Bone-Induced Her2 Promotes Secondary Metastasis in HR+/Her2− Breast Cancer

Abstract

Bone metastases can disseminate to secondary sites and promote breast cancer progression creating additional clinical challenges. The mechanisms contributing to secondary metastasis are barely understood. Here, we evaluate the prediction power of Her2-expressing (Her2E) circulating tumor cells (CTCs) after analyzing over 13,000 CTCs from a cohort of 137 metastatic breast cancer (MBC) patients with initial HR+/Her2− status and employ preclinical models of bone metastasis (BM) to validate the role of Her2E CTCs in multi-organ metastases. While Her2 expression was higher in patients with bone metastasis, experimental analyses revealed that Her2E CTCs derived from bone lesions were more dependent on Her2 activity and more susceptible to anti-Her2 therapy. Targeting the bone-mediated Her2 induction reduces CTC detection and abrogates secondary metastasis from bone. Overall, we elucidate that Her2E CTCs can serve as a non-invasive biomarker for BM formation with high therapeutic benefit for HR+ MBC patients.

Statement of Significance

Given the urgent need for alternative strategies to block metastasis progression, we demonstrate that blocking Her2-mediated secondary metastasis improves clinical outcome and establish Her2 as a biomarker for bone metastasis in patients with initial HR+/Her2− breast cancer which represents ~70% of all cases.

Introduction

Breast cancer can quickly transition from latency to a highly aggressive metastatic state. As the primary site of recurrence in HR+ breast cancer, bone metastasis remains a clinical challenge with most trials revealing controversial results on the overall survival of patients (1–5). With accumulating evidence suggesting cancer cell dissemination from bone to other metastatic sites, we are just beginning to understand the mechanisms involved in the secondary metastasis process(6,7).

To adapt to the bone microenvironment, cancer cells undergo epigenetic and phenotypic changes which may further confer potential for therapeutic escape. This is evidenced through known drivers of therapeutic resistance like receptor tyrosine kinases (RTKs) such as epithelial-, fibroblast-, and platelet-derived growth factor receptors (e.g., Egfr, Her2, Fgfr1, and Pdgfra/b) (8–11). The specific role of osteoblasts in promoting endocrine resistance in HR+ breast cancer is mediated by ER loss in the early stage of metastasis. However, the reversibility of ER status in macrometastases suggests other factors are required to maintain therapeutic resistance following the recovery of ER expression(6). These factors may contribute to cross-resistance events between distant metastatic sites. Hence, it remains urgent to identify stable markers associated with secondary metastasis seeding.

The aforementioned RTKs may provide alternative survival pathways during BM (9,12). This is particularly true for HR+/Her2− metastatic breast cancer (MBC) patients who develop Her2-expressing (Her2E) tumors during disease progression. While anti-Her2 therapy was traditionally considered poorly beneficial for Her2− breast cancer, some studies suggested otherwise(13). Additionally, recent clinical trials targeting Her2-low cancers using more advanced therapeutic approaches, revealed advantageous effects on both progression-free and overall survival (Destiny 4, NCT03734029) (14,15). Her2 expression in HR+ tumors is associated with increased stemness as depicted by ALDH1 and SOX9, and regulates lung metastasis tropism via CXCR4 (16,17). Preclinical studies suggest that the majority of disseminated tumor cells (DTCs) in the bone marrow are Her2+ with higher metastatic potential (18). Despite these associations, Her2 expression in primary tumors does not predict BM occurrence in breast cancer and associates with visceral metastasis instead (19). In contrast, HR+ tumors display higher bone tropism while BM couples with increased CTC number (20). Strong expression of Her2 in CTCs was found to align with worse survival in HR+ MBC patients (DETECT Trial, NCT01619111) (21). However, the role of bone metastasis in CTC accumulation and multi-organ metastasis remains unclear.

Our study demonstrates that the bone microenvironment promotes the release of Her2E CTCs which contributes to secondary metastasis formation in HR+ breast cancer. While Her2 expression is essential for the seeding process from bone to other organs, we uncover neuregulin 3 (NRG3) as a mediator of a sustained autocrine circuitry that maintains Her2 signaling during metastasis progression. Additionally, using bone-targeting Her2 antibodies eliminates Her2E CTCs and reduces multi-organ metastasis in established BM models. Overall, we identify Her2E CTCs as BM markers in HR+ breast cancer, which promotes secondary metastasis and creates new therapeutic vulnerabilities.

Results

1. Genomic mutations of ERBB2 are devoid of bone metastasis prediction potential

The clinical association of bone with metastasis recurrence is proven in breast cancer(22). However, whether Her2 is involved in that process remains unexplored. First, we decided to determine whether specific mutations of the ERBB2 gene could contribute to the bone tropism by analyzing the MetTropism dataset which provides genomic information on primary and metastatic lesions from multiple cancer types (23). We curated patients with ERBB2 mutations and stratified them based on their bone metastasis status. Despite the significant association of BM with worse overall survival outcomes in breast cancer, both the distribution and frequency of ERBB2 mutations were similar in MBC patients with either visceral metastasis only (VM) or BM (Fig. 1A,B). We extended the analysis to lung, colorectal, and bladder cancers and the results corroborated our finding in breast cancer with P-values of 0.000137, 0.001279, and 1.75×10⁻¹⁰, respectively (Fig.S1A–C). The mutated sites were highly variable between cancer types but very consistent in BM and VM groups. The most frequent driver mutations (L755S, Y772_A775dup, R687Q, and S310F/Y) were homogeneously represented in both groups. Therefore, we evaluated whether other genetic mutations could contribute to the worse overall survival of BM+ breast cancer patients. Despite focusing on the topmost altered genes, no mutation was significantly enriched in BM (Fig.S1D). Additionally, ~90% of patients with BM were hormone receptor-positive (HR+) at diagnosis (Fig.S1E). Altogether, these results demonstrate that the ERBB2 mutations are not sufficient to predict bone metastasis progression. Therefore, strategies different from circulating tumor DNAs (ctDNAs) need to be explored.

Figure 1: Her2-expressing (Her2E) CTC analysis shows a non-invasive diagnostic potential for bone metastasis in luminal breast cancer.

A- Mutation diagram showing mutation types and position on the Her2 gene (ERBB2). The dots are presented with different colors, specific to each type of mutation as indicated on the figure annotation. Cancer hot spots and OncoKB curated mutations are shown in orange and blue, respectively. The y-axis represents the percentage of ERBB2 mutation detected in each cohort of patients; Bone metastasis-positive patients (bottom) and other metastases (top).

B- Survival plot of breast cancer patients with ERBB2 mutations based on bone metastasis status (Bone met vs Other met). The number of patients (n) is shown on the graph. Significant P-value < 0.05 (Log-rank test).

C- Diagram summarizing the experimental approach used for CTC analysis in patients with ER+/Her2− primary breast cancer.

D- Demographic and clinical characteristics of 137 patients with HR-positive Her2-negative MBC.

E- Experimental design of CTC collection and processing from MBC patients.

F- Representative images of CTCs after CellSearch processing. CTCs are CK positive, DAPI positive, and CD45 negative. Her2 expression was evaluated using the FITC anti-Her2 antibody, and categorized as Her2 negative (0), weak (1+), moderate (2+), and strong (3+). CK (green) and DAPI (purple) are shown as merged images.

G- Quantification of Her2E CTC fraction in breast cancer patients with bone metastasis (BM+) or visceral metastasis (BM-). This is based on 137 patients with a total of 13,372 CTCs. P-value: Chi-Square test. The Donut plot shows the proportion of Her2E CTCs according to bone metastasis status.

H- Heatmap showing the mRNA expression of epithelial growth factor receptors in human CTCs from patients with luminal breast cancer (GSE86978).

I-L: Dot plots comparing the expression of ERBB2, ERBB3, ERBB4, and EGFR in CTCs derived from patients with bone (BM+) or visceral metastasis (VM+). P-values: unpaired one-tailed t-test with Welch’s correction. Error bars: mean +/− SD.

2. Increased frequency of Her2-expressing (Her2E) CTCs in patients with bone metastasis

To evaluate the non-genomic changes associated with bone metastasis, we decided to focus on breast cancer patients with initial HR+/Her2− primary tumors for 2 reasons: (i) HR+ breast cancer patients have a higher tropism to bone and (ii) we have previously demonstrated that cancer cells can disseminate from bone to other distant sites (7). The detection of Her2E CTCs in patients with HR+/Her2− status led us to evaluate whether increased Her2 expression could serve as a diagnostic strategy for BM (Fig. 1C). Hence, we retrospectively analyzed Her2 protein expression levels in CTCs collected from peripheral blood specimens of 137 metastatic HR+/Her2− breast cancer patients (Fig. 1D). A systematic approach was adopted for CTC classification based on Her2 immunofluorescence intensity (Fig. 1E,F). The degree of Her2 expression was variable and fluctuated between 0 and +3 (Fig.S1F). Despite the Her2− status of these patients, ~ 18% of CTCs collected from all individuals with BM were moderately to strongly Her2+ (i.e. 2+ and 3+) against only 5% for patients with VM only (P=0.004). This represents a three-fold increase in circulating Her2E CTCs found in BM patients (Fig. 1G and Fig.S1G). To further validate the results and minimize the effect of CTC variations between patients, we evaluated the proportion of Her2E CTCs in individual patients. We found higher proportions of Her2E CTCs in BM patients (P=0.022) which is consistent with the cumulative analysis (Fig.S1F). However, to establish Her2E CTCs as good markers for bone metastasis, we needed to evaluate Her2 specificity, relatively to other human epithelial growth factor receptor family members. Therefore, we analyzed a CTC transcriptomic dataset from HR+/Her2− breast cancer patients (24), and found that ERBB2 and ERBB3 were the most detectable (Fig. 1H). However, only ERBB2 expression was significantly increased in BM, comparatively to VM patients (Fig 1I,J). In contrast, ERBB4 and EGFR were largely undetected in most CTCs (Fig. 1K,L). Together, the transcriptomic analysis of CTCs supports the association of ERBB2 upregulation in patients with bone metastasis.

3. The bone microenvironment promotes Her2E CTCs

To demonstrate that bone mediates the increased expression of Her2 in CTCs, we decided to use an experimental model known as intra-iliac artery (IIA) injection to deliver cancer cells to the hind limb of mice for bone metastasis formation (Fig. 2A and Fig.S2A). IIA injections specifically enrich cancer cells in bone without physical alteration of the bone matrix integrity (25). We selected two non-metastatic HR+/Her2− breast cancer models (MCF7 and ZR75-1) that require estrogen to grow in vivo and do not produce detectable levels of CTCs from orthotopic tumors. Intriguingly, however, we identified high numbers of CTCs from the peripheral blood of mice with established bone metastasis (Fig. 2B,C). Immunofluorescent imaging revealed that CTCs derived from bone metastasis-bearing mice were predominantly Her2E, although MCF7 and ZR75-1 models are classified as Her2− (Fig. 2D). To ascertain the Her2-promoting effect of the bone microenvironment on CTCs, we adopted two major strategies for control purposes. First, we assessed the impact of short-term exposure of cancer cells to in vivo conditions by injecting cancer cells in the left ventricle of mice before harvesting blood-containing CTCs from the right ventricle. This approach allowed cancer cells to undergo at least one round of systemic circulation. Only ~15% of collected CTCs could be classified as Her2E CTCs compared to ~80% in mice with established BM (Fig. 2E). Secondly, we delivered cancer cells to the lung via tail-vein injection (Fig.S2B). As expected, lung metastasis remained poorly detectable even at 8–12 weeks post-injection, confirming the poor metastasis potential of MCF7 and ZR75-1 to the lung. We could not capture CTCs in these models except in a few random cases (1/18 mice for ZR75-1 and 1/16 mice for MCF7) where the identified CTCs were found Her2 negative (Fig.S2C,D).

Figure 2: Bone metastasis promotes Her2E CTCs in HR+/Her2− xenografts that can be blocked by anti-Her2 therapy.

A- Diagram describing the experimental strategy for bone metastasis and CTC collection. Cells were transplanted via intra-iliac artery injection.

B- Table showing CTC counts in ZR75-1 and MCF7 models of bone metastasis.

C- Representative fluorescent images of CTCs collected from HR+/Her2− breast cancer bone metastasis bearing mice. Her2 (red), Cytokeratin 8 (K8) green; DAPI (blue). Scale bars, 10 a.u.

D- Quantification of bone metastasis-derived CTCs based on Her2 status. P-value: two-tailed paired t-test. Mean +/− SD.

E- Quantification of CTCs following a short-term in vivo exposure via intracardiac injection (IC). P-value: two-tailed paired t-test. Mean +/− SD.

F- Bone metastasis treatment strategy using a bone-targeting anti-Her2 antibody conjugate (Tras-ALN).

G- Tumor growth in IIA-induced bone metastasis models (ZR75-1). Control (ALN-treated; n=14 mice) and anti-Her2 (Tras-ALN; n=14 mice). Treatment started 2 weeks post bone metastasis initiation. P-value: ANOVA; Tukey’s multiple comparisons test.

H- Similar to panel g, using MCF7 as bone metastasis model. (ALN-treated; n=11 mice) and anti-Her2 (Tras-ALN; n=12 mice). Treatment started 2 weeks post bone metastasis initiation. P-value: ANOVA; Tukey’s multiple comparisons test.

I- μCT images illustrating tumor-mediated bone loss in mice.

J- Dot plots showing bone volume changes in tumor-bearing femurs based on μCT images. Each dot represents a mouse. Representative samples for ZR75-1 (N=14) and MCF7 (N=7). P-value: unpaired two-tailed t-test (P<0.05: significant).

4. Blocking the bone-mediated Her2 expression reduces secondary metastasis

Cancer cells with copy number amplification of the Her2 gene (Her2+ BC) generally rely on Her2 signaling for survival and growth as proven by the high therapeutic benefit of anti-Her2 drugs. Hence, the acquired Her2 expression in bone metastasis suggests possible dependencies, especially due to the endocrine resistance phenotype observed in bone. We have previously developed a bone-targeting antibody by conjugating trastuzumab to alendronate (Tras-ALN)(26). The inhibitory effect of Tras-ALN was superior to trastuzumab (Tras) on Her2+ models of bone metastasis. We speculated that targeting the bone-mediated Her2E cells in HR+ breast cancer models may reduce secondary metastasis seeding. Hence, we allowed cancer cells to colonize the hind limb for 2 weeks before treatment (Fig. 2F). Despite the early exposure of tumor-bearing mice to estrogen, Tras-ALN reduced BM growth (Fig. 2G–H) and improved the tumor progression-free survival in both MCF7 and ZR75-1 models (Fig.S2E). This was further illustrated by microCT analysis of tumor-bearing bones (Fig. 2I,J and Fig.S2F). Besides bone, the number of metastases to other organs was also diminished following Tras-ALN as over 75% of mice with multiorgan metastasis came from the control (ALN-treated) group (Fig.S2G–J). Hence, we performed an extensive analysis of metastasis distribution using an ex vivo imaging approach. While 60% of control mice had secondary metastasis in visceral organs (e.g., lung, liver, ovary, and kidney (including the adrenal cortex)) no visceral metastasis was detected in the Tras-ALN treated mice (Fig.S2K–M).

5. ER blockade reinforces the MSC-driven Her2 expression in bone metastasis

We previously observed that Her2 expression was higher in bone-derived breast cancer cells using MCF7 as a BM model (6). Both total and phospho-Her2 proteins were affected (Fig.S3A). Using additional HR+ models, ZR75-1 and HCI-011, we confirmed that Her2 was specifically induced in bone metastasis and not in orthotopic tumors (Fig. 3A). This acquired expression was independent of copy number amplification as demonstrated by Her2-FISH experiments (Fig. 3B and Fig.S3B). However, no acquired Her2 expression was observed in metastasis to the lung and ovary (Fig.S3C). To test whether the bone-mediated effect was relevant to TNBC, we analyzed bone metastasis from SCP21, a model previously derived from MDA-MB-231 cells (27). Her2 remained indetectable in orthotopic and bone lesions, and no changes were found at the transcriptional level when comparing ERBB2 expression in bone, lung, and orthotopic lesions (Fig.S3D).

Figure 3: Mesenchymal Stem Cells (MSCs) induce Her2 expression in HR+/Her2− bone metastasis models of breast cancer.

A- Immunostaining of Her2 in primary tumors and bone metastasis models of ZR75-1 and HCI011. Scale bars: 100μm.

B- FISH analysis of Her2 in control and bone-derived MCF7. scale bars, 50μm.

C- TRAP-sequencing analysis of ERBBs’ expression in MCF7 organoids in monoculture or coculture with MSCs.

D- Effect of ER ligands on ERBB2 mRNA expression in MCF7 monoculture (MSC-) or coculture (MSC+). E2 (Estradiol), ICI (fulvestrant), 4OHT(4-hydroxy-tamoxifen). P-values: two-tailed student t-test. Error bars: mean +/− SD. The linear curve represents the Log10(P-value) in each treatment condition.

E- Diagraph depicting the experimental approach used to assess the response of bone metastases to Her2 inhibitors after two weeks of estrogen depletion and fulvestrant pre-treatment.

F- Representative images showing the response of IIA-induced bone metastasis to anti-Her2 therapy following a fulvestrant pretreatment. Treatment groups: Alendronate (ALN), Trastuzumab (Tras), antibody conjugate (Tras-ALN), and control (PBS).

G- Response of fulvestrant-resistant tumors to vehicle (PBS), Alendronate (ALN), Trastuzumab (Tras), bone-targeting Trastuzumab (Tras-ALN). Error bars: mean +/− Standard Error (SE). BLI is based on whole mouse imaging. Significant P-value < 0.05.

H- Area Under Curve (AUC) analysis of metastasis growth in Alendronate (ALN), Trastuzumab (Tras), and bone-targeting Trastuzumab (Tras-ALN), relatively to the control group (PBS). Significant P-value < 0.05.

I- Proportion of mice with multi-organ metastasis. Control (n=10), Tras (n=9), Tras-ALN (n=9), ALN (n=10). P-value: Fisher’s exact test.

J- Representative immunostaining of metastatic lesions in ovaries from all treatment conditions based on Cytokeratin 19 (K19). Scale bars: 1mm for whole tissue scans and 500um for magnified areas.

Therefore, we sought to investigate whether osteogenic cells can transcriptionally regulate Her2 expression by analyzing a previously published TRAP-sequencing dataset, where mesenchymal stem cells (MSC) were cultured with MCF7 cells in 3D (GSE137270). Relatively to other epithelial growth factor receptor family members, only ERBB2 was specifically upregulated in cancer cells exposed to MSCs (Fig. 3C). Surprisingly, this remained true even in presence of estrogen, but the anti-estrogenic drugs, fulvestrant and tamoxifen, displayed a more significant induction of ERBB2 (P=0.0006 and 0.0019, respectively) (Fig. 3D). The levels of ERBB3 and EGFR were not significantly altered in any treatment conditions (Fig.S3E,F). In contrast, ERBB4 expression was ligand-dependent in monoculture, but this effect was neutralized in the presence of MSCs, leading to a ligand-independent inhibition of ERBB4 (Fig.S3G).

6. Endocrine-resistant bone metastases are more responsive to Her2 inhibition

Based on the preponderant effect of endocrine treatment on Her2 expression in MSC conditions (Fig. 3C,D), we decided to evaluate whether establishing similar conditions in vivo would give rise to stronger responses to anti-Her2 therapy. Cancer cells were delivered to the bone via intra-iliac artery injection and metastasis was allowed to form for 2 weeks before endocrine treatment was applied (Fig. 3E). While bone exposure is sufficient to drive endocrine resistance, applying therapeutic pressure forces resistant cancer cells to rely on other mechanisms for survival. As predicted from the MSC coculture experiment, the inhibitory effect of anti-Her2 therapy was much stronger following neoadjuvant fulvestrant treatment (Fig. 3F–H). Both Trastuzumab (Tras) and Tras-ALN could block bone metastasis (Fig.S3H,I) and multi-organ metastasis (Fig. 3I,J and Fig.S3J). However, the effect of the bone targeting Tras-ALN was more prominent and led to a prolonged inhibition of the metastasis burden (Fig. 3F,G). As previously shown (26), we found that Tras-ALN was more enriched in the bone microenvironment, a property that supports its superior inhibitory effect (Fig.S3K).

7. Pre-existing Her2-expressing cells are poorly responsive to Tras-ALN

The strong response of bone metastasis to anti-Her2 therapies (Fig. 3E–H) led us to question whether similar responses could be observed by targeting the pre-existing Her2+ cancer cell populations that are known to associate with plasticity(16). In contrast to MCF7 which classified as Her2 “0” based on the Hercept test, ZR75-1 had a score of “1+” (28). These results were confirmed by confocal imaging and contrasted with the Her2+ breast cancer model MDA-MB-361. However, phospho-S6K expression suggested a negligible difference in Her2 signaling between the two HR+/Her2− models (Fig.S4A). Hence, we exposed ZR75-1 to different treatment conditions (ALN, Tras, Tras-ALN, and PBS/Control) for 72 hours in 3D culture. Intriguingly, no inhibitory effect was observed following anti-Her2 treatment (Fig.S4B,C). Rather, cells treated with Tras-ALN grew slightly faster than the control cells, contrasting with our bone metastasis experiments (Fig.S4C). Together, these results indicate that the bone-induced Her2E cells have different properties compared to pre-existing cells.

8. The inhibitory effect of Tras-ALN on bone metastasis is independent of immune intervention.

As antibody-based therapies can trigger an immune response in patients (29), we exposed immunocompetent mice (C57BL/6) to Tras-ALN or ALN for 3 weeks before immune profiling. No remarkable changes were observed between the two groups, except for CD4+ T cells where the ALN group has a slightly increased proportion (P=0.022) (Fig.S4D). Due to previously demonstrated leakiness events in immune-deficient mice (30), nude mice were exposed to similar treatments. Besides a slight increase of CD4+ T cells in the Tras-ALN group, no significant changes were observed (Fig.S4E). Together, these results suggest a direct inhibitory effect of Tras-ALN on BM progression.

9. Acquired Her2 contributes to metabolic changes in breast cancer bone metastases.

To determine differences between pre-existing and bone-mediated Her2-expressing cells, we performed single-cell RNA-seq on naïve and bone-derived MCF7 cells. Using Harmony (31) to ensure a stringent data integration, we identified multiple genes (top 10) that were significantly upregulated in bone-derived cells (Fig. 4A). Several of these markers were involved in cellular metabolism (e.g., MTND4, MTCO2, MTCO1, MTCYB, NR2F2, MTND5, MTND6, CA12, CAV1 and CRABP2), transcription (e.g., ID1, ID3, RPL13A, and ZMYND8), chemotaxis (e.g., CXCL12), cell adhesion (e.g., WISP2, FREM2, ANXA2), growth and survival (e.g., NRG3). Gene Ontology revealed the predominance of metabolic changes in bone metastasis (Fig.S4F). This was confirmed by seahorse analyses revealing an increased oxygen consumption rate (OCR) in bone-derived cells (Fig.S4G). In parallel, the hallmark analysis within each single-cell cluster identified MYC signature as the most enriched in bone-derived cells (4/8 clusters (C), P= (C0) 5.05e-4; (C1) 1.21e-11; (C2) 1.59e-17; (C4) 5.6e-4)) with cluster 2 (C2) being the most significant (Fig. 4B). Considering the well-established role of MYC in metabolism regulation (32–34), these results argue for a MYC-mediated metabolic perturbation in bone metastasis.

Figure 4: Bone-mediated Her2E cells are transcriptionally different from pre-existing Her2E cells.

A- Heatmap representing the top differentially expressed genes in all cell clusters identified from naïve and bone-derived MCF7 cells based on single-cell RNA-seq. The data was deposited under GSE230612.

B- Bar plot based on MYC-signature significance (P-value) in major cell states (UMAP, Resolution: 0.4) identified from single-cell RNA-seq.

C- Ridge plot showing Her2-signaling scores in major cell states. Clusters are indicated as C0 to C7.

D- Gene expression analysis of bone-associated genes identified from scRNA-seq analysis following Her2 overexpression in MCF7 (GSE111246). Significant P-values are shown as *(<0.05), **(<0.01), and ***(<0.001).

E- Violin plot representing neuregulin 3 (NRG3) expression in different cell states. Differential expression between Naïve (red) and bone-derived (blue) MCF7 cells is significant for most clusters. P-values from cluster 0 to 7 are: P₀<2e-16, P₁<2e-16, P₂<2e-16, P₃=3.9e-14, P₄=4.2e-15, P₅=1.2e-0.7, P₆=0.028, P₇=0.133.

F- Western blot showing the effect of NRG3 on heterodimerization-mediated Her3 phosphorylation (p-Her3) in cancer cells following a 4-hour exposure.

G- Bar plots based on quantitative level of p-Her3 relatively to total Her3, following MCF7 and SCP2 exposure to NRG3.

H- Time course analysis of p-Her3 expression under NRG3 treatment.

I- Experimental design and histograms showing quantitative phospo-Her3 level relatively to total Her3 in naïve (MCF7 and SCP2) and bone-derived MCF7 (Bo-MCF7 and Bo-SCP2) cells.

J,K- Forest plots depicting the risks of breast cancer relapse based on the mRNA expression ratio (j) or the mean expression (k) of indicated paired genes. Hazard ratios are plotted with a 95% confidence interval.

L,M- Experimental design and actin normalized densitometry analyses of Her2 and pS6K expression in naïve and bone-derived models of HR+/Her2− breast cancer.

N,O- Representative PLA images showing Her2/Her3 heterodimerization in CTCs collected from bone metastasis-bearing mice (N), or MCF7 treated with 100ng/ml NRG3 for 24h (O).

P- Dot plot showing the mean expression of neuregulins (NRGs) in breast cancer patient-derived CTCs and metastatic lesions. Each dot represents a sample. Data was analyzed from GSE113890. P-value: two-tailed paired t-test.

Q- Representative Bioluminescence images at day 0 and 49 following intra-cardiac (IC) injection of control MCF7 or NRG3 shRNA (sh16 and sh72) cells.

R- Area under curve analysis of metastasis progression in NRG3 depleted conditions relatively to control mice. Control mice (n=9), sh16 (n=8), sh72 (n=8).

S- Bar plot showing the percentage of mice with multiorgan metastasis (MoM) in control (n=9) and NRG3-depleted conditions (n=16).

Surprisingly, we found that C2 also had the most enriched Her2 signaling score, suggesting a link between Her2 and MYC activity (Fig. 4B,C). As Her2 increases in bone metastasis, we reasoned that Her2 could promote the increased metabolic state in bone-derived cells via MYC. To genetically test this hypothesis, we overexpressed Her2 in MCF7 cells before assessing its impact on bone-induced genes. Her2 overexpression upregulated multiple genes (Fig. 4D). We then performed OCR experiments to ascertain the role of MYC in mediating Her2 activity. While overexpressing Her2 promotes the Maximum Respiration Capacity of MCF7 cells, this was abrogated by MYC inhibitor (10058-F4) and Her2 inhibitor (Tras-ALN) (Fig.S4H). Combining both drugs was not superior to single treatments (Fig.S4H). Altogether, these data are consistent with an association between MYC and Her2 signaling in mediating bone-induced respiration, without ruling out independent regulatory mechanisms on oxygen consumption.

10. Bone-induced NRG3 promotes the heterodimerization of human epithelial growth factor receptors.

One of the most induced genes in bone-derived cancer cells was neuregulin 3 (NRG3) (Fig. 4E). Single-cell transcriptomics analysis revealed that while NRG3 expression was upregulated in bone, other neuregulins (NRG1, NRG2, and NRG4) remained unaltered (Fig.S4I). As a surrogate ligand for Her4, NRG3 does not directly bind to other Her2 family members (35), but can activate both Her2 and Her3 when these receptors are co-expressed with Her4, suggesting a prominent role in promoting Her4 heterodimerization (36). While Her3 has a functional ligand binding domain, it relies on heterodimerization to be activated due to the lack of a catalytic kinase domain (37,38). Therefore, we assessed the level of phospho-Her3 (p-Her3) as a marker of receptor heterodimerization in naïve and bone-derived cancer cells. Exposing breast cancer cells to NRG3 recombinant induced p-Her3 (Fig. 4F,G). The specificity of NRG3 in promoting p-Her3 was further validated by a time course experiment (Fig. 4H). Interestingly, p-Her3 was elevated in bone-derived cells (Fig. 4I), arguing for increased heterodimerization in bone metastasis. To determine the clinical relevance of heterodimerization, we curated clinical datasets for HR+ breast cancer patients (N=902) who have been exposed to endocrine treatment (39). As shown by the forest plot, gene expression ratios of ERBB2, and ERBB3 over ERBB4 (P=0.0003 and P=0.02, respectively), but not mean expression of gene pairs, were associated with increased risk of breast cancer relapse (RFS) (Fig. 4J,K). Hence, the relative proportion of Her2 family receptors may dictate survival outcomes in ER+/Her2− breast cancer.

11. NRG3 sustains mTOR signaling via autocrine activation of human epithelial growth factor receptors.

Since several studies have identified stromal cells as the primary source of NRGs (40–42), we reasoned that the acquired NRG3 expression may enhance cell-autonomous properties in bone-derived cells by sustaining Her2 signaling during metastasis. Of note, Her2 is an orphan receptor that relies on heterodimerization to activate downstream signaling. To test this, we cultured cancer cells in serum-free media for 48h to ensure NRG3-free conditions before assessing the level of phosphorylated (Thr389) Ribosomal protein S6 kinase beta-1 (p-S6K) as a readout for Her2-mediated AKT/mTOR signaling. Surprisingly, Her2 expression and its associated mTOR activation marker, p-S6K, were both sustained in bone-derived cancer cells while remaining low in naïve MCF7 and M7-SCP2 cells (Fig. 4L,M). We found that a 30-minute exposure to NRG3 was sufficient to induce p-S6K level (Fig.S4J). Similarly, overexpressing Her2 increases p-S6K level in the same cells (Fig.S4K).

To determine whether Her2 heterodimerization was also relevant in CTCs, we performed a PLA experiment to visualize this process. Similarly to 3D cultured cells, we found that Her2 directly heterodimerizes with Her3 in CTCs collected from both MCF7 and ZR75-1 bone metastasis-bearing mice (Fig. 4N and Fig.S4L). Transcriptomic analysis of CTCs from MBC patients revealed a positive association between ERBB2 and ERBB3 in CTCs derived from BM but not VM patients (P=0.0322), further supporting our results. (Fig.S4M) (24). We then assessed the impact of NRG3 on the heterodimerization process using the same PLA approach. Interestingly, exposing cancer cells to NRG3 remarkably increases Her2/Her3 heterodimerization (Fig. 4O). The involvement of NRG3 was supported by clinical evidence showing an increased expression of NRGs in CTCs compared to metastatic lesions (GSE113890) (Fig. 4P).

Still, the role of intrinsic NRG3 remained questionable. Hence, we used shRNA approaches to inhibit NRG3 expression in MCF7 cells before assessing their seeding potential via IC injection (Fig. 4Q and Fig.S4N). While p-Her3 was reduced following NRG3 depletion (Fig.S4O), both the overall metastasis and multiorgan metastasis potential of NRG3-depleted cells (sh16 and sh72) were significantly inhibited (Fig. 4R,S and Fig.S4O). Together, these results highlight the involvement of NRG3-associated autocrine circuitries that may contribute to bone-mediated secondary metastasis formation by maintaining a stronger dependency of bone metastasis on Her2 signaling for survival.

12. Pre-existing Her2E cells are reversible and not required for bone-mediated secondary metastasis formation.

Since Her2-expressing cells were detected in the naïve cancer model, we sought to investigate their role in bone metastasis formation. Cells were sorted by flow cytometry, based on Her2 expression, and classified as Her2-Low, Medium (Med), and High (Fig. 5A). While ~2% of Her2-Low and 37% of Her2-Med cells could initiate colony formation, fewer Her2-High cells managed to form colonies (<0.01) (Fig. 5A). Hence, we expanded Her2-Low and Her2-Med MCF7 sublines and re-assessed Her2 level by flow analysis. Surprisingly, most Her2-Med cells become Her2-, similarly to Her2-Low cells (Fig. 5B). In fact, the Her2 distribution was highly similar in both sublines with no statistical differences in various classes of Her2 (Low P= 0.1306; Med P= 0.0983; High P=0.2424) (Pie plots Fig. 5B). Of note this contrasts with Her2+ CTCs collected from ER+/Her2− patients where the majority (~90%) of the expanded cells remained Her2+ (43). Additionally, half of Her2− CTCs collected from the same patients could expand into Her2+ CTCs in vitro, contrasting with the Her2-Low cells that predominantly maintain a Her2− status. As both Her2-Med and Her2-Low skew towards a Her2− status, similarly to parental cells, these results demonstrate that pre-existing Her2E cells behave differently from Her2E CTCs.

Figure 5: Her2E cells are not required for BM initiation from naïve cells.

A- Fluorescence-activated cell sorting analysis of MCF7 cells based on Her2 expression (Low, Medium, and High). N= number of events (cell count) for each group. Cells were seeded in 5 cm dishes for expansion. The percentage of detectable colonies over the seeded cells is shown in parathesis.

B- Flow analysis of expended MCF7 Low and Medium cells based on Her2 expression. The pie plots represent the proportion of Her2 Low, Medium, and High from Her2-sorted sublines (N: 4 biological samples from 2 independent experiments). P-values (two-tailed Student’s t-test) comparing Her2 distribution in MCF7-Low vs MCF7-Med: P=0.1306 (Low vs Low), P=0.0983 (Med vs Med), P= 0.2424 (High vs High).

C- Cell proliferation analysis of Her2-Low and Her2-Med MCF7 cells based on BLI in 2D.

d- Representative BLI images of bone metastasis-bearing mice post-IIA injection.

E- Assessment of multiorgan metastasis (MoM) in control MCF7 (Parental cells), Her2-Low, and Her2-Med 8–10 weeks post IIA injection; (n=5 mice per group). P-value: Fisher’s exact test.

F- Flow analysis of Her2E cells in naïve and bone marrow (BMa)-derived MCF7 at day 3 post-IIA injection.

G- Ridge plot showing Her2 expression in naïve and MCF7-derived DTCs at day 3 post-IIA.

H-I: Percentage of Her2E cells in DTCs collected from bone marrow (BMa) or bone matrix at day 3 and 14 post IIA-mediated bone metastasis initiation; (n=3 mice per group). P-value: Two-tailed Student’s t-test.

J-M: Flow analysis of Her2E cells in naïve and BMa-derived ZR75-1 cells (J). Ridge plot showing Her2 expression in naïve and ZR75-1-derived DTCs at day 3 post-IIA (K). Percentage of Her2E cells in DTCs derived from BMa and bone matrix at day 3 and 14 post-IIA injection (I-M); (n=3 mice per group). P-value: Two-tailed Student’s t-test.

To test whether pre-existing Her2E cells were required for bone metastasis initiation, Her2-Low, and Her2-Med sublines were injected via IIA, along with the parental cells. While no proliferation difference was observed in vitro, Her2-Med appeared to grow faster in bone relatively to Parental and Med-Low MCF7 cells (Fig. 5C,D and Fig.S5A). This corroborates previous observations from patients-derived CTCs (43). Surprisingly, however, the growth advantage was not associated with a superior degree of multiorgan metastasis with both sublines showing a 40% multiorgan metastasis rate (Fig. 5E and Fig.S5B,C). Moreover, the initial Her2 level did not impact the bone metastasis initiation potential of these cells (Fig.S5B,C). Considering the potential of bone to induce Her2 expression, these results suggest that the pre-existing Her2E cells are not required for bone colonization.

Next, we sought to determine the temporal impact of bone on Her2 expression. MCF7 cells were transplanted to bone via IIA injection before DTC collection at various time points. We observed an increased expression of Her2 (Fig. 5F,G) and the proportion of Her2E cells was strikingly elevated in bone marrow (BMa) and bone matrix (Fig. 5H,I) only 3 days post-IIA. We found similar results for the ZR75-1 model (Fig. 5J–M). To further assess the involvement of osteogenic cells in this process, we adopted a 3D coculture approach. MCF7 cocultured with MSC displayed a lower but significant increase in the proportion of Her2E cells (Fig.S5D). Using a trans-well assay, we found that a direct interaction was not required for Her2 induction when cells were grown in 3D (Fig.S5E,F). While Her3 was also increased, it remained unsignificant (Fig.S5G).

Our previous studies identified the N-cadherin/E-cadherin heterotypic junction as a promoter of mTOR activation in the early stage of bone metastasis formation. As Her2 is increased in bone metastasis, we decided to determine whether its expression is required for mTOR activation in the osteogenic niche when heterotypic junctions are established between cancer cells and osteoblasts. We first generated Her2 shRNA models of MCF7 (Fig.S5H). Then, control and Her2 shRNA models of MCF7 were co-cultured with osteogenic cells in 3D for 4 hours to mimic the osteogenic niche, before assessing pS6K level (Fig. S5I,J). Intriguingly, while pS6K was undetectable in osteogenic cells, the reduced Her2 level in shRNA models did not prevent mTOR induction in the same coculture conditions, suggesting that Her2 is dispensable for mTOR activation when the heterotypic junction is established. However, when the heterotypic junction is lost, our results indicate that the Her2 upregulation and maintenance in bone-derived cells confers alternative mechanisms.

13. Her2 promotes multiorgan metastasis in HR+/Her2− breast cancer models

As the bone-induced Her2E was observed in CTCs, we decided to evaluate whether DTCs in visceral organs of IIA-induced bone metastasis mouse models could reflect similar phenotypes (Fig.S5k). Based on immunofluorescence, we identified several Her2E DTCs in the lung of bone metastasis-bearing mice (Fig.S5L). We also identified NRG3+ microlesions (Fig.S5M). Hence, to determine whether Her2 is essential for secondary metastasis seeding, we genetically engineered MCF7 and ZR75-1 models with gain or loss of Her2 expression (Fig.S6.1A–H). As Her2 genetic inhibition reduces bone metastasis formation (Fig.S6.1I), we directly injected MCF7 models via IC to determine the essentiality of Her2 on metastasis seeding and progression. Ectopic expression of Her2 displayed a minimal impact on metastasis formation. In contrast, various Her2 shRNAs inhibited metastasis growth in vivo (Fig.S6.2A–C). The ex vivo profiling of metastatic lesions revealed a remarkable impediment to visceral metastasis (Fig. 6A,B) and multiorgan metastasis (Fig. 6C). More specifically, lung and ovary metastases were significantly reduced in Her2 shRNA conditions (Fig. 6D). Genetic modulations of Her2 did not significantly impact bone metastasis formation despite a decreased tendency in both overexpressing (Her2-OE) and depleted models (Fig.S6.2D). We found similar results in ZR75-1 despite a lesser bone metastasis potential following IC injection (Fig.S6.2E,F). Compared to Control and Her2-OE cells, both multiorgan metastasis and metastasis to visceral organs were also inhibited (Fig. 6E and Fig.S6.2G,H). Organ-specific analyses revealed significant changes in kidney (including the adrenal cortex) and lung metastasis frequencies of the ZR75-1 model (Fig. 6F).

Figure 6: Her2 contributes to visceral metastasis seeding.

A- Ex vivo bioluminescence imaging of visceral organs following IC injection of MCF7 cells with Her2 overexpression (Her2-OE), or Her2 depletion (sh53, sh78, ShA, and ShB). Organs were harvested 3 to 4 months post-injection.

B- Bar plot depicting visceral metastasis distribution in control, Her2-OE, and Her2-targeting shRNA groups, based on ex vivo imaging. Control (n=8), Her2-OE (n-5), sh53 and sh78 (n=4 and n=3, respectively), shA (n=5), and shB (n=4).

C- Proportion of multiorgan metastasis in mice injected with control MCF7 (n=8), Her2-OE (n=5), and shRNAs (n=16). P-value: Fisher’s exact test.

D- Impact of Her2 genetic modulation on lung and ovary metastasis from MCF7.

E- Proportion of multiorgan metastasis in mice injected with control ZR75-1 (n=5), Her2-OE (n=5), and shRNAs (n=12). P-value: Fisher’s exact test.

F- Impact of Her2 genetic modulation on kidney (including the adrenal cortex) and lung metastases from ZR75-1.

Surprisingly, while liver metastasis was a rare event in both MCF7 and ZR75-1 models, we found a significant difference between Her2-OE and shRNA models of MCF7 (P=0.047), while similar trends were observed in other organs (Fig.S6.2I–K). Hence, while a high level of Her2 may be dispensable for bone metastasis formation, it may enhance metastasis take rate in certain visceral organs. Together, these results demonstrate that Her2 is essential for visceral metastasis seeding and multiorgan metastasis formation.

14. Bone dysregulates miRNAs involved in Her2 post-transcriptional regulation.

Based on ATAC-seq analysis, we found evidence that Her2 expression was not driven by increased chromatin accessibility in bone-derived cells (Fig.S7.1A). However, previous studies have identified several miRNAs that can influence Her2 signaling in breast (44,45), prostate (46), and colon cancers(47), suggesting that miRNAs could impact Her2 expression in bone metastasis. Hence, we extended the list of miRNAs with Her2-targeting potential using in silico approaches (48) (Fig.S7.1B) and performed a functional enrichment analysis in MCF7 cells exposed to osteogenic cells (MSCs) (GSE137270). Intriguingly, 5/8 predicted miRNAs showed decreased signature in MCF7 cells exposed to MSCs (Fig. 7A). More specifically, MIR125A/B (NES:−5.62; q-val:0), LET7C (NES:−5.56; q-val:0), MIR133A/B (NES: −4.68; q-val:0), MIR331 (NES:−4.23, q-val:0), MIR18A (NES:−3.87; q-val:0), and MIR99A/B (NES:−1.94; q-val:0.007) signatures were all reduced in MSC conditions (Fig. 7A). These results indicate that osteogenic cells directly inhibit Her2-targeting miRNAs in 3D coculture. As Her2 expression is increased in the absence of heterotypic junction, we profiled several miRNAs in trans-well coculture conditions. The results revealed that direct cell-cell interaction was not required for MIR133B, MIR125B and LET7C inhibition in coculture (Fig.S7.1C), hence suggesting the implication of miRNAs in osteogenic cell-mediated Her2 expression.

Figure 7: Involvement of miRNAs in bone-mediated Her2 expression.

A- Gene set enrichment analysis (GSEA) of mir-133a/b, miR-125b, and Let-7c pathway enrichment in naïve (MSC-) and MSC cocultured MCF7 cells (MSC+). Normalized enrichment score (NES) is significant if the FDR q-value <0.05. Only significant NES (FDR, q<0.02) are presented.

B- Bar plot summarizing the association of bone-inhibited Her2-targeting miRNAs with overall survival (OS) in breast cancer (METABRIC). A total of 726 HR+ and endocrine-treated breast cancer patients were analyzed. Significant P-value (Log-Rank <0.05).

C- Experimental design indicating the origin of metastatic and CTC samples from breast cancer patients with distant metastases (GSE113890). 8/14 specimens derive from BM+ MBC patients.

D- Bar plots showing the distribution of positive (EXP+) and negative (EXP-) specimens based on miRNA detection. Specific miRNAs are annotated on each graph. (N=14); Significant P-value <0.05; two-sided Fisher’s exact test.

E- Heatmap showing miRNA expression in prostate primary lesions (PCa) and metastasis to bone (BM) (GSE230278).

F- Bar plots showing the distribution of positive (EXP+) and negative (EXP-) specimens based on miRNA detection in primary prostate cancer (PCa) and bone metastasis (BM).

G- Dot plots showing miRNA expression in primary prostate cancer (PCa) and bone metastasis (BM). Significant P-value <0.05; two-tailed Mann-Whitney U-test.

H- Representative ex vivo bioluminescence imaging (BLI) of brain metastases following IC injection of control (negative mimics) or miR-133a/b mimic-treated bone-derived ZR75-1 (Bo-ZR75-1) models. Representative whole mouse images at day 0 post-IC are shown.

I- Percentage of brain metastasis in control and miR-133a/b mimic-treated groups (n=4 mice per group).

J, K: Ex vivo BLI of lung metastases post-IC injection of control (negative mimics) or miR-133a/b mimic-treated bone-derived MCF7 (Bo-MCF7) models. Bar plot showing the impact of miR-133a/b on lung metastasis formation (K); (n=5 mice per group).

L, M: Similar to panel “J and K’ using bone-derived SCP2 (Bo-SCP2); (n=5 mice per group). For I, K, and M: P-value = ‘N-1’ Chi-squared test.

N- Summary graph depicting key mechanistic insights associated with acquired Her2 expression in bone and contributing to secondary metastasis. While cell-cell interaction may contribute to mTOR activation in the osteogenic niche (51), we demonstrate how the bone-mediated NRG3/Her2/Her3 axis facilitates CTC seeding in visceral organs. Pharmacological blockade of Her2 using bone-targeting Trastuzumab (Tras-ALN) impedes Her2E CTC dissemination from bone to visceral organs.

To determine whether the bone-inhibited miRNAs are relevant for breast cancer progression, we assessed their association with overall survival using a cohort of 726 HR+ and endocrine-treated breast cancer patients (METABRIC)(49,50). Interestingly, high levels of MIR99A, LET7C, MIR125B, and MIR133A/B were associated with better OS; MIR133A/B was the most significant (P<0.0001) (Fig. 7B). To evaluate whether the miRNA alteration was long-lasting, we assessed the level of MIR133B in bone-derived breast cancer cells that were maintained in culture for several weeks. Interestingly, MIR133B was decreased in bone-derived MCF7 and M7-SCP2 cells (Fig.S7.1D). Similar changes were found in the Her2+ breast cancer models MDA-MB-361 (Fig.S7.1D).

Hence, we assessed the impact of these miRNAs on Her2 level in our models. Treating bone-derived cancer cells (Bo-SCP2 and Bo-MCF7) with a library of miRNA mimics (miR133a, miR133b, miR125b, and Let7c) reduced Her2 expression in bone-derived models (Fig.S7.1E,F). Similar results were observed by flow analysis on bone-derived ZR75-1 cells (Fig.S7.1G). Hence, we assessed the impact of miRNA mimics and inhibitors on Her2 level following a 72-hour-treatment. While miRNA mimics could reduce Her2 level in bone-derived cells, they did not affect Her2 level in naïve cells (Fig.S7.1H). In contrast, the miRNA inhibitors rescued Her2 expression in naïve but not in bone-derived ZR75-1 cells (Fig.S7.1H). We found that miR133a/b specific-mimics could more efficiently inhibit Her2 in bone-derived cells (Fig.S7.1I). Together, these results suggest that the bone microenvironment can impact miRNAs with Her2 regulatory potential. However, their involvement in secondary metastasis was still unclear.

15. Loss of Her2-targeting miRNAs contributes to Her2 maintenance in CTCs.

We speculated that the bone-imprinted miRNA alterations should also translate to CTCs, considering the increased level of Her2 in patients with bone metastasis. Hence, we assessed miRNA expression in CTCs relatively to metastasis in breast cancer patients using published datasets (GSE113890) (Fig. 7C). While most selected miRNAs were lowly expressed in the metastatic setting (Fig.S7.1J,K), we found that the frequencies of MIR133B, MIR331, MIR99A and MIR125B were reduced in CTCs relatively to metastatic lesions (Fig. 7D). Other miRNAs were not significantly altered in CTCs (Fig.S7.1L,M). Hence, we have demonstrated that while multiple miRNAs can target Her2 expression and signaling, some bone-inhibited miRNAs remain low in CTCs allowing a sustained Her2 level. While MIR133B, MIR125B, and LET7C expression decrease in metastasis compared to primary tumors, our results suggest further loss of expression in CTCs from stage 4 MBC patients.

16. miR-133 associates with bone metastasis in prostate cancer.

To determine whether the identified miRNAs were relevant to other cancer types that frequently metastasize to bone, we analyzed a prostate cancer (PCa) dataset (GSE230278) with primary (benign and local) and bone metastasis (BM) samples (Fig. 7E). Consistent with breast cancer, PCa analysis revealed a remarkable decrease of MIR133A/B in both frequency and expression in BM compared to primary tumors (Fig. 7F,G). No major changes in detection frequency were found for other miRNAs (Fig.S7.2A). Intriguingly, however, gene expression analysis revealed a reduced expression MIR125A/B in BM samples (Fig.S7.2B,C), suggesting that it may influence Her2 signaling in prostate cancer. Taken together, these results indicate that the bone microenvironment can modulate MIR133- and MIR125-related miRNAs which can interfere with Her2 signaling in prostate and breast cancers.

17. miR-133 hinders visceral metastasis in breast cancer

Since MIR133A/B decreases in bone metastasis and CTCs of patients with multiorgan metastasis, we reason that miR-133a/b mimics could impede metastasis progression. To mimic the seeding process of cancer cells from bone to other organs, we transfected bone-derived models (Bo-MCF7, Bo-ZR75-1, and Bo-SCP2) with miR-133a/b or non-targeting mimics (Control) for 72h before IC injection. Following metastasis formation, visceral organs were evaluated ex vivo. The results were intriguing. While brain metastasis from naïve cancer cells was uncommon, we found that 75% of Bo-ZR75-1-bearing mice had developed brain metastasis in the control group. Similarly, all mice injected with control Bo-MCF7 or Bo-SCP2 had lung metastasis, a phenotype not observed in naïve MCF7 and SCP2 cells. Yet, we found a drastic decrease in visceral metastasis when we analyzed mice transplanted with miR-133a/b-treated cells (Fig. 7H–M). More specifically, no brain metastasis was observed in miR-133a/b mimic-treated Bo-ZR75-1 (Fig. 7H,I and Fig.S7–2A) while only 20 % of mice displayed lung metastasis in miR-133a/b-treated Bo-MCF7 and Bo-SCP2 (Fig. 7J–M). As Her2+ breast cancers are more brain tropic, we evaluated Her2 level in brain micrometastasis derived from Bo-ZR75-1. While brain micrometastases maintained Her2 expression, NRG3 was strongly enriched in the extracellular matrix surrounding the lesion (Fig.S7.2D,E). As no enrichment was observed in tumor-free brains (Fig.S7.2F), these findings further support the importance of the Her2/NRG3 axis in visceral metastasis formation. Overall, these results demonstrate that re-establishing Her2-targeting miRNAs in bone metastasis can reduce secondary metastasis seeding to visceral organs.

Discussion

Cancer cell cross-seeding exacerbates metastasis progression. In this study, we provide evidence that the bone microenvironment promotes Her2+ CTCs in MBC patients with initial HR+/Her2− status and identify a new tumor-promoting circuitry involving NRG3 autocrine signaling that sustains mTOR signaling beyond the osteogenic niche. We further demonstrated that targeting bone-mediated Her2E metastasis impedes secondary dissemination and provides long-term survival benefits (Fig. 7N).

Until now, the diagnosis of bone metastasis relies on advanced imaging approaches that may not capture early metastasis. Late diagnosis is common, often permitted by unexpected bone pain or fracture. Tools to identify patients with bone metastasis will guide therapeutic strategies before morbidity occurs. The role of the bone microenvironment as a primary site of recurrence compels for better approaches to detect early bone metastasis. Multiple studies identified Her2 expression discrepancies between primary tumors and bone metastases. The bone-mediated Her2 activation in the luminal setting provides a unique opportunity to evaluate the secondary impact of these cells on multi-organ metastasis and to predict bone metastasis. We have demonstrated that DNA-based approaches (e.g., ctDNA focusing on ERBB2 mutations) would not be a good strategy to predict bone metastasis (Fig. 1A,B and Fig.S1D). To identify non-invasive markers for the diagnosis of bone metastasis, we collected and analyzed thousands of CTCs from HR+/Her2− MBC patients. Our analysis revealed that Her2 is generally highly expressed in CTCs from MBC patients with BM and could serve as a relevant biomarker for patients with initial Her2− status.

Acquired Her2 expression is a benchmark of endocrine resistance in breast cancer, and the association of Her2 with stemness, survival, and proliferation has been established (16,17). Patients with Her2 amplification commonly develop metastasis in visceral organs (e.g., brain), contrasting with HR+ tumors that predominantly recur from bone. Therefore, our results suggest that the bone-mediated Her2 induction in patients with initial HR+/Her2− status can change the tropism of CTCs toward non-skeletal sites, leading to metastasis expansion.

In previous studies, we demonstrated that the formation of heterotypic adherent junctions between N-cadherins and E-cadherins in the osteogenic niche promotes bone metastasis via mTOR activation (51). While this heterotypic junction-mediated mTOR activation process was independent of the initial Her2 level of cancer cells (Fig.S5I–K), we highlight the essentiality of Her2 for secondary metastasis progression from bone. Our results suggest that in advanced stages of bone metastasis, when cancer cells lose their interaction with osteogenic cells and disseminate to other organs, the acquired Her2 expression drives a compensatory mechanism to ensure CTC survival and metastasis expansion in visceral organs. This is corroborated by the high sensitivity of CTCs to AKT and mTOR inhibitors in colorectal cancer (52).

The potential of cancer cells to carry and maintain therapeutic resistance traits (e.g., Her2, NRG3) exemplifies the possibility for cross-resistance between multiple organs. The remarkable expression of NRG3 in bone-derived cells provides further evidence that CTCs may develop autocrine mechanisms to maintain growth factor receptor signaling, independently from the tumor microenvironment. We reveal that NRG3 promotes Her2 heterodimerization with other Her2 family members that serve as surrogate receptors (52). NRGs most commonly bind Her3 and Her4, both of which preferentially dimerize with Her2. However, relatively to ERBB4, ERBB3 is transcriptionally higher in CTCs and fully depends on heterodimerization to be functional. Our results also indicate a decreased level of ERBB4 which is often associated with luminal A breast cancers and better prognosis(53). Therefore, the ratio between ERBBs may be more determinant for clinical responses in HR+/Her2− breast cancers.

We identified miR-125b and miR-133a/b as potential regulators of Her2 expression in bone. Our results were supported by previous findings demonstrating the regulatory role of miR-125b and Let-7c in endocrine resistance and Her2 expression in luminal breast cancer (44). Hence, we show that miRNAs, by post-transcriptionally regulating Her2 in bone metastasis, can modulate metastasis progression through activation of mTOR signaling. We further demonstrate that the Her2-targeting miRNAs could impact the visceral metastasis potential of bone metastasis models.

While Tras-ALN can slow down treatment-naïve bone metastasis growth, our study suggests a better outcome in bone metastasis pre-exposed to endocrine therapy. This process increases the resistance phenotype and forces cancer cells to rely more on alternative pathways for survival. In these conditions, anti-Her2 treatment abrogates bone metastasis and multiorgan metastasis. These results are in line with the DESTINY-Breast04 trial where pretreated HR+ MBC patients with low Her2 tumors revealed improved overall survival compared to physician’s choice chemotherapy regiments (15). As such, we provide evidence that Tras-ALN can efficiently impede endocrine-resistant bone metastases and eliminate Her2E CTCs with tumor-initiating properties. While using Trastuzumab alone may not be sufficient to eliminate all Her2 low lesions in bone, future studies may target novel therapeutics such as T-DM1 and Enhertu to the bone to increase therapeutic benefits.

In summary, our results provide insight into mechanisms of resistance that can be transmitted from one organ to another without genetic implications. The activation of autocrine signaling increases possibilities for microenvironment-independent survival, a process that facilitates metastasis seeding and tissue colonization. We have identified and characterized Her2 as a traceable marker that associates with bone-mediated secondary metastasis. Together, these findings may help improve diagnosis and guide therapeutic strategies in HR+/Her2− MBC patients.

Methods

Breast Cancer Patients Cohort

We analyzed 137 patients with HR-positive Her2-negative MBC who had CTC collection before the initiation of the next treatment line. Patients were enrolled between 2013 and 2019 under the Investigator Initiated Trial NU16B06, at the Robert H. Lurie Comprehensive Cancer Center at Northwestern University (Chicago, IL). The median age at diagnosis of MBC was 56 [Interquartile range (IQR): 49–63] years. De-novo MBC was diagnosed in 21% (n=29) of patients. Bone metastasis was detected in 77% (n=105) of patients, while visceral metastasis only was detected in 23% (n=32) of patients. The study protocol was approved by the Thomas Jefferson University and the Robert H. Lurie Comprehensive Cancer Center at Northwestern University Institutional Review Boards. A written informed consent was obtained from each participant.

Human Circulating Tumor Cell (CTC) Staining and Analysis

CTCs were collected from 7.5ml of peripheral blood and Her2 immunofluorescence labeling was carried out using the CellSearch^® system (Menarini Silicon Biosystems) coupled with the CellSearch^® Circulating Tumor Cell kit and Tumor Phenotyping Reagent Her2/neu (Menarini Silicon Biosystems), as previously reported (54). Imaging output from the Celltracks Analyzer II^® system was then uploaded into the open-source program Automated CTC Classification Enumeration and PhenoTyping (ACCEPT) (55). ACCEPT helps classify Her2 expression in CTCs by mean intensity immunofluorescence as follows: negative (Her2 mean intensity of 0) and positive (Her2 mean intensity of >0). The Her2 positive CTCs were further subdivided into 1 to 2+ (mean intensity of >0 and ≤100) and 3+ (mean intensity of >100). Clinical and pathological variables were reported using descriptive analysis. Bone metastases were detected by imaging. CTC count is described as cells per 7.5mL of peripheral blood. The association between frequencies of Her2-positive CTCs per patient and the presence of bone metastasis was tested using the Mann-Whitney U test. All analyses were performed using IBM SPSS Statistics (Version 29.0). CTC counting and Her2 immunophenotyping characteristics for the study cohort included patients with 0 to 37 CTCs per 7.5 mL of peripheral blood.

Mouse CTC collection, Processing, and Analysis

Mouse experiments were approved by the Institutional Animal Care & Use Committee of Baylor College of Medicine and Mount Sinai. Mice were euthanized using isoflurane followed by cervical dislocation. Using a 31G syringe, 750 μl of blood was collected from the right ventricle into EDTA-coated vials on ice. This was followed by two rounds of RBC lysis and 3 washes with PBS. The cell pellet was resuspended in 300 μl of cold PBS and 100ul was placed on Poly-L-Lysine (Electron Microscopy Science 19320-A) coated slides for 30 to 60 min at room temperature to allow cell attachment to the slide. The excess liquid was then aspirated and 200 μl of 4% PFA was used to fix cells at room temperature for 10 min. The slides were then washed twice with PBS before storage at 4° Celsius or used for downstream analysis. CTCs were blocked and permeabilized with 0.4% Triton X in 10% normal serum for 30 min at room temperature and incubated overnight at 4° Celsius using antibodies against Her2 (RRID:AB_331015; 1:200), Keratin 8 (RRID:AB_2891089; 1:500), Keratin 19 (RRID:AB_2133570; 1:500) or pan Keratin (RRID:AB_10983023; 1:1000). After 3 rounds of washing, secondary antibody (1:1000) incubation was performed a room temperature for 1 hour. Imaging and analysis were performed on a Rarecyte scanner using the following channels: DAPI/Hoechst, FITC/AF488, PE/AF594, and APC/AF647. The CD45 antibody was included for optimization purposes. The report generated from Rarecyte was manually screened for true CTCs before validation. Selected CTCs were re-imaged with a 40X objective to generate representative images.

Cell Lines and Cell Culture

The human breast cancer cell lines MCF7 (RRID:CVCL_0031, Cat# HTB-22), MDA-MB-231(RRID:CVCL_0062), and MDA-MB-361(RRID:CVCL_0620, Cat# HTB-27) were obtained from ATCC. The subline M7-SCP2 (SCP2) was generated in the Zhang Lab by clonal expansion from parental MCF7 cells. ZR75-1 (RRID:CVCL_0588), was kindly provided by Dr. Rachel Schiff. MCF7, SCP2, and MDA-MB-361 were maintained in High Glucose DMEM (Gibco) supplemented with 10% FBS and 1X Penicillin-Streptomycin (PS) (Thermo, Cat# 15140122) while MDA-MB-231, and ZR75-1 were maintained in RPMI media containing 10% FBS and 1XPS. The bone-derived models of Bo-MCF7, Bo-SCP2, and Bo-MDA-MB-361 models were generated from bone metastasis lesions following Intra-iliac artery injection of naïve cells to mice hind limb. The human Mesenchymal Stem Cell (hMSCs) (RRID:CVCL_1D56; Cat# PCS-500-012) and FOB1.19 (RRID:CVCL_3708, Cat# CRL-3602) were obtained directly from ATCC and maintained in phenol red-free DMEM/F12 with 10% FBS, 1X PS. The hMSC cell cultures were supplemented with 1% B27 (37°C, 5% CO2) and hFOB1.19 with 0.3 mg/ml G418 (34°C, 5% CO2). Cells were routinely submitted to mycoplasma testing using the ATCC universal Mycoplasma Detection kit (ATCC; Cat# 30-1012K).

Protein Ligation Assay (PLA)

This experiment was performed using the Duolink in situ Red Starter kit Mouse/rabbit (Sigma-Aldrich; DUO92101-1KT). Following CTC collection on slides as described above, cells were permeabilized with 4% Triton X in PBS for 30 min, washed twice with 200ul PBS (5 min each), and blocked for 1h at room temperature with 2% Triton X-PBS solution containing 10% serum. Primary antibodies against Her2 (Invitrogen; #MA5-13675) and Rabbit Her3 (Cell signaling; 12708S) were diluted in a 1/500 concentration in the Duolink Antibody diluent and applied to each slide following blocking and buffer removal. The remaining steps were described in the manufacturer protocol. Briefly, incubation was performed at 4°C overnight. Next, slides were washed in 1x Wash Buffer A (WBA) at room temperature (2×5min) before Duolink PLA probes were diluted and applied to each sample. Incubation was performed in a pre-heated humidity chamber for 1 h at 37°C. Then slides were washed twice with WBA at room temperature before ligation in a pre-heated humidity chamber (30min at 37°C). Following a washing step with WBA (2X5min), samples were amplified using a polymerase buffer (100min at 37°C). A final wash using 1x Buffer B (2×5min) and 0.01X Buffer B (1×1min) was performed before mounting and fluorescence imaging. Images were acquired using the BZ-X fluorescence microscope (Keyence). The same protocol was used for NRG3-treated cells using 96-well optical plates (Thermo Fisher; 152028).

Delivery of miRNA mimics and inhibitors

mirVana miRNA mimics and inhibitors highly specific to miR-133a (MC10413 and MH10413), miR-133b (MC10029 and MH10029), miR-125b-3p (MC12582 and MH12582), miR-125-5p (MC10148 and MH10148), and negative control (#4464058) were obtained from Ambion while mimics and inhibitors specific to let7c (MIM0006 and INH0006) were purchased from Active motif. Cell transfection was performed using 10nM of miR-mimics and 50nM of miR-inhibitors in Lipofectamine RNAiMax (Invitrogen; #3778030) for 72h before western blot or Flow Analysis.

Her2 FISH and IHC staining

For Her2 FISH experiments, 1×10⁶ naïve and bone-derived breast cancer cells were collected and processed into TMA. Her2 gene amplification was assessed from FFPE slides using the Her2 IQFISH pharmDx (Dako, K573111-5). For IHC, tissues were harvested, fixed overnight with 4% PFA, and embedded in paraffin before sectioning. In most cases, antigen retrieval was performed through heat-induction in Sodium citrate 10mM, pH6 or Tris/EDTA pH9.0. ErbB2/Her2 (RRID:AB_10980124, 1:100) was used for Her2 staining.

Plasmids and lentivirus preparation

The following Her2-targeting shRNA plasmids (shA and ShB) were obtained from Origene (TL320342). Additional Her2-targeting shRNA plasmids with CMV-tGFP promoters and puromycin resistance were used: sh53 (Sigma Aldrich; #TRC000332953) and sh78 (Sigma Aldrich; #TRC000039878). Similar constructs were used for NRG3 targeting shRNAs: sh72 (Sigma Aldrich; TRCN0000150772) and sh16 (Sigma Aldrich; TRCN0000155516). Lentiviruses were generated from HEK293-T cells using 10ug of plasmid, 5ug of pMD2.G, and 5ug of psPAX2. Infected breast cancer cells were selected with Puromycin for 10–14 days until all cells became GFP+. For Her2 overexpression, the pHAGE-ERBB2 plasmid was used(56), and cells were selected with puromycin.

In vivo drug treatments

Bone metastasis was generated using intra-iliac artery injection. We transplanted 5×10⁵ breast cancer cells to 5–6 week-old female mice (RRID:IMSR_JAX:002019). To facilitate tumor establishment, estrogen (8 μg/ml) was orally supplied via drinking water for 2 weeks. Doses of fulvestrant (250mg/kg, 1/week) were subcutaneously delivered for 2 weeks as a neoadjuvant, before mouse randomization into 4 treatment groups: Alendronate/ALN (10 μg/kg, retro-orbital), Trastuzumab/Tras (1mg/kg, retro-orbital), Trastuzumab-Alendronate/Tras-ALN (1mg/kg, retro-orbital), and sterile PBS (100 μl, retro-orbital). Tumor burden was acquired on a weekly basis via luminescence imaging (IVIS Lumina II).

Seahorse Analysis

For Maximum Respiration capacity, 1–2.5×10⁴ cells were seeded in XF96 plates and treated for 24 hours with vehicle (DMSO), 1ug/ml Her2 inhibitor (Tras-ALN), or 20uM Myc inhibitor (10058-F4) before OCR experiment using the XFe96 Extracellular Flux Analyzer and the XF Cell Mito Stress Test kit (Agilent; 103015-100).

Analysis of Bone-targeting Antibody Conjugates Distribution

1mg/kg of Tras and Tras-ALN were injected retro-orbitally (n=3) twice a week for 4 weeks. Tissues were harvested, fixed in 4% PFA for 10 min, snap frozen in O.C.T., and kept at −80 °C until sectioning and usage. To assess the tissue distribution of Tras and Tras-ALN, the following antibody against Tras was used: F(ab’)2-Goat anti-Human IgG Fc Secondary Antibody, FITC (RRID:AB_2536550). Images were obtained via confocal microscopy (Leica TCS SP5) and quantified using Fiji (RRID:SCR_002285).

Micro Computed Tomography (microCT)

For microCT, bilateral hindlimb specimens were dissected to isolate the femurs by disarticulating at the hip joints and removing adjacent soft tissues. The specimens were fixed in 4% formaldehyde and kept in 70% Ethanol in individual plastic tubes until microCT scanning. High-resolution microCT scans of the entire femur specimens were acquired with a spatial resolution of 6.5 μm (Skyscan1174v2; Bruker, Kontich; Belgium). The X-ray source was set at a voltage of 50kV, a current of 200μA, and a rotation step at 0.2°. 3D visualization was performed using the NRecon and CTvox software (Bruker, Kontich; Belgium), whereas total bone volume and mineral density were calculated using the CTAn software at the density thresholds of 68 g/cm² and 120 g/cm² representing low- and highly mineralized bone respectively, following calibration against standard mouse density microCT phantoms. The average mineral density of the entire femur was also calculated. MicroCT results of the unilateral samples with metastasis were compared between the study groups.

Flow Cytometry

To perform immunophenotyping, we collected 60ul of peripheral blood, retro-orbitally, from mice that were treated with ALN or Tras-ALN for 4 consecutive weeks. Samples were directly collected in 940 μl of RBC lysis buffer, incubated for 10 min at room temperature, and centrifuged at 600g for 5 minutes. Next, cells were blocked on ice with 100 μl anti-CD16/32 antibody (RRID:AB_2621443, 0.2 mg/mL, 1:100) for 10 minutes. After centrifugation and removal of the supernatant, 100 μl of pre-conjugated antibody cocktail was added per sample, transferred to a round bottom 96-well plate, and incubated for 15 minutes at 4 degrees Celsius. A cell pellet was then generated through centrifugation and resuspended in 250 μl of 2% FBS buffer, and flow was performed on the BD LSR Fortessa. Immune cells were profiled using the following antibodies CD45-VF450 (RRID:AB_2621947, 1:200), CD11b-APC/Cy7 (RRID:AB_2621625, 1:200), Ly6g-Percp/Cy5.5 (RRID:AB_2621899, 1:200), CD3e-PE (RRID:AB_2621730, 1:200), CD4-APC (RRID:AB_2621543, 1:200), CD8a-FITC (RRID:AB_2621671, 1:200). We gated each immune cell subtype as follow: Neutrophil: CD45+ CD11b+ Ly6G+ (Ly6Cmid-low); Monocyte: CD45+ CD11b+ Ly6Chigh (Ly6G-); Macrophages: CD45+ CD11b+ LY6G− Ly6C - F4/80+ CD11C+ MHII+; CD4 T cells: CD45+ CD11b− CD3+ CD4+; CD8 T cells: CD45+ CD11b− CD3+ CD8+.

Temporal Assessment of Her2 in DTCs

To assess Her2 expression in DTCs, cancer cells were delivered via IIA injection to the right hindlimb of mice. Bone marrow was collected by flushing out all cells with PBS using a 26G syringe. Skeletal tissues were mechanically dissociated using a pestle and mortar before processing with the MACS dissociation kit following manufacturer’s instructions. The dissociation buffer was supplemented with 0.1mg of collagenase I and II per ml of dissociation buffer. Tissues were incubated for 40 min at 37°C. A second dissociation was performed, and cells were vortexed for 10 seconds before single-cell filtration and collection by centrifugation at 1200RPM for 5 min. All samples (bone marrow and bone matrix). Cells were treated with 2ml of 1xRBC lysis buffer and incubated at room temp for 5 min before centrifugation (1200RPM for 5 min). Cells were washed twice with PBS containing 3% BSA before staining and flow analysis. Samples were collected at days 3 and 14 post-injection and non-injected parental cells were used as control.

Single-cell RNA sequencing library preparation

For single-cell RNA sequencing experiments, 2×10⁵ cells were seeded in triplicate in 6 well plates for 24 hours. Next, we collected cells using trypsin and washed them twice with PBS containing 3% BSA before adding 100 μl of multiplexing oligos to each pellet. Following dissociation, cells were incubated for 5 min at room temperature before subsequent washes with 2 ml of PBS containing 3% BSA. Cells were resuspended in 500 μl of PBS containing 1% BSA, counted, and 1.5×10⁵ cells from individual samples were pooled into one library. After cell viability assessment, the library was centrifuged (1200 RPM for 5 min), the supernatant removed, and the pellet was suspended to a final concentration of 1.5 million cells per ml. We found that most samples had over 90% cell viability, therefore, samples were directly processed for library preparation without the need for cell sorting based on DAPI. For multiplexing, we used CellPlex, a cholesterol-modified oligonucleotide (CMO) cell tagging-based approach from 10x Genomics. We barcoded each sample (MCF7 and bone-derived MCF7) before pooling samples for single-cell library preparation. This allowed us to multiplex cells to minimize batch effect between samples. Samples were sequenced by the Genomic and RNA Profiling Core (GARP) at Baylor College of Medicine. Cell hashing was performed with CellRanger using HTODemux, a K-medoid clustering approach based on normalized hashtag oligo counts which maximized the CMO assignment (57).

Single-cell RNA-Sequencing data quality control and preprocessing

Sequenced fastq files were aligned, filtered, barcoded and UMI counted using Cell Ranger Chromium Single Cell RNA-seq version 7.1.0, by 10X Genomics with Cell Ranger, GRCh38 database (version 2020-A) as the human genome reference. Cells were demultiplexed into separate datasets by sample using the Seurat HTODemux function. Cells classified as negative and doublets, based on CMO tag counts, were filtered out. Only single cells with at least 1000 UMIs, ≥400 genes expressed, and <25% of the reads mapping to the mitochondrial genome were retained for further analysis. UMI counts were then normalized so that each cell had a total of 10,000 UMIs across all genes and these normalized counts were log-transformed with a pseudocount of 1 using the “LogNormalize” function in the Seurat package. The top 2000 most highly variable genes were identified using the “vst” selection method of “FindVariableFeatures” function and counts were scaled using the “ScaleData” function. Datasets were processed using the Seurat package (version 4.0.3) (RRID:SCR_007322)(58).

Single-cell RNA-Sequencing data dimensionality reduction and integration

Principal component analysis was performed using the top 2000 highly variable features (“RunPCA” function) and the top 30 principal components were used in the downstream analysis. Diffusion maps were generated as implemented in the destiny (version 3.4.0) R package (RRID:SCR_001905)(59) with default parameters and using 10,000 subsampled cells from each integrated dataset. Datasets for each sample were integrated separately by using the “RunHarmony” function in the harmony package (version 0.1.0) (RRID:SCR_022206)(60). K-Nearest Neighbor graphs were obtained by using the “FindNeighbors” function whereas the UMAPs were obtained by the “RunUMAP” function. The Louvain algorithm was used to cluster cells based on expression similarity. The resolution was set at 0.4 for optimal clustering.

Single-cell RNA-Sequencing data cell type annotation

Differential markers for each cluster were identified using the Wilcox test (“FindAllMarkers” function) with adjusted p-value < 0.01, absolute log2 fold change > 0.25, and minimum 10% of cells expressing the gene in both comparison groups using 1000 random cells to represent each cluster. For custom and differential markers, UMAP and TSNE plots are created using the FeaturePlot command. Violin and ridge plots showing expression value per cluster are created using ggplot2 (v3.3.5, RRID:SCR_014601)(61). Gene set enrichment analysis was performed on gene lists sorted by −log10 of p-value * log2 fold change using clusterProfiler (v4.2.2, RRID:SCR_016884), and enrichment of selected terms was visualized using gseaplot2 from enrichplot (v1.14.2) and bubble plots were generated by ggplot2(62). Enrichment scores for MYC signaling were generated using the “AddModuleScore” function from the Seurat package based on gene sets obtained from MSigDB(63).

Violin plots test statistics

Custom violin plots for marker genes were generated using the “VlnPlot” function from the Seurat package. P-values were added to the plots using the “stat_compare_means” function from the ggpubr package using the Wilcoxon test (version 0.4.0, RRID:SCR_021139).

Chromatin Accessibility Profiling

We analyzed ATAC-seq datasets available to the public (GSE160582). BAM files from passages 2, 6, and 12 of M7-SCP2 and bone-derived M7-SCP2 were merged using IGV.2.10.3. This process allows the identification of stable open chromatins following M7-SCP2 cell extraction from bone. Group autoscaling was applied to all conditions when analyzing chromatin accessibility for specific genes.

miRNA prediction, validation, and analysis

For miRNA prediction based on our target of interest (ERBB2), we used TargetScanHuman 8.0 (RRID:SCR_010845), an open source that searches for conserved sites of miRNAs and ranks best candidates based on biochemical models and cumulative weight. To identify relevant miRNAs, we evaluated their association with the overall survival (OS) of patients in HR+ breast cancer cohorts (METABRIC). Only miRNAs with significant impact on OS were selected for further assessments in the bone microenvironment. This was used as a complementary approach to published studies on ERBB2-targeting miRNAs. For miRNA assessment, total RNA was collected from 7.5×105 cells cultured in serum-free DMEM/F12 medium for 48 hours using Trizol. Reversed transcription was performed using the MicroRNA Reverse Transcription Kit (4366596). The following oligos were purchased from Thermofisher: hsa-miR-133b (Assay ID: 002247) and RNU6B (Assay ID:001093) as reference genes. TaqMan^™ Fast Advanced Master Mix (4444557) was used for sample preparation and real-time PCR was run on the CFX Opus 384 system (Biorad).

Data Availability

Genetic mutations were curated from the MetTropism data available on the cBioportal (RRID:SCR_014555). METABRIC was used for survival analysis and is freely available. Single-cell datasets have been deposited in the GEO database under GSE230612 and are accessible upon request to the lead author.

Statistical Analysis:

Statistical analyses were performed with GraphPad Prism or IBM SPSS Statistics (Version 29.0). The miRpower tool was used for miRNA-related survival analysis (RRID:SCR_018753). Statistical approaches are specified for each experiment and only P-values < 0.05 are considered significant.