Association of baseline ROR1 and ROR2 gene expression with clinical outcomes in the I-SPY2 neoadjuvant breast cancer trial

Purpose ROR1 and ROR2 are Type 1 tyrosine kinase-like orphan receptors for Wnt5a that are associated with breast cancer progression. Experimental agents targeting ROR1 and ROR2 are in clinical trials. This study evaluated whether expression levels of ROR1 or ROR2 correlated with one another or with clinical outcomes. Methods We interrogated the clinical significance of high-level gene expression of ROR1 and/or ROR2 in the annotated transcriptome dataset from 989 patients with high-risk early breast cancer enrolled in one of nine completed/graduated/experimental and control arms in the neoadjuvant I-SPY2 clinical trial (NCT01042379). Results High ROR1 or high ROR2 was associated with breast cancer subtypes. High ROR1 was more prevalent among hormone receptor-negative and human epidermal growth factor receptor 2-negative (HR-HER2-) tumors and high ROR2 was less prevalent in this subtype. Although not associated with pathologic complete response, high ROR1 or high ROR2 each was associated with event-free survival (EFS) in distinct subtypes. High ROR1 associated with a worse EFS in HR + HER2- patients with high post-treatment residual cancer burden (RCB-II/III) (HR 1.41, 95% CI = 1.11–1.80) but not in patients with minimal post-treatment disease (RCB-0/I) (HR 1.85, 95% CI = 0.74–4.61). High ROR2 associated with an increased risk of relapse in patients with HER2 + disease and RCB-0/I (HR 3.46, 95% CI = 1.33–9.020) but not RCB-II/III (HR 1.07, 95% CI = 0.69–1.64). Conclusion High ROR1 or high ROR2 distinctly identified subsets of breast cancer patients with adverse outcomes. Further studies are warranted to determine if high ROR1 or high ROR2 may identify high-risk populations for studies of targeted therapies. Supplementary Information The online version contains supplementary material available at 10.1007/s10549-023-06914-2.


Introduction
ROR1 encodes a developmentally restricted type I receptor tyrosine kinase-like orphan receptor, [1][2][3][4] which we identified was a receptor of Wnt5a. [5] ROR1 expression is prominent in embryogenesis, attenuates during fetal development, and is minimal in post-partum tissues. However, ROR1 is expressed by neoplastic cells of many cancer types making it a potential target for cancer therapy [5,6]. High-level expression of ROR1 on breast cancer cells has been associated with epithelial-mesenchymal transition (EMT), tumor cell proliferation, and metastases [7]. In chronic lymphocytic leukemia (CLL), high-level expression of ROR1 associates with more-rapid disease progression and shorter survival.

Patients and the I-SPY2 trial
We interrogated the clinical significance of high-level gene expression of ROR1 and/or ROR2 in 989 patients with stage II or III breast cancer and high-risk disease by clinical criteria (HR-HER2-or HER2 +) or high-risk disease according to the 70-gene signature. [15] Patients were enrolled in one of nine completed/graduated/experimental and control arms in the multi-center, multi-arm neoadjuvant I-SPY2 clinical trial (NCT01042379, IND 105,139) as depicted in Supplemental Fig. 1. Detailed descriptions of the I-SPY2 study design, eligibility, and assessments are as reported [16][17][18][19][20][21][22].

Ethics
Institutional Review Boards at all participant institutions approved the protocol. All patients provided signed informed consent to allow for research on their biospecimen samples in association with clinical outcome data.

Datasets
Platform corrected, log2-transformed, and normalized genelevel transcriptomic data generated from pretreatment tumor samples assayed on Agilent 44 K expression arrays were obtained from NCBI's Gene Expression Omnibus (GEO) (GSE194040). We obtained patient-level scores from expression signatures reflecting estrogen receptor signaling, HER2 signaling, and proliferation from the supplemental data of the associated publication. [22]

Statistical analysis
We assessed association between ROR1 or ROR2 gene expression levels and hormone receptor (HR) and human epidermal growth factor receptor 2 (HER2) defined subtypes using a Kruskal-Wallis test with post hoc pairwise comparisons by Wilcoxon-rank sum tests with default (Holm) adjustment for multiple hypothesis testing. We used logistic regression to assess association between ROR1 or ROR2 expression levels and pathologic complete response (pCR) with significance assessment, using the likelihood ratio test comparing models with or without the biomarker term. We performed analyses with univariate and multivariate models, adjusted for subtype and treatment, conducted withinsubtype analyses, with and without adjusting for treatment, as well as exploratory analyses within subtype and within arm. We used multivariate Cox proportional hazard modeling to assess association between ROR1/ROR2 expression levels and EFS with significance assessment, using the likelihood ratio test (comparing models with/without the biomarker term). These analyses were performed in the overall population adjusting for subtype and treatment and extent of residual disease (RCB-0/I vs. RCB-II/III [23]), and among RCB-0/I and RCB-II/III patients, adjusting for subtype and treatment; within-subtype analyses adjusting for treatment, for treatment and extent of residual disease, and among RCB-0/I and RCB-II/III patients. Association between ROR2 expression levels and subtype, pCR, and EFS 1 3 were similarly evaluated. Pearson correlation was used to assess correlations between the expression levels of ROR1 and ROR2 with expression levels of EMT-related pathway genes, including Hippo/Yap/TAZ, WNT5A, BMl1, BCL2, and GLI1, as well as two ER-related, two HER2-related, and two proliferation-related expression signatures. In addition, we also compared expression levels of these genes and signatures between patient subsets defined by ROR1 and ROR2 expression levels (above versus below the median) using the ANOVA F test. All analyses were performed using R version 3.6.3 without adjustment for multiple hypotheses testing.
The analysis reported here is a biomarker study of the gene expression of ROR1 and ROR2 leveraging data from the I-SPY2 clinical trial. The patient population, specimen collection, assay methods, and trial design were all previously described, and the sample size could not be changed for this study. REMARK criteria were used to report the data. [24].

Expression of ROR1 and ROR2 in breast cancer subtypes
We examined the expression levels of ROR1 at baseline by subtype of breast cancer, Fig. 1a. We observed a wide range of ROR1 expression levels in all subtypes. We noted HR-HER2-breast cancers expressed the highest levels of ROR1, followed by cancers with the HR-HER2 + subtype. HR + HER2-tumors expressed lower levels, which were not significantly different from that of HR + HER2 + tumors. In contrast to what we found for ROR1, the expression levels of ROR2 were lowest in HR-HER2-breast cancers and significantly lower than that found in other subtypes; the highest ROR2 expression levels were observed in the HR + HER2 + subtype, Fig. 1b. A weak-positive correlation was observed between the expression levels of ROR1 and ROR2 , Fig. 1c. When dichotomized at the median expression levels of ROR1 and ROR2 and divided into 4 subgroups, we observed an association between the subgroups defined by high-level expression of ROR1 and ROR2 in breast cancer subtypes, Fig. 1d. The choice of median cut-point for high-versus low-level expression was based upon our prior study using the median cut-point in analyzing ROR1 expression in breast cancer datasets before and after neoadjuvant treatment. [25,26] High-level expression of ROR1 and ROR2 was noted in a large percentage of HER2 + specimens, whereas high-level ROR1 and low-level ROR2 were more common in HR-HER2-tumors. Consistent with the enrichment for HR-HER2-tumors, the subset with high ROR1 and low ROR2 has the lowest expression levels of ER-and HER2-related signatures and the highest gene expression signatures associated with proliferation, Supplemental Table S1.

Expression of ROR1 and ROR2 and likelihood of pCR
Analysis of likelihood of pCR (RCB-0) [23] by overall population and by subtype revealed a wide range of ROR1 expression levels within both pCR and non-pCR groups. Although higher expression levels of ROR1 associated with non-pCR in the HR-HER2-subtype, Fig. 2a, the higher ROR1 expression observed in non-pCR patients did not retain significance when adjusted for treatment arm. Moreover, there was no apparent association with pCR in other breast cancer subtypes, Supplemental Table S2. Exploratory analysis of pCR in HR-HER2-patients by treatment arm indicated a trend toward negative association of high ROR1 expression and pCR in 5 of the 8 treatment arms with a notably strong signal in the 32 patients treated on the MK2206 (AKT inhibitor) arm, Fig. 2b. Analysis of ROR2 expression in relationship to pCR revealed that high-level ROR2 was not associated with pCR in the overall population or in any subtype, Supplemental Table S2. Therefore, neither high-level ROR1 nor high-level ROR2 was associated with the likelihood of pCR.

Association of ROR1 / ROR2 and EFS
Breast cancers from patients who had high-level expression of ROR1 had a worse EFS when adjusted for subtype and treatment arm (HR 1.2, 95% CI = 1.03-1.40, LRp = 0.02), Table 1. When assessed in the context of residual disease, high-level expression of ROR1 associated with a significantly worse outcome for patients with HR + HER2-tumors who had a high post-treatment residual cancer burden (RCB-II/III) (HR = 1.41, 95% CI = 1.11-1.80, LRp = 0.01). However, we did not , which may in part be attributable to a smaller number of events within the RCB-0/I group. We did not observe an association between high-level expression of ROR1 and worse EFS among patients with HR-HER2-or HER2 + cancer subtypes. Inclusion of ROR2 in the analysis model did not change these findings. Kaplan-Meier exploratory analysis by ROR1 above or below the median level indicated that HR + HER2-patients with high-level ROR1 at baseline and high-tumor burden after treatment (RCB-II/III) had significantly worse EFS, (HR = 0.55, 95% CI = 0.33-0.9, Fig. 3a. Kaplan-Meier plots further stratified by RCB class showed that high ROR1 in HR + HER2patients with RCB-III had the worst outcome, Supplemental Figure S2a. Patients with breast cancers exhibiting high-level expression of ROR2 did not have a significant difference in EFS compared to patients with tumors with low-level expression of ROR2 when adjusted for subtype and treatment arm (HR = 1.09, 95% CI = 0.94-1.27, LRp = 0.27), Table 1. However, after adjustment for subtype, treatment, and RCB class, patients with HER2 + subtype tumors and minimal residual disease after treatment (RCB-0/I) had significantly worse EFS (HR = 3.46, 95% CI = 1.33-9.02, LRp = 0.01,) Table 1, if their breast cancers expressed high levels of  359 1.32 (1.05-1.68) 0.02 1.33 (1.04-1.71) 1.11-1.80) 0.01 1.46 (1.14-1.88) 245 1.36 (1.05-1.75) 0.02 1.42 (1.09-1.86)

Expression of ROR1 associates with high-level expression of EMT-related genes
We performed hierarchical clustering of ROR1 with 24 EMTrelated genes including the Hippo signaling pathway genes from the MSigDB database [27] along with WNT5a, BMI1, BCL2, and GLI1 prompted by associations noted in prior studies on breast cancer or CLL [25,[28][29][30]. As shown in Fig. 4

Discussion
Using the annotated I-SPY2 transcriptome data from a cohort of nearly 1000 patients with newly diagnosed highrisk early breast cancer, we found that high-level pretreatment expression of ROR1 or ROR2 had a distinct subtypespecific association with adverse risk. High-level expression of ROR1 was highest in HR-HER2-subtype and was associated with worse EFS in HR + HER2-patients with high posttreatment residual cancer burden (RCB-II/III). High-level expression of ROR2 was lowest in the HR-HER2-subtype of breast cancers and higher ROR2 expression was associated with worse EFS in HER2 + patients with minimal residual disease after therapy (RCB-0/I). High-level ROR1 or high-level ROR2 each was associated with high-level expression of genes involved in EMT. Although not correlated with pCR, high-level expression of ROR1 or ROR2 distinctly identified breast cancer patients with different tumor subtypes with adverse outcomes. This study highlights the potential prognostic value in assessing the levels of ROR1 and/or ROR2 in untreated high-risk early-stage breast cancer and justifies further studies to evaluate the biology and possible value of targeting ROR1 and ROR2 with investigational treatments. Prior studies from our group showed an association of ROR1 signaling with stem cell features, EMT, proliferation, and metastases in preclinical models; moreover, the apparent reversal of such features by treatment with an inhibitory anti-ROR1 antibody justified correlative studies of ROR1 expression with response and clinical outcome [25]. Interrogation of tumor biopsies from 122 patients before and after neoadjuvant chemotherapy revealed the expression level of ROR1 was increased in residual breast cancer cells after surgery and was associated with enhanced expression of genes associated with EMT, proliferation, and cancer stemness. [25] Therefore, studies of the expression levels of ROR1 and ROR2 on post-treatment surgical specimens in the I-SPY2 transcriptome dataset, when it becomes available, may provide biologic insights, inform future clinical trials, and determine the optimal tissue and timing for assay.
An exploratory analysis of pCR in HR-HER2-patients by treatment arm indicated a negative trend for the association of high-level ROR1 with pCR in 5 of the 8 treatment arms with a notably strong signal in the 32 patients treated with MK2206, an AKT inhibitor. This strong signal with MK2206 is not surprising as ROR1 signaling activates the PI3K/AKT/MTOR pathway [25] and high-level expression of ROR1 may mitigate the anti-tumor activity of an AKT inhibitor in combination with chemotherapy. This observation suggests that investigations of ROR1 blockade with inhibitors of AKT signaling may be informative.
The results for ROR2 expression significantly extend the prior observations that ROR2 signaling also may contribute to breast cancer progression and/or tissue invasiveness [9,10]. Studies have shown that ROR2 may regulate the balance of Wnt signaling and cellular heterogeneity during tumor progression. [31] To our knowledge, our study is the first to evaluate the expression levels of ROR1 and ROR2 in the same large clinical dataset and to evaluate the association of ROR2 expression with response and EFS by subtype. Our findings that elevated ROR2 expression associated with worse outcome in HER2 + patients with minimal post-treatment residual cancer burden (RCB-0/I) was based on a small number of events. As such, analyses of additional datasets are warranted to determine if high-level expression of ROR2 is associated with adverse outcomes and to determine other factors that may impact outcome in this patient subset.
Our study has several strengths and limitations. Strengths include the fact that the I-SPY2 trial platform includes robust correlative science on serial tumor biospecimens, an active control arm, and contemporary chemotherapy backbone [19]. As a result, we were able to evaluate associations of pretreatment ROR1 and/or ROR2 with chemotherapy response by pCR and clinical outcome by EFS. Of note, I-SPY2 eligibility requires that all tumors be clinically or molecularly high risk and patient performance status be excellent. The average age of enrolled patients is more than 10 years younger than that of typical breast cancer patients. [19] Therefore this study may not reflect ROR1 and ROR2 expression in the typical patient with breast cancer.
Potential limitations of our study include the analysis of pretreatment tumor specimens only, analysis of gene expression only, and use of Agilent 44 K platform which cannot distinguish between RNA isoforms of ROR1 or ROR2 that can or cannot be expressed as cell surface proteins [14]. Additionally, multiple hypothesis testing and small numbers of events in many categories limit statistical power. We analyzed the I-SPY2 transcriptome dataset of baseline pretreatment tumor specimens for expression of ROR1 and ROR2 as a transcriptome dataset for post-neoadjuvant surgical tumor tissue is not yet available. Breast cancer biology, hormone receptors, subtype frequency, and mutations can evolve over time under the pressure of systemic therapy; therefore, pretreatment tumor specimens may have different biomarker expressions than post-treatment tumor specimens. [32] However, current biomarkers with clinical utility in early breast cancer are based on assays of pretreatment tumor specimens justifying the current investigation of pretreatment specimens. Future studies of post-treatment surgical specimens, when available, and metastatic specimens are warranted to determine the optimal timing for assessment of ROR1 and ROR2 to inform clinical trials of targeted agents.
As noted, the expression or ROR1 or ROR2 may not accurately reflect the expression of ROR1 or ROR2 protein. Therefore, we examined the expression of genes that may be upregulated in breast cancer cells through activation of ROR1 or ROR2 signaling. [25,28] Hierarchical clustering reveals significant associations of ROR1 with 20 of 24 EMTrelated genes, and the strongest association is with WWTR1 (TAZ) a transcriptional coactivator in the Hippo signaling pathway. [33] Similar hierarchical clustering analysis of ROR2 expression and EMT-related genes reveals significant associations of ROR2 with 18 of 24 EMT-related genes with the strongest association with LATS1 (Rp = 0.57), hypothesized to be a tumor suppressor and the main kinase component in the Hippo signaling pathway [34]. Our analysis showed significant, but variable, correlations between WNT5a and ROR1 or WNT5a and ROR2. This observation is expected because WNT5a gene expression can be modulated by many different pathways [35] and not exclusively by ROR1-and ROR2-regulated pathways. Additional correlation analysis of ROR1/ROR2 expression groups by median cut-point high/low status with gene signatures revealed that the high ROR1 and low ROR2 group enriched for HR-HER2-tumors had the lowest expression levels of ER-and HER2-related signatures and the highest expression levels of proliferation signatures. EMT gene and signature expression that were significantly correlated with ROR1 and/or ROR2 expression were also differentially expressed between the four ROR1/ROR2 defined subsets.
Cancer cells may express ROR1 or ROR2 at levels not observed in normal post-partem tissues and, therefore, the protein antigens encoded by these genes could serve as potential targets for therapy. [25] Our group has developed a humanized monoclonal antibody, cirmtuzumab (now designated as zilovertamab), to ROR1. [36] A Phase 1 study in CLL showed that zilovertamab therapy reversed ROR1 signaling and stemness signatures with minimum apparent toxicity. [37] As a result, zilovertamab is currently under study in CLL and mantle cell lymphoma (NCT03088878) and in advanced breast cancer (NCT02776917) with no additional safety concerns reported to date. [38,39] Our group has also developed a ROR1 antibody conjugated to MMAE that has been found to be effective in a Richter's syndrome mouse model [40] and this compound, VLS-101, now zilovertamab vedotin, is under study in hematologic malignancies (NCT03833180) and in solid tumors (NCT04504916). Zilovertamab vedotin was found to have no unexpected toxicities in heavily pretreated patients with lymphoid cancers and to have clinical activity [41]. Other ROR1 [42] and ROR2 targeted therapies are in clinical trials (NCT03504488, NCT03393936).
In summary, we have shown in a cohort of almost 1000 high-risk early-stage breast cancer patients treated on the I-SPY2 platform that pretreatment expression of ROR1 was higher in HR-HER2-subtype, did not correlate with pCR, and was associated with worse EFS in HR + HER2-patients with high post-treatment residual cancer burden (RCB-II/III). We found that expression of ROR2 was lowest in HR-HER2-breast cancer, highest in the HR + HER2 + subtype, and did not correlate with pCR. High ROR2 identified a subset of HER2 + patients who had an excellent response to neoadjuvant treatment (RCB-0/I) but had a higher risk of relapse. Agents targeting ROR1 and ROR2 are in clinical trials and may provide new investigational opportunities. Importantly, these results warrant further studies to determine the value of using high-level expression of ROR1 and ROR2 as markers for poor outcome that may inform clinical trials of targeted therapies. Consent to participate All patients provided signed informed consent to allow for research on their biospecimen samples in association with clinical outcome data.
Consent to publish No individual patient data or individual patient images are included in this manuscript.

Electronic figure submission
Electronic figures were generated from R version 3.6.3, assembled and formatted using Adobe Illustrator, and embedded in the text body of the manuscript in Word as requested.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.