to false-negative results based on the HER2:CEP17 ratio in assays using a centromeric enumeration probe.¹⁵  Therefore, the 2013 ASCO/CAP guidelines indicate that alternative testing may be useful for excluding coamplification of the centromere for cases with a HER2 copy number ranging between 4 and 5.9 when the HER2:CEP ratio is less than 2 (because of a CEP17 copy number >2). The RAI1 gene is located on band 17p11.2 and appears to be a good candidate as a control probe with less than 1% high-level amplification, about 7% low-level gains, and less than 0.1% coamplification with HER2 based on TCGA analysis of breast cancers.¹⁶  In this study, gains of RAI1, either defined as an average RAI1 copy number greater than 2, as in our validated test, or as more than 2.6 per cell, as used by some others, was higher than found in previous studies.¹⁷  However, this study includes highly selected HER2 FISH equivocal cases, rather than a mix of breast cancers that form the basis of TCGA data. Two of the cases showed apparent amplification of RAI1 (>6 copies per cell), whereas 44.3% of all cases showed increased RAI1 (>2.6) per cell, similar to their corresponding CEP17 copy number. In these scenarios, if the HER2:RAI1 ratio were less than 2, the cases were reported as nonamplified, although the numbers were similar to those obtained with the FDA-approved probe set. The 2013 guidelines do not comment on the interpretation of the reporting of alternate probes. It is not known, in a clinical setting, whether retesting with yet another probe on chromosome 17 or elsewhere in the genome would identify additional cases that might benefit from targeted therapies.

Deletion of a control probe would result in ratios in the amplified range, even in cases with low (<4) HER2 copy numbers. RAI1 is reported to be deleted in one-half of the breast cancer cases in TCGA and in 37% of the subjects in a recent study.¹¹  The cases with deleted RAI1 are reported as unresolved. Deletion of RAI1 was much less than reported in the TCGA, with only 3.1% of cases not resolved because of deletion of RAI1. Deletion rates were also less than expected in previous studies in which additional alternate probes were tested based on increased CEP17 numbers. In those studies, they did not combine HER2 and reflex probe in the same assay but instead used the FDA HER2 value to compare to an alternate probe, including SMS/RAI1, increasing the variability.^11,12,17  The retesting decision in this study was based on the average HER2 copy number of 4 to 5.9 with a HER2:CEP17 ratio less than 2.0 and was not based solely on the average CEP17 copy number, possibly indicating a different patient population.

There are rare cases in which the HER2:CEP17 ratio is 2 or more with an average HER2 copy number less than 4.0 per cell (ie, CEP17 number <2 per cell, suggesting chromosome 17 monosomy). The 2013 guidelines recommend that these cases be classified as amplified because of Herceptin-adjuvant trials,¹⁸  which included such patients (albeit a small subset) who did not show any reduced benefit from targeted therapy. At the same time, the 2013 ASCO/CAP supplemental data indicate that several members of the committee, when faced with this scenario, “expressed concern about describing an invasive breast cancer as HER2 positive” ^6(p9) and recommended further testing. This highlights nonconsensus regarding the appropriate method for handling cases in this rare subgroup with low CEP17 copy numbers and HER2:CEP17 ratios in the amplified range. It is unknown whether recommendations for interpreting the FDA-approved control probe in these apparent monosomic cases can be extrapolated to alternate probes. We chose to report these cases as unresolved and recommended retesting, preferably on an alternate sample.

In this study, close to 40% of the equivocal cases using the FDA probe set HER2 FISH were reclassified as amplified using the HER2:RAI1 probe set. Jang et al¹¹  reclassified 35 of 46 cases (76.1%), a higher percentage of their HER2 equivocal cases as amplified by using 3 alternate probes (RARA, TP53, SMS/RAI1).¹¹  However, they selected 1 of the 3 ratios with the highest HER2 signal count and less than 2.6 signals for the control probe, which varied from case to case; that is, no single probe was found to consistently identify all amplified cases. Using alternate probes in an attempt to find an amplified ratio is not supported in the absence of clinical data showing benefit from targeted therapies.

Our laboratory-developed reflex assay contains differentially labeled probes for HER2 and RAI1, allowing for evaluation of HER2 and RAI1 on the same set of cells and likely reducing the variability caused by combining data from 2 separate FISH assays, as was done by Jang et al.¹¹  The HER2 probe used in the reflex test (Agilent Technologies) was dimmer and smaller (191 kb versus 218 kb) compared with the FDA-approved test (Dako IQFISH). This might have contributed to the slightly lower (4.32 versus 4.59), but not significantly different, average HER2 signals per cell in the different tests. In 2 cases in our study, the HER2:RAI1 ratio identified an average HER2 copy number per cell of 6.1 and 6.3 (amplified range), whereas the FDA-approved probe set found an average HER2 copy number per cell of 5.5 and 5.9, respectively. Reasons for this difference could include interobserver variation and tissue heterogeneity.

In this study, we reviewed our experience from a reference laboratory with the RAI1 alternate probe after equivocal HER2 FISH cases identified by an FDA-approved probe set. The 2013 guidelines suggested the use of a reflex test following equivocal results that includes an alternate probe to identify patients who might be eligible for targeted therapy. Although this alternate probe identified close to 40% of the original HER2 FISH equivocal cases as amplified, making them eligible for targeted therapies, it is largely unknown whether this subgroup of patients will benefit from targeted therapies. Thus far, there have been no clinical trials evaluating response in patients identified as amplified via an alternate probe. The lack of guidelines for the reporting of results with alternate probes is not only problematic for pathologists but also for clinicians, especially when no immediate alternate tissue is available for retesting. In addition, the use of alternate probes for resolving HER2 status can place additional burdens on laboratories performing the testing.