Introduction

Uterine serous carcinoma (USC) is an aggressive subtype of endometrial cancer which predominantly affects postmenopausal patients and arises in the setting of atrophy. Although it accounts for fewer than 10% of endometrial carcinomas, it is responsible for 40–80% of deaths [1,2,3]. Clinically, serous carcinoma is characterized by an aggressive disease course, often with extra-uterine disease even in the absence of myometrial invasion [4, 5]. Genetically, the hallmark of serous carcinoma is loss-of-function mutations in TP53, which are present in over 90% of cases [6, 7]. Chemotherapy is the mainstay of treatment for even early stage disease, although relapses and metastases are still common [3].

HER2 is a cell surface receptor in the epidermal growth factor receptor family and is encoded by ERBB2, which maps to 17q12. The protein is composed of an intracellular tyrosine kinase domain, a transmembrane domain, and an extracellular ligand binding domain [8]. Receptor activation triggers a number of pathways implicated in cell growth, apoptosis, and differentiation [9]. HER2 is overexpressed in several cancer types [8], and ERBB2 amplification correlates with a poorer prognosis in breast [10, 11], gastrointestinal [12,13,14], and endometrial carcinomas [15,16,17,18,19,20]. In USC, overexpression of HER2 by immunohistochemistry (IHC) ranges from 14 to 80% [15,16,17,18, 20,21,22,23,24] and gene amplification in 3–42% [6, 16,17,18, 20,21,22,23, 25, 26].

There are two categories of targeted therapies for tumors with HER2 overexpression: monoclonal antibodies against the extracellular domain of HER2, such as trastuzumab, and tyrosine kinase inhibitors of the intracellular domain, such as neratinib. Both approaches are used routinely in breast [27,28,29,30] and gastroesophageal tumors [31, 32]. Early reports described cases of USC with HER2 overexpression responding to trastuzumab therapy [24, 33, 34]. However, a subsequent phase II trial (GOG181B) failed to demonstrate any benefit [35]. This trial was subsequently criticized for a number of perceived limitations: it included a high number of non-serous subtypes, only 45% of cases had ERBB2 amplification, and it was statistically underpowered to detect clinically meaningful response rates [36]. A second phase II trial, however, addressed these shortcomings and demonstrated a 4-month increase in progression-free survival when patients with advanced stage or recurrent USC with HER2 overexpression were treated with trastuzumab [37]. Since then, anti-HER2 therapies have become standard of care for patients with advanced stage, recurrent, or metastatic USC [38].

Currently, HER2 overexpression and amplification in USC is assessed by IHC and in situ hybridization (ISH), respectively [39]. Next-generation sequencing (NGS) is frequently used to assess molecular alterations in endometrial carcinomas, including targetable point mutations, POLE mutation status [40], and microsatellite instability [41]. However, its ability to reliably detect ERBB2 amplification in endometrial carcinoma has not been previously studied. To address this question, we examined a cohort of USC and compared ERBB2 amplification as measured by NGS, IHC, and ISH.

Materials and methods

Case selection

Cases of USC, which had been previously tested by a targeted hybrid-capture NGS assay between 2014 and 2019 at Brigham and Women’s Hospital (Boston, MA), were identified by a retrospective search. Cases were included for study only if additional material was available from the same anatomic site for IHC, and if applicable, ISH. When possible, the same paraffin block was used for IHC, ISH, and NGS. However, for some cases which were seen in consultation, the only material available from the hysterectomy for subsequent IHC/FISH studies was from a hysterectomy block different than the one tested by NGS. This study was approved by the Brigham and Women’s Hospital Institutional Research Ethics Board.

Immunohistochemistry (IHC)

IHC for HER2 (SP3 clone; 1:75 dilution; Cell Marque, Rocklin, CA) was performed on 5-µm thick, full-slide sections of formalin-fixed paraffin-embedded tissue. IHC was scored independently by at least two pathologists using the 2018 ASCO/CAP clinical practice guideline for HER2 in breast cancer [42]. As per the guidelines, a positive (3+) result required intense, completely circumferential membranous staining in a contiguous focus representing at least 10% of tumor cells. In addition, intense basolateral staining was considered positive, as this pattern has been shown to correlate with amplification in USC [21]. Heterogeneity was assessed as a spatially discrete tumor population with either 2+ staining with amplification confirmed by FISH, or 3+ staining, on a background of negative tumor cells (0 or 1+). Discrepancies in interobserver scoring were resolved via consensus. Appropriate immunohistochemical controls were examined.

Fluorescence in situ hybridization (FISH)

All cases that scored 2+ (equivocal) by IHC were assessed by FISH for ERBB2 amplification. A subset of cases that scored 3+ by IHC was confirmed by FISH if material was available. Two 3–5-mm regions of tumor with the highest expression by IHC were marked for FISH evaluation. In cases with heterogeneous expression, the target probe was applied and assessed in both IHC-positive and - negative areas for confirmation of amplified and non-amplified populations. Five-micrometer sections of formalin-fixed paraffin-embedded tumor were tested with the Vysis PathVysion HER2 DNA Probe kit (Abbott, Abbott Park, IL), which includes the locus-specific ERBB2 probe (17q12) and the CEP17 (D17Z1) centromeric probe (17p11.1-q11.1). The previously marked areas of maximum HER2 IHC expression were evaluated by two observers examining at least 30 nuclei in a contiguous portion of the tumor, and the entire region marked by a surgical pathologist was scanned for heterogeneity in HER2 copy number (CN) status. Only cases with amplification in 10% or more of the tumor cells were considered positive [42]. Evaluation of each specimen followed the algorithm described in the 2018 ASCO/CAP clinical practice guideline for breast cancer [42]. Those guidelines define the dual-probe FISH groups as follows: group 1 is ERBB2/CEP17 ratio ≥2.0 and ≥4.0 ERBB2 signals/cell; group 2 is ERBB2/CEP17 ratio ≥2.0 and <4.0 ERBB2 signals/cell; group 3 is ERBB2/CEP17 ratio <2.0 and ≥6.0 ERBB2 signals/cell; group 4 is ERBB2/CEP17 ratio <2.0, and ≥4.0 and <6.0 ERBB2 signals/cell; and, group 5 is ERBB2/CEP17 ratio <2.0 and <4.0 ERBB2 signals/cell. Group 1 is positive, group 5 is negative, and groups 2–4 require correlation with IHC to determine the overall HER2 status.

Targeted NGS

Targeted NGS was performed as previously described [43, 44]. Briefly, areas of tumor were macrodissected from unstained slides and DNA was isolated from the tissue. The fraction of tumor nuclei in the area selected for study was estimated from hematoxylin and eosin stained slides, and only samples with a tumor percentage >20% were tested. Sequencing libraries were generated for the coding regions of at least 275 genes, including ERBB2, using solution-based hybrid capture (Agilent SureSelect; Agilent Technologies, Santa Clara, CA). Sequencing was performed using an Illumina HiSeq 2500 (Illumina, Inc, San Diego, CA). Cases included for study passed quality control metrics. The median average target coverage was 286.9 (range 68.3–635.1), and the median percentage of bases with >30× coverage was 98.4% (range 85.6–99.4%). ERBB2 CN by NGS was determined by the relative number of reads mapping to ERBB2 in tumor DNA compared to pooled normal (unmatched) control nonneoplastic DNA. ERBB2 CN was calculated using the estimated tumor percentage

$${\mathrm{CN}} = \left( {2\,{\times}\, \frac{{{\mathrm{Median}}\;{\it{ERBB2}}\;{\mathrm{copy}}\;{\mathrm{ratio}} - 1}}{{{\mathrm{Tumor}}\;{\mathrm{cell}}\;{\mathrm{fraction}}}}} \right) + 2.$$

A receiver operating characteristic curve analysis, using FISH as the gold standard, demonstrated an AUC of 1.0 between a NGS CN cutoff of 5.4 and 7. For simplicity, we chose to use the same threshold of 6.0 as the 2018 ASCO/CAP clinical practice guideline for reporting of breast cancer single probe HER2 ISH [42]; an ERBB2 CN ≥ 6 was considered amplified, and ERBB2 CN < 6 was non-amplified.

Statistical analysis

A post hoc power analysis was completed using the methods described by Liu et al. [45] and Tang [46] for a matched pair non-inferiority trial with binary outcomes. A sample size of n = 93 had 62% power at α = 0.05 to detect non-inferiority if the minimum ratio of amplification detection sensitivities between the two techniques resulting in non-inferiority was 0.8. With a minimum non-inferiority ratio of 0.9, the power to detect non-inferiority was 44%.

Results

In total, specimens from 93 patients were included for the study, including 71 hysterectomies, 8 endometrial biopsies, 1 curetting, and 13 metastases.

By IHC, 70 of 93 cases (75%) were negative (0 or +1), 19 (20%) were equivocal (2+), and 4 (4%) were positive (3+, Fig. 1 and Table 1). Of 9 cases (10%) with heterogeneous HER2 expression, two tumors were 3+, with intense staining in 20 and 70% of the tumor cells. In six tumors, the heterogeneity was of limited extent, with intense staining limited to an area representing <10% of the tumor, and were classified as equivocal (2+). In one additional tumor classified as IHC equivocal (2+), there was a discrete area of weak-to-moderate membranous staining (80% of tumor), which was found to be amplified by FISH.

Fig. 1: HER2 immunohistochemical staining in uterine serous carcinoma.
figure 1

A Score of 0 (negative), with faint staining in ≤10% of tumor cells. B Score of 1+ (negative), in which there is faint, incomplete membranous staining in >10% of tumor cells. C Score of 2+ (equivocal), with strong, circumferential staining in ≤10% of tumor cells. D Score of 3+ (positive), with circumferential, complete membrane staining that is intense and >10% of tumor cells.

Full size image
Table 1 Immunohistochemical (IHC) expression of HER2, ERBB2 amplification status determined by FISH and combined IHC/FISH assays in cases of uterine serous carcinoma compared to ERBB2 copy number as determined by next-generation sequencing (NGS).
Full size table

FISH was performed on a subset of cases (n = 24), which had IHC scores of 1+ (n = 3), 2+ (n = 19), and 3+ (n = 2) (Tables 2 and 3). The median ERBB2 copies per cell determined by FISH were 3.4 (mean 7.0, range 1.8–53.8). The median ERBB2/CEP17 ratio was 1.4 (mean 2.8, range 0.6–18.1). When classified according the 2018 ASCO/CAP ISH groupings, 6 (25%) were group 1, 1 (4%) group 2, 1 (4%) group 3, 2 (8%) group 4, and 14 (58%) group 5. FISH performed on tumors with heterogeneous IHC expression confirmed amplification in areas with 3+ pattern of staining, and absence of amplification in areas of tumor with 0 or 1+ pattern staining. A single case with 1+ expression was amplified by FISH, with ERBB2 CN 13.9 and ratio 3.9.

Table 2 Immunohistochemical (IHC) expression of HER2 compared to ERBB2 amplification status determined by FISH in 24 cases of uterine serous carcinoma.
Full size table
Table 3 Cases (n = 24) with ERBB2 amplification determined by both fluorescence in situ hybridization (FISH) and next-generation sequencing (NGS).
Full size table

Using the combined interpretation of the IHC and FISH assays, 8 (9%) cases showed ERBB2 amplification and 85 (91%) were non-amplified.

By NGS, the median ERBB2 log2 ratio for all cases was 0.035 (Fig. 2 and Supplementary Fig. 1). Using the estimated tumor fraction, ERBB2 CN was calculated, giving a median CN of 2.1 for all cases (mean 3.7, range 0.2–62.8). Eight (9%) cases were amplified (CN ≥ 6, mean 20.0, range 7.0–62.8), and eighty-five (91%) cases were non-amplified (CN < 6, mean 2.1, range 0.2–5.4). The concordance between NGS and the combined IHC and FISH interpretation was 100% (Table 1 and Fig. 3). A least squares linear regression, including all cases with FISH, of NGS CN as a function of FISH CN yielded a slope of 0.25, with an R2 of 0.509 (Fig. 4, gray line). When cases with heterogeneous amplification were excluded, the slope was 0.64 with an R2 of 0.758 (Fig. 4, black line), indicating that the CN in cases with heterogenous amplification was systematically underestimated by NGS compared to FISH.

Fig. 2: ERBB2 amplification detected by NGS.
figure 2

A Copy number (CN) plot of all chromosomes for a case of serous carcinoma with several copy number gains and losses, including ERBB2 amplification (red circle). The y-axis is the log2 ratio of the number of reads in the tumor sample compared to a set of normals. B ERBB2 amplification (CN = 13).

Full size image
Fig. 3: ERBB2 copy number (CN) in uterine serous carcinoma determined by next-generation sequencing (NGS) compared to traditional assays.
figure 3

A NGS compared to  immunohistochemistry (IHC), and B NGS compared to  combined IHC and fluorescence in situ hybridization (FISH). Dashed line is CN = 6.

Full size image
Fig. 4: Linear regression of ERBB2 copy number measured by NGS and FISH in uterine serous carcinoma (n = 24).
figure 4

Cases with heterogeneity in HER2 expression (n = 9) are indicated by ×, those without heterogeneity (n = 15) are marked by •. Linear regression including all cases is a gray line, while the regression excluding cases with heterogeneity is a black line. Dashed line is CN = 6.

Full size image

Discussion

Since trastuzumab was recently shown to have a progression-free survival benefit in ERBB2-amplified USC, it has become critical to correctly identify patients for targeted therapy. HER2 overexpression and amplification in USC has been studied for over 20 years using IHC, fluorescence [16, 18, 21,22,23, 33, 47, 48] and chromogenic [49, 50] ISH, and polymerase chain reaction [17, 18]. Endometrial carcinomas are routinely interrogated by NGS for targetable treatments, including point mutations and microsatellite instability. This is the first report which demonstrates that ERBB2 amplification in USC can be accurately detected using NGS. It follows previous work that has shown excellent concordance between ERBB2 amplification detection by NGS and IHC overexpression in breast [51], gastroesophageal [51], and colorectal carcinomas [52].

The frequency of ERBB2 amplification found here (9%) is in the lower range of that previously reported (3–42% [6, 16,17,18, 20,21,22,23, 25, 26]), and this discrepancy is likely partially due to a selection bias. The cases selected for inclusion in this study were previously tested using our in-house targeted NGS assay, which is performed by clinician request. Often, this testing is performed to identify a targeted treatment when a patient has recurrent or metastatic disease. If a case of USC was found to have overexpression of HER2 during routine clinical testing, the clinician may have opted to forgo NGS testing as a targeted therapy was already available to the patient. We are aware of several such cases that showed HER2 overexpression and were not referred for genomic profiling, thereby enriching the sequenced cohort for ERBB2 amplification negative cases. Other possible explanations for a lower proportion of positive cases include differences in the underlying patient population, the HER2 antibody clone used, or the scoring system selected. In any case, both the small fraction of HER2 positive cases and total sample size of this cohort are significant limitations, as this study was not sufficiently powered to conclusively demonstrate non-inferiority of NGS compared to IHC/FISH. Additional studies are necessary to confirm these findings.

In this study, HER2 overexpression and amplification in USC was assessed using the 2018 ASCO/CAP clinical practice guidelines for HER2 expression in breast cancer [42], an approach recommended by the College of American Pathologists [39]. This is in contrast to previous studies, including the recent phase 2 trial [37], which have used other criteria for evaluating HER2 expression in USC, including the modified 2007 ASCO/CAP guidelines, or the original US FDA criteria [53]. This decision will not affect the interpretation of most cases with straightforward positive and negative results. However, one advantage of the updated guidelines is that non-classical FISH results in groups 2–4 (i.e., those that are not clearly positive [ERBB2:CEP17 ratio ≥ 2 and ERBB2 CN ≥ 4] or negative [ERBB2:CEP17 ratio < 2 and ERBB2 CN < 4]) are adjudicated by the IHC interpretation, with the final HER2 status typically correlating better with the level of HER2 expression and ERBB2 CN [54]. Furthermore, the 2018 guidelines eliminate the equivocal ISH categories and reduce the false positive rate compared to the 2013 guidelines [55, 56]. It appears that our choice of guidelines may have helped to facilitate the excellent correlation observed between IHC/FISH and gene amplification determined by NGS. For example, of the four cases in groups 2–4 in our study, only the group 3 case (ERBB2:CEP17 ratio < 2 and ERBB2 CN ≥ 6) with an IHC score of 2+ was considered FISH-positive and amplified by NGS. It should be noted that according to the modified 2007 criteria used in the recent phase 2 clinical trial, cases with an ERBB2:CEP17 ratio ≥2 were considered positive (irrespective of CN) [21, 37]. If these group 2–4 cases had been assessed using this modified 2007 criteria, the group 2 case (ratio ≥ 2 and ERBB2 CN < 4) would have been considered positive by FISH but negative by NGS, while the group 3 case (ratio < 2.0 and ERBB2 CN ≥ 6.0) would be negative by FISH but positive by NGS, leading to discordances with our NGS results. Although these group 2–4 cases are relatively rare in breast cancer (~5%), they represent 17% of the cases tested by FISH in our small study of USC, suggesting that an appropriate strategy for how to approach these cases may be needed. Another difference between the 2018 and 2007 guidelines is the IHC criteria for an interpretation of positive (3+), which reduced the required proportion of tumor cells with intense staining from 30 to 10%. From our limited experience, it does not appear that the lower threshold for IHC-positive results had an impact on the sensitivity of NGS, as all four IHC-positive (3+) cases were NGS-amplified, including two IHC-positive heterogeneous cases with a 3+ pattern of staining in 20–70% of tumor cells. Although we recommend using the 2018 ASCO/CAP guidelines for breast cancer for HER2 testing in USC, we understand that some groups may be hesitant to do so given that the modified 2007 ASCO/CAP guidelines were used in the clinical trial, and that is the only scoring system which has demonstrated clinical response in a randomized trial. As such, select cases may be best managed by multidisciplinary discussion. Additional studies are needed to determine the optimal expression and amplification levels which correlate to response to anti-HER2 therapies.

One case demonstrated 1+ (negative) expression by IHC but was amplified by both FISH and NGS. This IHC result may represent a technical failure due to loss of antigenicity, as the only material available for testing was an old, archived unstained slide. This specimen may also have had inadequate fixation, as pathologists are generally not as mindful of preanalytical factors such as fixation time and cold ischemic time in hysterectomies, as compared to breast specimens. Another possible explanation is that this case had greater expression of a truncated variant of HER2 (p95HER2) that our extracellular antibody (SP3) does not detect. This shedding of the extracellular domain has been reported to be a more significant phenomenon in endometrial cancer than in breast cancer [57].

Detection of gene amplification by NGS has several advantages. We demonstrated a perfect concordance between ERBB2 amplification by NGS and the current gold standard, FISH. FISH is usually only performed if HER2 IHC is equivocal (2+), while a targeted NGS panel is increasingly routinely performed for advanced stage USC. This raises the possibility that additional actionable information can be extracted from an NGS assay that was already performed. Currently, patients can only receive trastuzumab therapy if their tumors show overexpression by IHC or amplification by ISH; amplification by NGS can act as a trigger for a second confirmatory study. As our collective experience with NGS-based tests increase, it may eventually be accepted in place of IHC or FISH assays. Although the optimal IHC scoring system for USC has yet to be established [39], the CN measured by NGS shows excellent concordance with FISH.

Assessment of gene amplification by NGS has several limitations. The calculated CN is directly proportional to the estimated tumor percentage, and therefore an accurate assessment of the proportion of tumor nuclei in the sample is critical to accurately determine CN. Also, both IHC and (F)ISH allow for assessment of single cells and therefore identification of subclones, whereas NGS provides an average measurement over a larger tumor area. Consequently, an ERBB2-amplified subclone, which comprises a small proportion of the tumor, may not be detected by NGS. In this series, all amplified cases (n = 8) were successfully identified by NGS, including an IHC-positive case with a 3+ pattern in 20% of tumor cells. Of note, there were six cases in this series with heterogeneity of limited extent (3+ patten in <10% of cells), which are best considered as negative according to the 2018 ASCO/CAP guidelines, and which were correctly classified as negative by NGS (CN < 6). Given these results, although the limit of detection was not rigorously evaluated in this study, it appears that NGS may have the capability to discriminate between clinically relevant and insignificant amplified populations.

Tumor sequencing using clinical NGS panels may be performed either with or without analysis of paired normal DNA from the same patient. Paired sequencing allows for more accurate filtering of germline variants and may improve detection of CN variants in some situations. Despite its advantages, paired tumor-normal sequencing is not routinely performed at most institutions, because of its increased cost and workload. While patient-matched analysis offers several benefits, we demonstrated here that unmatched sequencing offers satisfactory performance in the detection of ERBB2 CN variants in the vast majority of cases of USC.

NGS is highly specific for gene amplification, however it is less sensitive than IHC or FISH in assessing tumors with focal or low levels of ERBB2 amplification. The predictive and prognostic significance of this subclonal amplification in USC is unknown. Heterogeneous HER2 expression has been documented in a number of tumor types; while unusual in breast carcinoma [58,59,60], heterogeneity is more common in gastric [61,62,63], bladder [64], and lung [65] tumors with ERBB2 amplification. Intratumoral heterogeneity may be a relatively common phenomenon in USC as well, ranging from 11% of cases in this cohort to 31% of cases in a previous report [21]. Heterogeneity of ERBB2 amplification is one possible mechanism of resistance to HER2 therapies. Another important consequence of such heterogeneity is that HER2 status may vary between biopsy, hysterectomy, and metastases from the same patient [66]. We became aware of five such possible cases which were initially considered for inclusion in this study because they had been previously tested by NGS during routine clinical work. However, only limited additional material from a different site which was not tested by NGS was available for further IHC and FISH studies. In these five cases, the CN determined by NGS at one site differed from the combined IHC/FISH interpretation at a different site. In three of the cases, NGS was performed on a metastasis but only unstained slides were available from the hysterectomy (or vice-versa), and in two of the cases, NGS was performed on one block from the hysterectomy but additional material was only available from a different block from the hysterectomy specimen. In four of the cases, there was insufficient material available to perform confirmatory IHC, FISH, and sequencing on tissue from the same site. In the fifth case, sequencing was attempted on the specimen that already had IHC and FISH but failed sequencing due to poor quality control metrics.

At our institution, NGS testing is usually performed on the hysterectomy specimen, and therefore a non-amplified result (or negative HER2 IHC on the hysterectomy) should prompt consideration of retesting of a metastasis. Consequently, we suggest testing should be performed on a metastatic site if a patient is being considered for adjuvant therapy and material is available. Similarly, we recommend confirmatory IHC or FISH studies should be performed on the same block as the NGS testing. Ultimately, while the analytical validity of this technique is high, the clinical utility, likely determined by the implications of HER2 heterogeneity, is still unknown. Further work is needed to determine the predictive and prognostic implications of heterogeneous HER2 expression in USC.