Stephen V. Liu, MD:
NRG1 is a unique target in terms of its biology and detection.^1-4 When we talk about actionable targets in oncology, we usually are referring to receptor tyrosine kinases that are constitutively active with treatments designed to interrupt signaling. However, NRG1 is a very different paradigm because it serves as a ligand, typically for HER3 (ErbB3) or HER4. NRG1 has an EGF‑like domain that promotes activation of various cellular signaling pathways, often by engaging the HER3 receptor and causing HER3 to heterodimerize with the HER2 receptor.

We do not frequently discuss HER3 as an actionable target, as it has minimal intrinsic kinase activity. Instead, one main focus has been on the heterodimerization partner. Because the activation of HER2 triggers downstream signaling cascades, therapeutic strategies have mainly targeted HER2.

Stephen V. Liu, MD:
NRG1 fusions are pathologically heterogenous, and various fusion partners have been identified.⁵ The pathologic fusions that have been described retain the EGF‑like domain of NRG1, resulting in aberrant expression of the domain the fusion partner influences the localization of the NRG1 fusion, usually providing a transmembrane domain to tether the EGF‑like domain of NRG1 to the cell surface in proximity to HER3.

Stephen V. Liu, MD:
With NRG1 in proximity to its receptor, pathologic activation and pathophysiologic signaling occur, often mediated by HER2.

Stephen V. Liu, MD:
Binding of the EGF-like domain of NRG1 to HER3 in an autocrine or juxtacrine manner leads to the formation of HER2/HER3 complexes, which in turn activates downstream signaling pathways. NRG1 also can cause paracrine activation, which is largely driven by the fusion partner.⁶

NRG1 fusions are molecularly heterogenous events because NRG1 has many different fusion partners and comutations. Although the significance of specific fusion partners is currently unclear, it is likely that not all NRG1 fusions are equal. The fusion partner may be important in terms of how the therapeutic product is designed, but we are not yet at the point where fusion partners are used to guide therapy.²

Stephen V. Liu, MD:
In a retrospective analysis that I led with Dr. Sushma Jonna, we investigated the incidence of NRG1 fusions across multiple solid tumor types. We assessed tumor specimens using anchored multiplex PCR and targeted RNA sequencing. Across all cancer types investigated, we found that NRG1 fusions are rare, occurring in 0.2% of cases.² In this and other studies, there was enrichment for NRG1 fusions in invasive mucinous adenocarcinoma of the lung, with a reported incidence of approximately 10% to 30%.⁷

Also, nearly all NRG1 fusion–positive pancreatic adenocarcinomas are KRAS wild-type.^7-9 Overall, NRG1 fusions occur in many common cancer types, including pancreatic cancer, colorectal cancer, breast cancer, non‑small-cell lung cancer (NSCLC), cholangiocarcinoma, ovarian cancer, and certain types of sarcomas.

Stephen V. Liu, MD:
NRG1 fusions can coexist with other oncogenic drivers. For instance, NRG1 fusions occur concurrently with tumor suppressor genes, such as TP53 mutations.¹⁰ An assessment of the immunophenotype of patients with NRG1 fusion–positive lung cancer demonstrated that the median tumor mutational burden in lung cancer was lower than that for patients with other gene fusions, such as NTRK1/2/3 fusions: 0.9 mutations/megabase (95% CI: 0.9-1.8) vs 4.9 mutations/megabase (95% CI: 3.0-10.5); P <.003.¹¹ In this study, PD-L1 was expressed in 13 of 46 (28%) NRG1 fusion–positive lung cancer cases, with a PD-L1 expression ≥50% observed in 4% of cases.¹¹ However, the significance of these findings is uncertain.

Some investigators think that NRG1 fusions can mediate acquired or intrinsic resistance to other actionable drivers. The true incidence of the coexistence of NRG1 fusions with other oncogenic drivers in solid tumors is difficult to estimate because NRG1 fusions can be a challenge to detect with standard assays, and the tests that can detect them are not always performed. It is important to perform very specific types of tests to optimize the detection of NRG1 gene fusions. Dr. Benayed, can you help explain that?

Ryma Benayed, PhD:
When looking for a common gene fusion event that occurs frequently in the tested tumor samples, single-gene testing using reverse transcriptase PCR, FISH, or IHC is a reasonable approach. Of note, reverse transcriptase PCR is not ideal if the fusion can involve multiple gene partners because in such situations, a multiplexing approach can be complex to design, making FISH or IHC better choices. In this context, if the result is negative, it could be suspicious, so further confirmation will need to be obtained by NGS. However, in some cases the identification of the gene fusion partner may be needed for a patient to be eligible to participate on a clinical trial or receive a particular treatment. In such cases, NGS also will need to be performed for verification.

As Dr. Liu indicated, if a gene fusion event is known to occur at a low frequency, such as with NRG1 fusions, then a comprehensive NGS approach is required. Usually, the first step is to perform targeted DNA sequencing to look for any driver or targeted event involving any class of alterations, which could be a mutation, copy number alteration, or gene fusion.

When DNA sequencing is performed and no oncogenic driver is identified, the next step should be to perform targeted RNA sequencing to make sure that gene fusions are not missed. Although very rare, there are times when confirmation of RNA sequencing results will need to be performed by IHC or FISH.

Ryma Benayed, PhD:
One of the ways to perform FISH is the break-apart method, which involves 2 probes labeled with 2 different fluorescent dyes.^12,13 One probe is designed to bind to the 5’ end of the fusion break point, and the other binds to the 3’ end. The presence or absence of a gene rearrangement or gene fusion is determined using the signal patterns. The presence of a split signal indicates that a gene fusion event has taken place, and the absence of a split signal indicates that the NRG1 gene has remained intact.

FISH-based assays have been widely used for many years. FISH can be rapidly performed, with a turnaround time as short as 1-2 days. However, FISH is not scalable, and only a single gene can be detected with an assay. Multiplexing with FISH cannot be easily done. With FISH, if a gene fusion is identified, the identity of the fusion partner remains unknown.

Ryma Benayed, PhD:
The detection of NRG1 fusions by IHC relies on the identification of the phosphorylated ErbB3 (HER3) using an antibody. For the detection of NRG1 gene fusions, IHC has excellent sensitivity of 100%, with a 97% reported specificity in a cohort of formalin-fixed paraffin-embedded (FFPE) lung cancer samples.¹³ It also has a high correlation with FISH results. IHC has been widely used for many years, and it is fast and easy to implement in clinical laboratories. However, similar to FISH, it does not inform about the identity of the fusion partner. Although it is possible to perform multiplex IHC assays, doing so in the context of gene fusion identification has not been clinically proven or validated. So, comprehensive testing using IHC is still challenging.

Ryma Benayed, PhD:
DNA sequencing is a comprehensive approach for the detection of drivers and targetable alterations, and the hybridization capture‑based NGS approach is a popular targeted DNA–based sequencing method. Using this approach, the first step is to extract DNA from tissue, and the extracted DNA is used to create libraries, which allow for the analysis of multiple exons or genomic regions at the same time. The libraries include patient-specific index sequences, which allow for the analysis of multiple patient samples simultaneously, thus increasing throughput. Gene‑specific capture probes are added to specifically enrich for the genes of interest.

Of importance, when we design the capture probes for NRG1 gene fusions, we must design them specifically to target intronic sequences. This is because most of the DNA structural rearrangements involve intronic regions that are not present in the exons.

Ryma Benayed, PhD:
The targeted DNA sequencing assay at our institution has probes capturing 70 introns from 20 genes known to be involved in rearrangements.^14-15 However, NRG1 is not among the 20 genes because it is technically challenging to design capture probes for the introns of NRG1. To partially compensate for this, we can create probes to detect most, although not all, CD74-NRG1 fusions. If CD74 is not the fusion partner, our DNA-based sequencing platform will not detect the NRG1 fusion.

Ryma Benayed, PhD:
For DNA-based NGS assays, it is technically challenging to capture the introns in some genes, such as NTRK3 and NRG1, that are known to have very large introns. Hence, designing capture probes for those large genomic regions is quite challenging and will lead to excessive sequencing.¹⁶ Also, there could be repetitive elements within those large introns, causing difficulties in aligning the sequences back to the genome, and designing probes for those repetitive elements is not an easy task. Sometimes the biopsy sample may have low tumor purity, whereby it is difficult to detect the gene fusion in the isolated DNA due to limited assay sensitivity. Furthermore, some gene fusions result from complex genomic events that cannot be captured in the DNA. Taken together, all these identified issues can be addressed by performing RNA-based NGS.

Of the 60,000 tumor specimens that underwent molecular profiling by DNA-based NGS at our institution, only 0.04% were positive for NRG1 fusions. As previously noted by Dr Liu, NRG1 fusions are rare, with a prevalence of 0.2% across all tumor types.² This further indicates that DNA-based NGS is not the optimal approach for the comprehensive detection of NRG1 fusions.

Ryma Benayed, PhD:
Our approach is to use a PCR‑based technology for targeted RNA-based NGS called anchored multiplex PCR.¹⁷ This approach, using a validated custom panel of primers to detect gene fusions, is unique because there is no need to design a primer for every possible gene fusion partner if a unidirectional primer for the gene targets is included.

For NRG1 gene fusion detection, a primer will be designed for NRG1, and another primer will be designed to bind the sequencing barcode ligated to the complementary DNA libraries created. With the use of this approach, everything between these 2 primers will be amplified and sequenced, and any new, known, or novel NRG1 fusion partner gene will be identified.

Because NRG1 has multiple isoforms resulting from alternative splicing or the use of different promoters, it can be challenging to design NRG1 primers that can capture the various isoforms. The most common isoform, however, is NM_004495, and in my institution, the target exons for this isoform are exons 1, 2, 3, and 6. Exons 2 and 6 are the most commonly reported break points in NRG1 across solid tumors.¹⁸ For some cancer types, such as hepatobiliary and lung cancers, exon 1 has been implicated as a common break point in NRG1.^18,19 The primers are designed in the 5’ direction because the gene partner for NRG1 is upstream. Using this technology, we can include anywhere from 50-200 genes in a panel.

Ryma Benayed, PhD:
We performed a study at the Memorial Sloan Kettering Cancer Center in which we molecularly profiled 2522 lung adenocarcinoma samples using targeted DNA sequencing.²⁰ Of these, 1933 were positive for oncogenic drivers, and 589 (23%) were driver negative. Of these, only 254 samples had adequate tissue for RNA extraction with sufficient RNA quality, and we experienced a technical failure in 22 cases. In total, RNA-based NGS was performed on 232 samples previously identified as driver negative by targeted DNA–based NGS, and 36 of these samples were driver positive by targeted RNA–based NGS. In 33 of these cases, the identified alterations showed actionable in-frame fusions, including MET exon 14 skipping alterations (n = 6), ROS1 (n = 10), and NRG1 (n = 5). Of note, tissue was unavailable for RNA extraction in 314 (53%) of the cases.

So, 5 NRG1 fusion events were missed by DNA-based NGS but detected by RNA-based NGS. Different fusion partners were identified, but 2 of these were CD74-NRG1 fusions. As I previously stated, targeted DNA-based NGS does not capture every single intron in every gene, such as CD74.

Ryma Benayed, PhD:
Across different tumor types tested at the Memorial Sloan Kettering Cancer Center, NRG1 fusions were detected in 34 samples by RNA-based NGS. In 28 of those samples, NRG1 fusions were undetected by DNA-based NGS. There was concordance between RNA-based NGS and DNA-based NGS in only 2 samples. In 4 of the samples, DNA-based NGS was not performed. This further illustrates and highlights the importance of and differences between comprehensive DNA and RNA sequencing in detecting NRG1 fusions in tumor samples.

Ryma Benayed, PhD:
The most common NRG1 gene fusion partners reported in lung adenocarcinoma are CD74, SLC3A2, and SDC4.^2,7,19 In pancreatic cancer, ATP1B1 seems to be the most common fusion partner. In other tumor types—such as breast cancer, sarcoma, and cholangiocarcinoma—there are other different fusion partners, of which some already are known and some are novel. As Dr. Liu stated earlier, the significance of each fusion partner remains unclear.

Ryma Benayed, PhD:
At our institution, we have implemented a workflow where we perform complementary DNA- and RNA-based NGS for the optimal detection of NRG1 fusions.

As initially stated, due to the inadequate availability of tumor tissue for RNA extraction, we were unable to perform the needed RNA sequencing in >50% of samples in our lung cancer study.²⁰ Hence, for these samples, we could not ensure that no gene fusion events were missed. To avoid this, our approach is to initially extract DNA from FFPE samples that we receive. We deparaffinize the sample, lyse the cells, and use the lysate for automated DNA extraction. Rather than discarding the remaining 60 microliters of lysate left in the vial by the robot, we keep it for future RNA extraction. We have validated that the lysate remains stable at 4°C for 2 years.

The extracted DNA is subjected to targeted DNA sequencing, and if an actionable oncogenic driver is identified, the patient goes on to receive targeted therapy or enrolls in a clinical trial. If no driver is identified by DNA sequencing, the remaining lysate sample is reflexed for RNA sequencing. We have been doing this consistently since 2018. In this time, 4116 of 27,000 samples (15.2%) have undergone RNA sequencing because no oncogenic drivers were detected by DNA sequencing.

Stephen V. Liu, MD:
The detection of NRG1 fusions is clinically relevant and is not just an esoteric classification. We recently published the results from a global eNRGy1 registry.¹¹ This retrospective study characterized 110 pathologically confirmed cases of NSCLC harboring NRG1 fusions. It is the largest reported series of NRG1 fusion–positive cases to date.

In this study, we described patterns of distant metastases in NRG1 fusion–positive lung cancer, including metastasis to bone, brain, and kidney, and noted that these events are more common in nonsmokers (57%). We confirmed that there are many different NRG1 fusion partners, and the clinical significance of this observation is still unclear.¹¹

Stephen V. Liu, MD:
The most important findings were the clinical outcomes, as patients with lung cancer harboring NRG1 fusions did not seem to respond particularly well to standard therapies.¹¹ The objective response rate to platinum doublet chemotherapy was 13%, which is far lower than expected in advanced NSCLC. The objective response rate for single-agent immunotherapy was 20% and for targeted treatment with afatinib was 25%. None of the patients responded to chemoimmunotherapy. Although there are limitations with retrospective studies with several other potential confounding factors, overall the clinical outcomes with chemotherapy and/or immunotherapy were rather poor in this study.¹¹

Stephen V. Liu, MD:
Afatinib is a second-generation, irreversible pan‑HER kinase inhibitor that blocks the trans-phosphorylation of HER3 and downstream signaling pathways.^23,25 Although case reports have shown some dramatic and durable responses with afatinib in patients with NRG1 fusion–driven solid tumors,²³ the challenge with case reports and case series is that they are associated with significant reporting bias. So, the actual response rate with afatinib for NRG1 fusion–driven solid tumors remains unclear. However, the case studies demonstrate proof of concept that afatinib has the potential to be an effective drug for patients with solid tumors harboring NRG1 fusions.

We need prospective studies to demonstrate the efficacy of afatinib in this setting. The ongoing TAPUR study is a multiarm, prospective phase II study investigating currently available, targeted anticancer agents for patients with advanced cancer with tumors harboring a genomic variant that is known to be a drug target or known to predict disease sensitivity to a drug (NCT02693535). In one of the arms of the TAPUR trial, patients with cancers harboring NRG1 fusions will receive afatinib. This trial will inform on the response rate of NRG1 fusion–positive solid tumors to afatinib.

Stephen V. Liu, MD:
Unlike afatinib and tarloxotinib, which target the kinase domain of HER, seribantumab targets the HER3 receptor. Seribantumab is a fully human anti-HER3 monoclonal antibody that was previously explored in breast cancer. Seribantumab prevents ligand-dependent activation and phosphorylation of ErbB3. It also acts by inhibiting HER2/HER3 dimerization and downstream signaling pathways.²⁷

Stephen V. Liu, MD:
In a phase I dose-escalation study of seribantumab monotherapy for 44 unselected patients with advanced or refractory solid tumors, the recommended phase II dose of seribantumab was a 40-mg/kg loading dose followed by 20-mg/kg weekly maintenance dosing.²⁸ In this study, seribantumab monotherapy was safe across all studied dose levels.

Seribantumab is being investigated in the ongoing, prospective, multicohort, single-arm phase II CRESTONE trial for patients with recurrent locally advanced or metastatic solid tumors harboring NRG1 fusions (NCT04383210). This study will assess seribantumab at various dose levels to determine the objective response rate.

Stephen V. Liu, MD:
Zenocutuzumab (MCLA‑128) is a bispecific antibody directed at HER2 and HER3. It inhibits NRG1 engagement with HER3 and HER2/HER3 heterodimerization. This agent also has enhanced antibody-dependent cellular cytotoxicity activity. It has shown some early efficacy signals and tolerable safety profile across different tumor types.^5,24 In January 2021, zenocutuzumab was granted FDA fast track designation for NRG1 fusion–positive cancers.

Stephen V. Liu, MD:
An ongoing open-label, multicenter phase I/II trial is investigating the pharmacokinetic/pharmacodynamic profile, efficacy, safety, and tolerability of zenocutuzumab in patients with solid tumors harboring NRG1 fusions (NCT02912949).