Clonal heterogeneity in ER+ breast cancer reveals the proteasome and PKC as potential therapeutic targets

Beumers, Lukas; Vlachavas, Efstathios-Iason; Borgoni, Simone; Schwarzmüller, Luisa; Penso-Dolfin, Luca; Michels, Birgitta E.; Sofyali, Emre; Burmester, Sara; Heiss, Daniela; Wilhelm, Heike; Yarden, Yosef; Helm, Dominic; Will, Rainer; Goncalves, Angela; Wiemann, Stefan

doi:10.1038/s41523-023-00604-4

Download PDF

Article
Open access
Published: 02 December 2023

Clonal heterogeneity in ER+ breast cancer reveals the proteasome and PKC as potential therapeutic targets

npj Breast Cancer volume 9, Article number: 97 (2023) Cite this article

1481 Accesses
2 Altmetric
Metrics details

Subjects

Abstract

Intratumoral heterogeneity impacts the success or failure of anti-cancer therapies. Here, we investigated the evolution and mechanistic heterogeneity in clonal populations of cell models for estrogen receptor positive breast cancer. To this end, we established barcoded models of luminal breast cancer and rendered them resistant to commonly applied first line endocrine therapies. By isolating single clones from the resistant cell pools and characterizing replicates of individual clones we observed inter- (between cell lines) and intra-tumor (between different clones from the same cell line) heterogeneity. Molecular characterization at RNA and phospho-proteomic levels revealed private clonal activation of the unfolded protein response and respective sensitivity to inhibition of the proteasome, and potentially shared sensitivities for repression of protein kinase C. Our in vitro findings are consistent with tumor-heterogeneity that is observed in breast cancer patients thus highlighting the need to uncover heterogeneity at an individual patient level and to adjust therapies accordingly.

PERCEPTION predicts patient response and resistance to treatment using single-cell transcriptomics of their tumors

Article 18 April 2024

Predictability of B cell clonal persistence and immunosurveillance in breast cancer

Article Open access 02 May 2024

KRAS mutation: from undruggable to druggable in cancer

Article Open access 15 November 2021

Introduction

Breast cancer remains the most commonly diagnosed malignancy in women worldwide¹ and comprises a heterogeneous group of diseases characterized by distinct clinical, histopathological, and molecular features². Accordingly, breast tumors are initially classified by immunohistochemistry (IHC) based on the expression of estrogen receptor (ER), progesterone receptor (PR), human epithelial growth factor receptor 2 (HER2) and the proliferative marker Ki-67³. Around 60–70% of tumors are characterized by ER expression⁴. Patients with ER+ tumors are efficiently treated with endocrine therapies to abrogate the tumor promoting effects of estrogen⁵, leading to overall survival rates of over 90%⁶. First-line endocrine therapy consists of selective ER modulators (SERMs) such as Tamoxifen and of aromatase inhibitors (AIs) for pre- and postmenopausal patients, respectively. Despite high therapeutic efficacy, up to 40% of patients with late-stage diagnoses relapse within 20 years of treatment start⁷ and recurrences can be caused by diverse resistance mechanisms⁸. Along these lines, we recently identified ATF3, a member of the ATF/CREB family of transcription factors, as novel driver of endocrine therapy resistance in vitro and in vivo⁹, and pinpointed candidate methylation sites which are clinically associated with endocrine therapy resistance¹⁰.

Tumor cell heterogeneity contributes to therapy resistance by increasing the likelihood of cells having a resistant phenotype or the capacity to acquire such a resistant phenotype are present prior to treatment^11,12. Next-generation sequencing (NGS) has fundamentally uncovered genetic heterogeneity also in breast tumors^13,14. However, tumor heterogeneity and response to therapy can be driven by genetic, epigenetic and stochastic events in the tumor cells, and by local or systemic factors such as tissue context and clonal expansion in healthy tissue^15,16. Stochasticity in gene expression and other mechanisms triggering protein abundance may lead to variable protein expression states among individual cells and induce heterogeneity of cellular phenotypes¹⁷. Tumor heterogeneity may destine therapeutic resistance via two mechanisms: Pre-existent resistant clones may be selected through the applied therapy (i.e., primary resistance)^18,19 or initially treatment-persisting cells may adapt to and then acquire resistance to the applied therapy (secondary resistance)²⁰. While the former mechanism points at the presence of alterations that confer a resistance phenotype in some tumor cells before therapy, the latter is rather connected to stochastic events that are selected and then fixed in tumor cells carrying eventually dominating molecular states.

Here, we leveraged cellular barcoding, cell cloning and integrative multi-omics analysis to disentangle clonal adaptions to endocrine therapies. We barcoded two treatment-sensitive ER+ breast cancer cell lines and then rendered them resistant to Tamoxifen (TAMR) or long-term estrogen deprivation (LTED), the latter mimicking clinically applied aromatase inhibition. We find evidence of both, the selection of pre-existing resistant clones and for acquired resistance in persister cell populations. Individual clonal populations were isolated and characterized at the phenotypic, gene expression (RNA-Seq) and phospho-proteome levels. There, we identified private and potentially shared contributors to resistance, opening potential therapeutic vulnerabilities. Sensitivity to the proteasome inhibitor Bortezomib was private and correlated with activation of the unfolded protein response (UPR). In contrast, protein kinase C (PKC) isoforms were commonly activated in TAMR and LTED clones originating from one cell line, while treatment with the pan PKC inhibitor Sotrastaurin was significant just with one LTED clone. Clinical relevance of our findings was established in patient-oriented analysis of the CPTAC-BRCA proteogenomic cohort²¹ upon stratification by ER− vs ER+. PKC activation in particular, was evident in two patients with ER− disease, which closely resembled our in vitro models. Taken together, we developed cellular models of endocrine resistance that revealed private and potentially shared contributors to endocrine therapy resistance thus supporting the need for molecular stratification at the single patient resolution for treatment of advanced luminal breast cancer.

Results

Clonal in vitro models of endocrine therapy resistance

In this study, we wanted to disentangle the heterogeneity of endocrine therapy resistance in individual clones using in vitro models. Molecular changes should be uncovered at the gene expression and phospho-proteomic levels, and new potential vulnerabilities be identified. To this end, we transduced treatment-naïve ER+ breast cancer cell lines MCF7 and T47D with the ClonTracer library²² at an MOI of 0.05 to favor single barcode integration. Following short-time expansion, cultures were split into twenty replicates for every cell line: Five replicates of each were frozen down to later assess the barcode complexity in the initial cell population (Supplementary Fig. 1). Five replicates of each cell line were treated for eight months with either 4-OHT or estrogen deprivation (-E2) to induce Tamoxifen resistant (TAMR) and long-term estrogen deprived (LTED) cell lines, respectively. The latter mimics aromatase inhibition. Control cells were kept in standard media (+E2). Tamoxifen reduced the proliferation of cells, however, did not induce a quiescent state. In contrast, cells kept in estrogen deprived conditions entered a quiescent state for several months and re-gained proliferative capacity after six months.

At eight months of cultivation, treated replicates had regained a more proliferative phenotype compared to +E2 replicates that had been challenged with endocrine therapies only for a short time, indicating successful induction of endocrine resistance in the former (Fig. 1a). Bulk sequencing of barcodes in every replicate discerned the barcode- and the inferred clonal complexities. At least 5 × 10⁶ barcode reads were obtained for all cell lines (Supplementary Table 1). The theoretical maximum of approximately 1*10⁵ transduced barcodes was found in the initial cell population of the barcoded MCF7 cell line (1.15*10⁵ unique barcodes), while the complexity was lower (4*10⁴ unique barcodes) in T47D (Fig. 1b). Prolonged cultivation of replicates led to a strong reduction in the complexity of barcodes even in the absence of endocrine therapy (i.e., the +E2 controls). Treatment with 4-OHT and particularly with estrogen deprivation (-E2) induced a significant further decrease in the numbers of distinct barcodes, however, also an enrichment of a few barcodes, while no barcodes were enriched in the initial barcode library (Supplementary Fig. 2, Supplementary Files 1–3). This supports the intended rather homogenous representation of barcodes in the initial barcoded pool as well as neutral drift potentially having occurred in some replicates.

**Fig. 1: Initial analysis of barcoded endocrine resistance models.**

Next, we inspected the identity of barcodes in the respective replicates and found extensive heterogeneity regarding numbers and recurrence of barcodes between individual replicates. In Fig. 1c, for MCF7 the five and for T47D the 25 most abundant barcodes in each replicate are color coded. Other barcodes are shown in gray. The apparent barcode- and clonal complexity appeared to be higher in MCF7 than in T47D. In the MCF7 TAMR replicates only 16 barcodes along all TAMR replicates were enriched with a frequency greater than five percent. None of those barcodes was enriched in more than one replicate, pointing to resistance acquisition of distinct clones that were selected during treatment. In contrast, 4/5 replicates in the T47D TAMR model were dominated by a clone carrying the purple barcode and this was prominent also in the fifth replicate (Fig. 1c). The recurrent enrichment of clones carrying the purple barcode could suggest that the respective cells had been Tamoxifen-resistant already prior to the onset of treatment as this barcode was overrepresented neither in the T47D + E2 controls nor in the LTED conditions (Supplementary Fig. 2).

We then isolated single clones from selected replicates (Supplementary Fig. 3a and Fig. 1c). The barcodes in these clones were sequenced and associated with the same color code that was used for the whole replicate. At least three clones carrying the identical barcodes could be isolated from two independent replicates each of the T47D TAMR and LTED conditions. Clones carrying the orange barcode were isolated from two LTED replicates a well as from the +E2 control replicate. While clonal populations having been isolated from the same replicate had been cultivated independent of one another for a period of at least six months, clones isolated from different replicates of any condition had been cultivated independently for at least 14 months (compare Supplementary Fig. 1).

Barcode sequencing of individual clones revealed that most T47D clones carried only one barcode, as was expected given the MOI of 0.05 that had been used in the generation of the barcoded libraries. However, the purple T47D TAMR clones carried two barcodes that were found at the same ratio in all replicates. The MCF7 blue TAMR_1 clones, for example, had even 17 barcodes integrated into their genomes (Supplementary Fig. 3b), requiring an NGS approach to identify all barcodes. A list of clones and their integrated barcodes is provided in Supplementary Table 2. Some barcodes detected in sequencing of MCF7 clones were not detected in the initial barcoded cell pools, potentially because some barcodes did not match the expected pattern of ([AT][GC])x15²². As we determined the barcodes in individual clones only for those MCF7 replicates and clones that are indicated in Fig. 1c, the apparent clonal complexity in other replicates is likely exaggerated in the figure. Yet, we refrained from isolating single clones from the MCF7 control (+E2) and instead, employed pools of three MCF7 + E2 replicates (i.e., +E2_1, +E2_3, and +E2_5) as controls in any experiments using MCF7 cells throughout the study.

Finally, we wanted to uncover the chromosome positions virus-encoded barcodes had integrated into, to potentially exclude viral integration into genomic loci as drivers of resistance. To this end, we amplified and sequenced integration sites in MCF7 replicates TAMR_2 and LTED_5 as well as T47D replicates +E2_2, TAMR_2, and LTED_2. Read counts of LAM-PCR²³ libraries (File ‘Integration site analysis.xlsx’ at Zenodo)) were matched with barcode read counts in the same replicates suggesting viral integration sites for particular barcodes and clones. Matches between integration sites and barcodes were validated for the individualized clones using PCR (Supplementary Fig. 3c–f). None of the viral integration sites (Supplementary Table 3) were associated with significant expression changes of nearby genes (|log2FC| ≥ 0.05 and p-adj. ≤ 0.05), except for integration of a barcode into Chr11:31,426,888 (GRCh38/hg38) in the orange barcode in T47D LTED and +E2 clones. This barcode had integrated into intron 4 of DNAJC24. The gene was significantly higher expressed in T47D orange LTED clones compared to orange + E2_5 clones (data not shown). Since the orange clonal populations were all characterized by the same barcode and integration site, however, upregulation of DNAJC24 expression was likely a treatment-specific event (LTED vs +E2) and not related to the viral integration. We, therefore, excluded viral integration and potential deregulation of genes in close proximity to the viral integration sites as potential drivers of endocrine therapy resistance for the isolated clones.

Gene expression are stable over time in endocrine resistant clones

Next, we wanted to know if the observed heterogeneity of clones in the different conditions and replicates was mirrored at the gene expression level. To this end, RNA was sequenced from clones sharing the same as well as having different barcodes. Unsupervised hierarchical clustering of the top 200 most variable genes across all samples in each dataset (Supplementary Files 4 and 5) revealed stable clonal lineages, however, also extensive heterogeneity between clonal populations having different barcodes. Individual clones sharing the same barcode could be robustly discriminated from clones that had any other barcode, both for MCF7 (Fig. 2a) and T47D (Fig. 2b). This suggests that gene expression patterns were stable in clones sharing the same barcode even though the span between clone isolation and molecular profiling was at least six months of continuous culture. While this might be expected for clones isolated from the same replicate, this was also apparent for clones from different replicates thus having been cultured independently for at least 14 months (compare Supplementary Fig. 1). This could implicate that the respective constant stressor (i.e., 4-OHT and estrogen-deprivation, respectively) induced fixation of particular, yet for different clones individual, gene expression patterns.

**Fig. 2: Gene expression trajectories are stable in clonal lineages.**

Clones having derived from the same parental cell display distinct phenotypes after selection

Having identified stable clonal trajectories in the RNA-Seq data, we next assessed the proliferative capacity of individual clones. Here, we observed similar but distinct proliferative capacities of clones even when these carried the same barcodes and had clustered in the RNA-seq experiment (Fig. 3a–c). Hence, the rather homogenous gene expression patterns present in clones carrying the same barcodes were not mirrored by a similar homogenous distribution of proliferative phenotypes. Next, we focused specifically on the orange clones, where we had isolated five clones from the +E2_5, and four and another three clones from the LTED_1 and LTED_2 replicates of the T47D cell line, respectively. These clones had all derived from the same initial cell as they carried the same barcode integrated into the identical chromosomal location (Fig. 3f). We tested the clones in cellular assays informing on proliferation, cell cycle progression and apoptosis (Fig. 3d–f). This revealed consistent clonal patterns within and also between replicates of the same condition (i.e., LTED_1 and LTED_2). The orange barcoded T47D + E2_5 showed robust proliferation, however, this was strongly reduced upon short-term estrogen deprivation. The respective T47D LTED clones showed a similar phenotype having limited proliferative capacity when grown without estrogen. Conversely, strong proliferation could be induced by short-term supplementation of estrogen (Fig. 3d). The short- and long-term estrogen deprived orange barcoded T47D clones thus displayed a ‘cycling persister’ phenotype²⁴ characterized by a reduced but prominent fraction of cycling cells and a strong increase in dying/dead cells compared to non-treated counterparts from the same clonal origin (Fig. 3e, f). Given the rather similar growth patterns observed with orange + E2 and LTED clones, we next utilized our RNA-Seq data to identify deregulated pathways.

**Fig. 3: MCF7 and T47D TAMR and LTED clones display distinct phenotypes.**

Pathway activities in MCF7 and T47D TAMR and LTED clones reflect intra- and inter-tumor heterogeneity

Here, wanted to investigate if the differences in clonal gene expression patterns were reflected by similar differences in pathway activities and if these could hint at common or private molecular contributors to drug resistance. To this end, we considered clones isolated from the same replicate and sharing the same barcode as biological replicates. PROGENy^25,26 was employed with RNA-seq data to estimate the activity of 14 major signaling pathways in drug-treated cells compared to untreated controls. Estrogen signaling was significantly repressed in all clones (Fig. 4), reflecting the therapeutic block imposed by 4-OHT and estrogen deprivation. Mutation of ESR1 has been determined as one mechanism of acquired endocrine therapy resistance²⁷. Consistent with the observed downregulation of ER−signaling in our resistant models, we did not observe ESR1 hotspot mutations in any endocrine therapy resistant clones (Supplementary Fig. 4). Hence, we next focused on other pathways and mechanisms that might be common or contribute to resistance just in particular clones.

**Fig. 4: Pathway analysis of clonal populations reveals cell line- (MCF7 vs. T47D), treatment- (TAMR vs. LTED), and clone-specific activities.**

JAK-STAT signaling was strongly upregulated only in the LTED clones of MCF7 (Fig. 4a), while this pathway was consistently downregulated in all T47D clones (Fig. 4b). Hence, this regulation was different for the two treatments in MCF7, and in opposite directions for the two cell lines. Similarly, a significant activation of NFκB signaling was shared among all MCF7 TAMR and LTED conditions, while this pathway was not changed or even downregulated in T47D TAMR and LTED clones. Both, JAK-STAT and NFκB signaling have been associated with the tumor (immune) microenvironment^28,29.

In summary, we have uncovered substantial commonalities in pathway activities between cell lines and treatment conditions, however, also specific differences that might be related to inter- and intra-tumor heterogeneity. We next wanted to test potential common as well as private therapeutic vulnerabilities in the endocrine therapy-resistant clones, focusing on the MCF7 TAMR and LTED cell models.

Sensitivity to the proteasome inhibitor Bortezomib correlates with unfolded protein response as a private vulnerability

To find therapeutic vulnerabilities, we next estimated transcription factor activities using DoRothEA³⁰ and decoupleR³¹. Furthermore, we analyzed kinase activities using matched phosphoproteomic data with Omnipath³² and decoupleR³¹. The latter resulted in the identification of at least 17,000 phosphosites (class 1, localization probability >95%) in every sample (MCF7: 20,207–25,874; T47D: 17,118–24,449). Of these, between 46–74% were shared between all clones sharing identical barcodes (Supplementary Fig. 5a). Additionally, unsupervised hierarchical clustering of all consistently identified phosphopeptides showed that clones with the same barcode clustered together (Supplementary Fig. 5b), corroborating also our results from gene expression analysis (see Fig. 2).

We had found NFκB signaling differentially regulated in endocrine therapy resistant MCF7 vs. T47D clones (compare Fig. 4a, b). Analysis of phospho-proteomic data however did not reveal patterns indicating potential activation of the IκB kinases IKKα, IKKβ, IKKε or TBK1 (Fig. S6a) suggesting that potential pathway activation should be downstream of IκB kinases. The key downstream transcription factor p65/RELA³³ was estimated from the RNA-Seq data to be relatively activated in MCF7 LTED clones and, to a lesser extent in the TAMR clone (Fig. S6b). Knock-down of RELA (Fig. S6c) induced similar phenotypes in the endocrine sensitive +E2_5 cells and in TAMR as well as LTED clones as proliferation and cell viability as well as the percentage of cycling cells were consistently decreased while dying/dead cells were increased (Fig. S6d). These findings point at an essential role for p65 even in endocrine sensitive cells, suggesting that complete inhibition of NFκB signaling might be too harsh in view of a potential window of therapeutic targeting.

In contrast, MCF7 blue TAMR_1 clones also showed a significant enrichment in the expression of genes that are connected to the unfolded protein response (UPR), pointing to its activation (Fig. 5a) and to a potential dependency on the proteasome. Accordingly, we utilized the proteasome inhibitor Bortezomib which is applied to treat NFκB-driven multiple myeloma³⁴ and is hypothesized to abrogate also the UPR³⁵. We initially assessed the sensitivity for Bortezomib in MCF7 clones and one +E2 replicate (+E2_5), which revealed the highest sensitivity (Fig. 5b) to Bortezomib treatment for the MCF7 blue TAMR_1 cells that had also shown a predicted activation of UPR. No association between activation of NFκB signaling and sensitivity to Bortezomib was evident. We next assessed the effects Bortezomib treatment on cell proliferation, viability, and the ratios of cycling cells in the MCF7 TAMR and LTED clones (Fig. 5c). We used Bortezomib at a 1.85 nM concentration as we had observed the strongest difference in sensitivities for the respective cell clones there. A strong and significant effect was only detected in the blue TAMR cells while the other cells were not at all or just slightly (red LTED_5) affected. Taken together, sensitivity to Bortezomib correlated with activation of the UPR just in the MCF7 blue TAMR_1 clones thus highlighting activation of the proteasome and inhibition by Bortezomib as private and druggable endocrine therapy resistance contributor to a more complex, multifactorial resistance phenotype.

**Fig. 5: Sensitivity to Bortezomib treatment correlates with activation of the unfolded protein response (UPR) in MCF7 models.**

PKC activation as potentially shared contributor to endocrine therapy resistance in different clones

Kinase activity analysis revealed significantly higher activation of several isoforms of protein kinase C in MCF7 TAMR as well as in LTED clones compared to the +E2 control replicates (Fig. 6a). Treatment of the cell lines with the pan-PKC inhibitor Sotrastaurin showed higher sensitivity of the yellow MCF7 LTED_2 clone (Fig. 6b) compared to the other clones and replicates. Treatment with 2.50 µM Sotrastaurin reduced the proliferative capacity of these cells by almost 60% compared to +E2_5 cells (Fig. 6c). Accordingly, viability and cell death were reduced and enhanced, respectively, while the number of cycling cells was not affected. This suggests that the inhibitory effect of Sotrastaurin was mostly conferred by increased induction of cell death. Cellular viability was significantly reduced also in MCF7 blue TAMR_1 and red LTED_5 clones upon treatment with 2.50 µM Sotrastaurin and showed a trend when compared to Sotrastaurin-treated +E2_5 control cells (Fig. 6c). Yet, activation of PKC by TPA was significantly and similarly repressed by 66–80% in all endocrine sensitive and resistant clones tested (Fig. 6d), pointing at a similar inhibition of PKC signaling. The consistent effects in reduced cellular viability in MCF7 yellow LTED_2 and to a lesser extent also for red LTED_5 clones, thus point to PKC activation as a potentially shared contributor to endocrine therapy resistance.

**Fig. 6: PKC is a potentially shared contributor to endocrine therapy resistance.**

Mechanisms contributing to clonal endocrine resistance have potential clinical relevance

Finally, we assessed clinical relevance of our in vitro findings. To this end, we retrieved and analyzed data from the CPTAC-BRCA cohort²¹, which contains paired gene expression and phospho-proteomic data from 120 treatment-naïve breast cancer patients with known ER−status, i.e., ER− (n = 39) and ER+ (n = 81) disease. Here, we were focused on patients with ER− disease; tamoxifen treatment and estrogen deprivation had abrogated estrogen signaling in our in vitro models, rendering them ‘quasi’ ER−, and ‘subtype switching’ has been reported for luminal A and luminal B patients upon endocrine therapy³⁶. Pathway analysis of the CPTAC RNA-seq data showed significantly elevated NFκB signaling in patients of the ER− subtype compared to patients with ER+ disease (Fig. 7a), in line with the observations we made in all MCF7 blue TAMR_1 and LTED clones. Similarly, TNFα, JAK-STAT, and hypoxia signaling were associated with ER- disease, also matching our in vitro findings in MCF7. Further, p65 activity was significantly associated with ER− patients, while the UPR showed only a trend. Different from our in vitro models, activation of different PKC isoforms was not associated with either ER− or ER+ disease in the patients (Fig. 7b). We thus performed data analysis on a per patient basis to correlate our in vitro results with alterations in individual patients within a potentially heterogeneous cohort.

**Fig. 7: Activation of UPR and PKC is reflected in breast cancer patients from the CPTAC cohort.**

Indeed, patient-specific pathway activity analysis highlighted strong inter-patient heterogeneity in both the ER− (Fig. 7c) and the ER+ cohorts (Supplementary Fig. 7a–d). JAK-STAT signaling was significantly activated and downregulated in 15 and 12 patients with ER− disease, respectively. This heterogeneity is consistent with our in vitro models, where we had observed upregulation of JAK-STAT signaling only in the MCF7 LTED clones while this pathway was not activated in the blue TAMR_1 clones (compare Fig. 4a). Significant activation and repression of p65 activity was evident in 16 and 2 patients with ER− disease, respectively (Fig. 7c). Similarly, 16 and 2 patients with ER- disease showed significant activation and repression of UPR, respectively. Inter-patient heterogeneity was, however, observed also in ER+ disease. There, 6 patients showed significant up- and another 19 downregulation of p65 activity (Supplementary Fig. 7b), while 9 and 27 patients had activated and repressed UPR, respectively (Supplementary Fig. 7c). Finally, we investigated PKC activity at the single patient level and indeed observed strong inter-patient heterogeneity (lower panel in Fig. 7c). Two ER− patients (#31 and #33) showed consistent activation of PKCα, PKCβ, PKCγ, and PKCε. The same kinases were upregulated also in several other patients with ER− disease, yet not reaching significance. In contrast, PKCδ was hardly activated in any patients compared with the reference sample utilized by Krug et al.²¹ and was significantly activated in only four out of the 120 CPTAC patients, all with ER+ disease (Supplementary Fig. 7d). In contrast, we found activation of PKCδ in all MCF7 TAMR and LTED clones, however, this did not correlate with sensitivity to Sotrastaurin. PKCδ might thus not be the main determinant of Sotrastaurin sensitivity in our cell line models which is corroborated by the observation that the activities of the other PKC isoforms in several ER− patients were consistent with our in vitro models. It is tempting to speculate that these patients might benefit from (pan-) kinase inhibitors of PKC. Intriguingly, all PKC activities, also of PKCδ, were downregulated in some patients (e.g., #21), further supporting the need to assess pathway activities at the single patient level. Taken together, our in vitro models were in agreement with clinical findings from the CPTAC-BRCA cohort and particularly with ER−negative disease.

Discussion

Endocrine therapy resistance remains an urgent clinical problem as up to 41% of patients aged younger than 75 years who originally presented with advanced disease (i.e., pT2N ≥ 2: tumor diameter >2 cm and at least 4–9 positive lymph nodes) relapse within 20 years after initial diagnosis and treatment start³⁷. Recent findings have highlighted the importance of tumor heterogeneity in disease progression^11,38. In our study we set out to answer two questions: First, we aimed at deciphering the contribution of primary and secondary resistance to commonly applied first line endocrine therapeutics, namely Tamoxifen and aromatase inhibitors. Second, we wanted to unravel clinically relevant contributors to endocrine therapy resistance in individual clones. To this end, we (i) utilized a bulk barcoding approach in two luminal breast cancer cell line models, (ii) isolated individual clones from complex cell pools, and (iii) validated private and potentially shared contributors to endocrine therapy resistance. We are aware of some limitations our study has as we employed tumor cell lines and did not perform in-depth genetic characterization of the clones.

Our bulk analysis showed that Tamoxifen resistance was achieved by either selection of primary resistant clones or acquisition of secondary resistance. Recurrent outgrowth of clones carrying identical barcodes was prominent in the T47D TAMR and LTED models suggesting primary resistance in these cell line models. In contrast, the lack of commonly selected barcodes in any MCF7 replicates pointed at the acquisition of secondary resistance as the likely mechanism there. The latter observation is not in line with ref. ¹⁸, who had shown predominant selection of primary resistant clones to Tamoxifen and the estrogen degrader Fulvestrant. In our LTED models, we observed a combination of likely primary (e.g., T47D orange LTEDs) as well as secondary resistance (e.g., MCF7 red LTED_5 and T47D brown LTED_2). In all cell line models, different clones were selected in TAMR vs. LTED conditions, indicating independent modes of resistance to the two treatment regimes.

In the T47D model, the dominance of orange barcoded cells in the untreated +E2_5 replicate suggested a fitness advantage, like stronger proliferation, over other clones in the same replicate. When analyzing single clones, however, green + E2_5 clones proliferated significantly stronger than their orange counterparts. These findings point at clonal interactions, cooperativity and competition³⁹, which is likely to influence dynamics of different clones and that is lost when cultures are grown at the single clone level.

Initial analysis of differential pathway activities based on the RNA-Seq data highlighted strong inter- and intra-tumor heterogeneity, as estrogen signaling was the only significantly suppressed pathway in all isolated T47D and MCF7 TAMR clones. RNA-Seq revealed activation of the UPR as a private event for MCF7 TAMR clones. In line, elevated expression of a UPR gene signature was previously found to correlate with worse relapse-free survival during Tamoxifen treatment⁴⁰. Conversely, proteasome inhibitors such as Bortezomib have been hypothesized to potentiate and induce a detrimental UPR³⁵. Accordingly, we observed strong sensitivity to Bortezomib for blue TAMR_1 cells, which showed an activation of the UPR.

While UPR activation was a private event in our models, activation of PKC was observed in all MCF7 TAMR and LTED clones. Treatment with 2.50 µM Sotrastaurin preferentially reduced the cellular viability of yellow MCF7 LTED_2 clones and showed a trend also for red LTED_5 clones. Accordingly, different PKC isoforms have previously been implicated in endocrine therapy resistance and/or estrogen independent growth, such as PKCα⁴¹ and PKCε⁴². Mechanistically, we found that treatment with 2.50 µM Sotrastaurin did not significantly impact cells passing S-phase, which is in line with recent findings⁴³, but rather induced cell death.

The analysis of patient data from the CPTAC-BRCA cohort supported the relevance of our in vitro findings, and the heterogeneity of breast tumors at pathway, kinase and transcription factor activity levels. JAK-STAT, p65, and UPR activation was evident in a number of individual patients with ER− disease corroborating our in vitro results. However, activation of PKCα, PKCβ, PKCγ and PKCε was observed in just two patients with ER− disease (#31 and #33) and resembled the activity patterns in MCF7 TAMR and LTED clones. Our data suggests that patients with activation in UPR or PKC might benefit from treatment with the proteasome inhibitor Bortezomib or the PKC inhibitor Sotrastaurin in respective first line therapies. Multi-omics analysis, including that of phosphoproteomic data, turned out as essential to fully capture the heterogeneity in vitro and in patients, supporting the use of proteomics in precision oncology⁴⁴.

Methods

Generation of barcoded endocrine resistant cell lines

MCF7 (Cellosaurus CVCL_0031) and T47D (CVCL_0553) were obtained from ATCC (LGC Standards GmbH, Wesel, Germany). Cells were kept at 37 °C with 5% CO2 in a humidified atmosphere in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% FCS, 50 units/ml penicillin and 50 g/ml streptomycin sulfate (Invitrogen AG, Carlsbad, CA, USA), and 10^-8M 17-ß-estradiol (E2, Sigma-Aldrich, Saint-Louis, MI, USA). The ClonTracer library²² (Addgene # 67267) was co-transfected with second generation viral packaging plasmids VSV.G (Addgene #14888), and psPAX2 (Addgene #12260) into HEK293FT (Cellosaurus CVCL_6911) obtained from Thermo Fisher Scientific (Waltham, MA, USA). 48 h after transfection, the supernatant was cleared by centrifugation (500 × g/5 min) and passed through a 0.45 µm filter. 1 × 10⁵ MCF7 and T47D cells were transduced with the viral particles in the presence of 10 µg/mL Polybrene (Merck, Darmstadt, Germany) at a multiplicity of infection (MOI) of 0.05. Generation of viral particles and the transduction of T47D and MCF7 were performed by the Cellular Tools Core Facility, German Cancer Research Center (DKFZ). 1 µg/mL Puromycin (Merck) was used to select for successful integration events and transduced cells were initially expanded. Five replicates each were either frozen down immediately, kept under control conditions (+E2), were treated with 100 nM 4-hydroxytamoxifen (4-OHT, Sigma-Aldrich) i.e., the biologically more active metabolite of Tamoxifen⁴⁵, or were deprived of E2 and therefore kept in DMEM (w/o phenol red) supplemented with 10% charcoal stripped FCS. Continuous exposure for 8 months to either endocrine treatment rendered the barcoded cells Tamoxifen-resistant (TAMR) or resistant to long-term estrogen deprivation (LTED). During this time, cells were passaged when they had reached 80% confluency or reseeded in a new dish if they had not reached this within 4 weeks. Medium was changed every week to ensure availability of nutrients and to replenish the respective treatment, i.e., +E2 with +E2 media, TAMR with +4-OHT media, and LTED with -E2 media, except when otherwise stated. Control cells (+E2) required passaging on average every 4 days throughout. All cell lines were regularly authenticated by STR profiling (Multiplexion GmbH Heidelberg, Germany) as well as tested for potential mycoplasma contamination.

Barcode composition analysis

DNA was isolated using the DNeasy Blood & Tissue kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The barcode region was amplified as described (https://www.addgene.org/pooled-library/clontracer/, accessed 22/12/12) with the adapted thermocycling program: 95 °C–5 min, 30x (95 °C–30 s, 69 °C – 15 s, 72 °C – 9 s) and 72 °C – 7 min. PCR products were cleaned with the Wizard® SV Gel and PCR Clean-Up System (Promega, Madison, WI, USA). Concentration and size were determined with Bioanalyzer 10 (Agilent, Santa Clara, CA, USA). Ten samples were multiplexed per lane and sequenced on the HiSeq 2000 V4 Single-read 50 bp with the custom sequencing primer: 5′-TCTACACACTGACTGCAGTCTGAGTCTGACAG-3′. Sequencing was performed by the DKFZ NGS Core Facility. Barcode analysis was performed as described (https://github.com/luca8651/Barcode_analyses-python, accessed: 22/12/07). Barcodes with at least 5 × 10⁵ combined reads in all initial, +E2, TAMR and LTED replicates for each cell line (MCF7 or T47D) were log2 normalized and depicted as violin plots using ggplot2 (version 3.4.1) in R (version 4.0.2).

Isolation of clonal endocrine resistant subpopulations

Eight barcoded treatment naïve or resistant MCF7 or T47D cell pools were used to isolate clonal subpopulations utilizing a Cytena single cell spotter (https://www.cytena.com/wp-content/uploads/2022/07/F.SIGHT-Deconvoluting-clonal-complexity-of-barcoded-cell-populations-Rainar-Will-2022.pdf) or manual dispension. Clonality was verified by assessment of images taken during spotting of single cells, and by Sanger sequencing of barcode regions once clones had grown out. For the latter, DNA and RNA was isolated using the AllPrep DNA/RNA Micro Kit (Qiagen) according to the manufacturer’s instructions. Concentrations were determined with a NanoDrop® ND-1000 UV-Vis Spectrophotometer (Thermo Fisher Scientific). Barcode regions were amplified using Phusion Hot Start II DNA-Polymerase (2 U/µL) (Thermo Fisher Scientific) according to the manufacturer’s instructions using the following primers: Fw: 5′-GCTGTGCCTTGGAATGCTAGTTGG-3′ and Rev: 5′-TCTGCTGTCCCTGTAATAAACCCG-3′. Thermocycling was performed as: 98 °C – 2 min, 32x (98 °C – 10 s, 71 °C – 20 s, 72 °C – 70 s) and 72 °C – 10 min. PCR clean-up and Sanger sequencing were performed by Eurofins Genomics Germany GmbH (Ebersberg, Germany). Sequences were analyzed using SnapGene software (Dotmatics, Bishop’s Stortford, UK). Sanger chromatograms of some MCF7 TAMR and LTED clones were too complex for unambiguous determination of barcodes, thus necessitating deep sequenced using Illumina MiSeq technology. To this end, published reverse primers (https://www.addgene.org/pooled-library/clontracer/, accessed: 22/12/12) were extended by three bases (CTG) at the 3′ end. PCR amplification was performed for 48 isolated clones with Phusion Hot Start II DNA-Polymerase (2 U/µL) according to the manufacturer’s instructions and thermocycling was performed as: 98 °C–2 min, 32x (98 °C–10 s, 69 °C–20 s, 72 °C–12 s) and 72 °C–10 min. PCR products were cleaned using the Wizard SV Gel and PCR Clean-up System (Promega) according to manufacturer’s instructions. Concentration and size were evaluated with the Agilent 2200 TapeStation System (Agilent). Samples were sequenced as single library on the MiSeq V2 300 Nano (Illumina, San Diego, CA, USA) with the custom sequencing primer: 5′-TCTACACACTGACTGCAGTCTGAGTCTGACAG-3′. Sequencing was performed by the DKFZ NGS Core Facility. Barcodes of interest were identified in MiSeq derived fastq files using a custom script. Using R (version 4.0.3)⁴⁶ and R studio⁴⁷, the ShortRead package⁴⁸ was used to read fastq files individually. For each of the files, occurrence of the respective barcodes in the DNA sequence was counted searching for the respective sequence. For MCF7 TAMR_1 clones, highly similar reads were manually clustered and violin plots generated in GraphPad Prism (version 9.4.1; GraphPad, Boston, MA, USA). One outlier was identified using the ROUT method (Q = 0.1%) and removed from the data.

Integration site analysis

Virus integration sites were identified in MCF7 + E2_3 and LTED_5 as well as T47D + E2_3, TAMR_2 and LTED_2 by Genewerk GmbH (Heidelberg, Germany) as previously described²³. The top predicted integration sites of individual complex barcoded cell pools were validated by a PCR-based assay using the Phusion Hot Start II DNA-Polymerase (2 U/µL) (Thermo Fisher Scientific) according to manufacturer’s instructions. To this end, a primer binding only to the viral 3′LTR with the sequence 5′-CCCAACGAAGATAAGATCTGC-3′ was combined with primers binding adjacent to the identified individual integration sites. Consequently, integration sites validated in the complex barcoded cell pools were investigated in isolated clones. Thermocycling was performed as: 98 °C–2 min, 32x (98 °C–10 s, 65 °C–20 s, 72 °C–60 s) and 72 °C–10 min. PCR products were subjected to Sanger sequencing (Eurofins Genomics) and sequences analyzed with SnapGene software.

Assessment of ESR1 hot-spot mutation status in individual clones

Sequence covering the hot spot mutation site (codon 536–538) in the ESR1 gene was amplified from single clones representing each of the isolated populations using Phusion Hot Start II DNA-Polymerase (2 U/µL) according to manufacturer’s instructions and the following primer pair: Fw: 5′-AATACCCACTCCTGCTTGGC-3′, Rev: 5′-TATCTGAACCGTGTGGGAGC-3′. Thermocycling was performed as: 98 °C–2 min, 32x (98 °C–10 s, 66 °C–20 s, 72 °C–60 s) and 72 °C–10 min. PCR products were subjected to Sanger sequencing (Eurofins Genomics) and sequences were analyzed with SnapGene software.

Analysis of cell proliferation and viability

Cells were plated in black clear-bottomed 96 well plates (Greiner Bio-One International GmbH, Kremsmünster, Austria). Cell numbers were assessed either by counting RFP-positive cells or counting of nuclei after Hoechst 33342 staining using a IXM XLS microscope (Molecular Devices, San Jose, CA, USA). Cells were detected and counted with Molecular Devices Software using default settings for minimum RFP intensity, size of nuclei, and staining intensity. Cell numbers were expressed relative to the seeding or treatment controls. Cell viability was determined using the CellTiter-Glo® Luminescent Cell Viability Assay (Promega) according to the manufacturer’s recommendation. Luminescence was measured using a GloMax Discover System (Promega).

Analysis of cycling and dying/dead cells

Cells in S-phase were determined using the Click-iT™ EdU Cell Proliferation Kit for Imaging with Alexa Fluor™ 594 dye (Thermo Fisher Scientific) following the manufacturer’s recommendation, however, adapted to 96 well format. Cells were plated in black clear-bottomed 96 well plates and pulsed for 21 h with 10 µM EdU 72 h post treatment start or transfection by replacing half of the media with fresh growth media containing 20 µM EdU. Cells in S-phase were determined as Alexa Fluor 594 positive and normalized to total cell numbers as determined by Hoechst 33342 staining using a IXM XLS microscope. Dying/dead cells were stained with 1 µg/mL DAPI 96 h post treatment start or transfection, and counted with a IXM XLS microscope. Total cell numbers were determined by Hoechst 33342 staining and ratios were calculated.

Inhibitor treatment and siRNA transfection

Cells were either treated with the proteasome inhibitor Bortezomib⁴⁹ or the pan-PKC inhibitor Sotrastaurin⁵⁰. Inhibitors were purchased from MedChemExpress (Monmouth Junction, NJ, USA) and resuspended in dimethyl sulfoxide (DMSO, Sigma-Aldrich). Resulting data were normalized to DMSO controls. A pool of four siRNAs targeting RELA (LQ-003533-00-0002) and a non-targeting control (D-001810-10-20) were obtained from Dharmacon (Lafayette, CO, USA) and cells were transfected using RNAiMAX® (Thermo Fisher Scientific) in media lacking penicillin and streptomycin.

RNA-Sequencing

RNA was isolated from clonal cell line populations using the AllPrep DNA/RNA Micro Kit (Qiagen) according to the manufacturer’s instructions. RNA-Sequencing (RNA-Seq) was performed by the DKFZ NGS Core Facility using NovaSeq 6 K PE 50 SP for MCF7 and NovaSeq 6 K PE 50 S1 for T47D clones (Illumina). Initial data QC of raw sequence data was done using FASTQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/, version 0.11.9). Alignment of reads to the human reference genome (“GRCh38.p13.primary_assembly_gencode” (https://www.gencodegenes.org/human/release_39.html) was done using the Rsubread R package (version 2.6.4)⁵¹. Gene expression counts were quantified using the featureCounts pipeline⁵² with default settings, implemented also in the Rsubread R package. Initial processing, alignment and quantification were performed by the Omics IT and Data Management Core Facility (ODCF) using the DKFZ Compute Cluster running under CentOS Linux 7 (Core) (https://odcf.dkfz.de/). Complete downstream RNA-Seq analysis was performed with R (R version 4.1.0)/Bioconductor software⁵³ separately for each cell line, using the same settings. Initially, a non-specific intensity procedure was applied (function filterByExpr::edgeR R package) to remove non-expressed genes (<10 reads in at least one of the barcoding clone biological conditions). Normalization of the remaining gene counts was performed using TMM (Trimmed Mean of M-values) from edgeR R package (version 3.36.0)⁵⁴, followed by the voom function from the limma R package (version 3.50.0), aiming to increase statistical power and to account for non-biological variability⁵⁵. Finally, differential expression analysis between the different cell lines and clones was performed using the limma R package⁵⁶.

Pathway and TF activity analysis

PROGENy (Pathway RespOnsive GENes) R package (version 1.16.0) was utilized to disentangle transcriptional heterogeneity related to differentially activated pathways. PROGENy infers the activity of 14 major signaling pathways^25,26. Differentially activated transcriptional factors were retrieved using the decoupleR R package (version 2.1.6) to implement robust statistical models³¹. Then, the DoRothEA gene regulatory network resource R package (version 1.6.0) was applied to predict transcription factor (TF)—target gene interactions (A to C confidence levels)³⁰.

Functional enrichment analysis

Gene Set Enrichment Analysis (GSEA) was applied using the clusterProfiler R package (version 4.2.2)⁵⁷ and MsigDB All Hallmarks(H) gene sets from the msigdbr R package (version 7.5.1) (https://github.com/DavisLaboratory/msigdb) using default settings.

RT-qPCR

cDNA was synthesized from the isolated RNA using the RevertAid RT Reverse Transcription Kit according to the manufacturer’s instructions. Quantitative Reverse Transcription-PCR (RT-qPCR) was performed using the Power SYBR Green PCR Master Mix (2x) (Thermo Fisher Scientific) and a QuantStudio™ 5 Real-Time PCR System (Thermo Fisher Scientific). Data analysis was performed using QuantStudio™ Design & Analysis Software v1.5.0. Relative changes were calculated by the comparative Ct method (ΔΔCt method)⁵⁸ and relative expressions were visualized as 2^−ΔΔCt. ACTB forward: 5′-ATTGGCAATGAGCGGTTC-3′, ACTB reverse: 5′-GGATGCCACAGGACTCCA-3′, GAPDH forward: 5′-GAGTCCACTGGCGTCTTCAC-3′, GAPDH reverse: 5′-GTTCACACCCATGACGAACA-3′, RELA forward: 5′-GCTTGTAGGAAAGGACTGCC-3′, RELA reverse: 5′-GCTGCTCTTCTATAGGAACTTGG-3′.

Mass spectrometry

Cells were rinsed with ice-cold PBS and lysed in RIPA buffer (Thermo Fisher Scientific) supplemented with 1x cOmplete EDTA-free protease inhibitor, 1x PhosSTOP phosphatase inhibitor (both Roche, Basel, Switzerland), 10 mM NaF (Bernd Kraft, Duisburg, Germany), 1 mM Na3VO4 (Sigma-Aldrich), 250 U/ml Benzonase (Merck) and 10 U/mL Rnase-Free Dnase (Qiagen). Lysates were incubated on ice for 30 min and cleared by centrifugation. Protein concentrations were determined using the Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions.

Proteins (320 µg protein/sample) were precipitated⁵⁹. Afterward, the protein pellet was digested first using Lysyl endopeptidase for 4 h (Lys-C, Fujifilm, Hong Kong, China) and then using trypsin for 16 h, both at 37 °C. Peptides were desalted using Sep-Pak C18 cartridges (Waters, Milford, MA, USA) and an Immobilized Metal Affinity Chromatography (IMAC) column (Thermo Fisher Scientific) charged with FeCl₃ was used for the enrichment of phosphorylated peptides⁶⁰. The fraction containing the phosphopeptides was collected, and vacuum centrifuged to dryness. Enriched phosphopeptides were dissolved in 0.1% TFA and desalted using Stop and Go Extraction tips (Supelco, Bellefonte, PA, USA)⁶¹. Each sample was analyzed via LC-MS/MS for a total of 120 min of analysis time. Peptides were separated according to their hydrophobicity by liquid chromatography (UltiMate 3000, Thermo Fisher Scientific) with a 102 min linear gradient of 2–28% acetonitrile (Waters, 186008795, BEH C18 130 Å 1.7 µm 75 × 250 mm). The LC system was directly coupled to a MS system (Orbitrap Exploris 480, Thermo Fisher Scientific) via electrospray ionization. MS analysis was performed using data-independent acquisition (DIA). MS1 scans were acquired at a resolution of 120 K covering the range from 350–1400 m/z. Maximum injection time was 45 ms and the automated gain control (AGC) target was set to 3e6. MS2 acquisition was performed using 48 precursor isolation windows of variable width and 1 m/z overlap that covered the range from 400–1200 m/z. Fragment spectra were acquired at a resolution of 30 K and a normalized collision energy of 26% was applied. Maximum injection time was 54 ms and the AGC target was set to 1e6.

Peptide and protein identification and quantification from DIA raw data was performed with directDIA in the Spectronaut software (version 15, Biognosys, Switzerland) using the built in identification and localization algorithm⁶². The search was performed against the human proteome fasta file (Uniprot, downloaded on July 14th, 2020, with 74,811 entries). Default settings were applied with phosphorylation (S,T,Y) as an additional variable modification (localization cut-off = 0). Identified phosphorylated peptides were site-collapsed using the Perseus⁶³ (version 1.6.2.3) plug-in PeptideCollapse⁶² and a localization cut-off at 0.95 was applied.

Phosphoproteomics data analysis

Raw data output from Perseus was initially processed using the PhosR R package (version 1.4.0)⁶⁴. Raw phosphosite intensities were normalized using variance stabilizing transformation from the vsn R package (version 3.62.0)⁶⁵. In addition, phosphosites were removed if they were not quantified in at least 2 out of 3 replicate samples in at least one condition. Differential phosphosite abundance analysis was performed with the limma R package implementing the limma-trend pipeline⁵⁶.

Differentially activated kinases were identified initially utilizing the OmniPath R package (version 3.2.5) to retrieve enzyme-PTM relationships, in form of molecular networks with directed interactions and effect signs³². Then, the decoupleR R package (version 2.1.6) was used to implement specific statistical models (normalized weighted mean) to infer kinase activities³¹. The estimated activity score for each kinase was approximated by weighting the molecular readouts of its targets by their *mode of regulation* (Activation or Inhibition) & their relative *likelihood*. Only kinases having at least 5 target proteins (phosphosite_target) present in the analyzed phosphoproteomics data were regarded.

Western blotting

Proteins were separated by SDS-PAGE⁶⁶ using PageRuler™ Prestained Protein Ladder (Thermo Fisher Scientific) as size maker. Afterwards, proteins were blotted using Trans-Blot Turbo™ Mini PVDF Transfer Packs and the Trans-Blot Turbo™ Transfer System (Bio-Rad Laboratories, Hercules, CA, US) according to the manufacturer’s instructions. Membranes were blocked in Blocking Buffer for Fluorescent Western Blotting (Rockland, Philadelphia, PA, USA): TBS supplemented with 10 mM NaF and 1 mM Na₃VO₄. Subsequently, the membranes were incubated with antibodies targeting ß-Actin (0869100-CF, MP Biologicals, Irvine, CA, USA, 1:1500 dilution), p65 (CST8242, Cell Signaling Technology, Danvers, MA, USA, 1:1000 dilution), and p(Ser) PKC Substrate (CST2261, 1:1000 dilution) in blocking buffer at 4 °C overnight. The next day, membranes were washed three times for 5 min each with TBS containing 0.1% Tween 20 (TBS-T, Merck), incubated with anti-mouse (SA5-35521, Thermo Fisher Scientific) or anti-rabbit antibodies (A-21077, Thermo Fisher Scientific) and washed again. Finally, (phosphorylated) proteins were visualized with Odyssey Infrared Imaging Software and quantification was performed using Image Studio Lite (LI-COR, Lincoln, NE, USA). Uncropped images of respective Western blots are shown in Supplementary Information.

Clinical dataset

Raw data from patient-derived gene expression and phospho-proteomics data of the CPTAC-BRCA²¹ clinical dataset were obtained from the cBioPortal open-source database^67,68. Initial data analysis was done as described for the in vitro data above. The CPTAC-BRCA cohort was subdivided for primary analysis into patients with ER− (n = 39) and ER+ disease (n = 81). Pathway and transcriptional factor activities were estimated also per patient basis, using as an input the scaled and normalized gene expression values from cBioPortal. To this end, Z-score transformation was applied prior to per patient analysis, to uncover higher or lower activities within individual patients compared to the total reference cohort. Single sample gene set enrichment analysis (https://github.com/broadinstitute/ssGSEA2.0)⁶⁹ was applied on the scaled gene expression values using Z-score transformation as above. The script “ssgsea-gui.R” was applied with default settings (sample.norm.type: none, weight: 0.75, statistic: area.under.RES, nperm: 1000, output.score.type: NES and correl.type: z.score) and as gene-set library the included hallmark signatures file (version 7.5.0) in gmt format was selected (https://github.com/broadinstitute/ssGSEA2.0/tree/master/db/msigdb).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The RNA sequencing data of the clonal and bulk MCF7 and T47D cell line samples generated in this study have been deposited in the European Genome-Phenome Archive under the accession number EGAS00001007123. The data are available under controlled access due to the sensitive nature of human sequencing data, and access can be obtained by contacting the appropriate Data Access Committee listed in the study for each dataset. Access will be granted to commercial and non-commercial parties according to patient consent forms and data transfer agreements. We have an institutional process in place to deal with requests for data transfer. A response to requests for data access can be expected within 14 days. After access has been granted, the data is available for two years. The mass spectrometry phospho-proteomics data from clonal and bulk cell line samples have been deposited to the ProteomeXchange Consortium via the PRIDE⁷⁰ partner repository with the dataset identifier PXD040478. The uncropped Western blot images shown in Fig. 6d and Supplementary Fig. S6c are provided in the Source Data file with this paper. The remaining data are available within the article, Supplementary Information, Source Data file, Zenodo (https://doi.org/10.5281/zenodo.7780930) or github (Jasonmbg/BRCA_Clonal_Resist_Project and Jasonmbg/PublicData.CPTAC.BreastCancer.PhosphoP).

References

Siegel, R. L., Miller, K. D., Wagle, N. S. & Jemal, A. Cancer statistics, 2023. CA Cancer J. Clin. 73, 17–48 (2023).
Article PubMed Google Scholar
Loibl, S., Poortmans, P., Morrow, M., Denkert, C. & Curigliano, G. Breast cancer. Lancet 397, 1750–1769 (2021).
Article CAS PubMed Google Scholar
Inwald, E. C. et al. 4-IHC classification of breast cancer subtypes in a large cohort of a clinical cancer registry: use in clinical routine for therapeutic decisions and its effect on survival. Breast Cancer Res. Treat. 153, 647–658 (2015).
Article CAS PubMed PubMed Central Google Scholar
Dunnwald, L. K., Rossing, M. A. & Li, C. I. Hormone receptor status, tumor characteristics, and prognosis: a prospective cohort of breast cancer patients. Breast Cancer Res. 9, R6 (2007).
Article PubMed PubMed Central Google Scholar
Andrahennadi, S., Sami, A., Manna, M., Pauls, M. & Ahmed, S. Current landscape of targeted therapy in hormone receptor-positive and HER2-negative breast cancer. Curr. Oncol. 28, 1803–1822 (2021).
Article PubMed PubMed Central Google Scholar
SEER_Program. SEER 5-Year Relative Survival Rates, 2011–2017, https://seer.cancer.gov/statfacts/html/breast-subtypes.html (2022).
Davies, C. et al. Long-term effects of continuing adjuvant tamoxifen to 10 years versus stopping at 5 years after diagnosis of oestrogen receptor-positive breast cancer: ATLAS, a randomised trial. Lancet 381, 805–816 (2013).
Article CAS PubMed PubMed Central Google Scholar
Hanker, A. B., Sudhan, D. R. & Arteaga, C. L. Overcoming endocrine resistance in breast cancer. Cancer Cell 37, 496–513 (2020).
Article CAS PubMed PubMed Central Google Scholar
Borgoni, S. et al. Time-resolved profiling reveals ATF3 as a novel mediator of endocrine resistance in breast cancer. Cancers (Basel) 12, https://doi.org/10.3390/cancers12102918 (2020).
Soleimani Dodaran, M. et al. Candidate methylation sites associated with endocrine therapy resistance in ER+/HER2- breast cancer. BMC Cancer 20, 676 (2020).
Article CAS PubMed PubMed Central Google Scholar
Marusyk, A., Janiszewska, M. & Polyak, K. Intratumor heterogeneity: the rosetta stone of therapy resistance. Cancer Cell 37, 471–484 (2020).
Article CAS PubMed PubMed Central Google Scholar
Pasha, N. & Turner, N. C. Understanding and overcoming tumor heterogeneity in metastatic breast cancer treatment. Nat. Cancer 2, 680–692 (2021).
Article CAS PubMed Google Scholar
Yates, L. R. et al. Subclonal diversification of primary breast cancer revealed by multiregion sequencing. Nat. Med. 21, 751–759 (2015).
Article CAS PubMed PubMed Central Google Scholar
Casasent, A. K. et al. Multiclonal invasion in breast tumors identified by topographic single cell sequencing. Cell 172, 205–217.e212 (2018).
Article CAS PubMed PubMed Central Google Scholar
Hayford, C. E. et al. An in vitro model of tumor heterogeneity resolves genetic, epigenetic, and stochastic sources of cell state variability. PLoS Biol. 19, e3000797 (2021).
Article CAS PubMed PubMed Central Google Scholar
Pich, O. et al. The translational challenges of precision oncology. Cancer Cell 40, 458–478 (2022).
Article CAS PubMed Google Scholar
Gaudet, S., Spencer, S. L., Chen, W. W. & Sorger, P. K. Exploring the contextual sensitivity of factors that determine cell-to-cell variability in receptor-mediated apoptosis. PLoS Comput Biol. 8, e1002482 (2012).
Article CAS PubMed PubMed Central Google Scholar
Hinohara, K. et al. KDM5 histone demethylase activity links cellular transcriptomic heterogeneity to therapeutic resistance. Cancer Cell 34, 939–953.e939 (2018).
Article CAS PubMed PubMed Central Google Scholar
Shaw, A. T. et al. Resensitization to crizotinib by the lorlatinib ALK resistance mutation L1198F. N. Engl. J. Med. 374, 54–61 (2016).
Article CAS PubMed Google Scholar
Hata, A. N. et al. Tumor cells can follow distinct evolutionary paths to become resistant to epidermal growth factor receptor inhibition. Nat. Med. 22, 262–269 (2016).
Article CAS PubMed PubMed Central Google Scholar
Krug, K. et al. Proteogenomic landscape of breast cancer tumorigenesis and targeted therapy. Cell 183, 1436–1456.e1431 (2020).
Article CAS PubMed PubMed Central Google Scholar
Bhang, H. E. et al. Studying clonal dynamics in response to cancer therapy using high-complexity barcoding. Nat. Med. 21, 440–448 (2015).
Article CAS PubMed Google Scholar
Schmidt, M. et al. Detection and direct genomic sequencing of multiple rare unknown flanking DNA in highly complex samples. Hum. Gene Ther. 12, 743–749 (2001).
Article CAS PubMed Google Scholar
Oren, Y. et al. Cycling cancer persister cells arise from lineages with distinct programs. Nature 596, 576–582 (2021).
Article CAS PubMed PubMed Central Google Scholar
Schubert, M. et al. Perturbation-response genes reveal signaling footprints in cancer gene expression. Nat. Commun. 9, 20 (2018).
Article PubMed PubMed Central Google Scholar
Holland, C. H., Szalai, B. & Saez-Rodriguez, J. Transfer of regulatory knowledge from human to mouse for functional genomics analysis. Biochim. Biophys. Acta Gene Regul. Mech. 1863, 194431 (2020).
Article CAS PubMed Google Scholar
Li, S. et al. Endocrine-therapy-resistant ESR1 variants revealed by genomic characterization of breast-cancer-derived xenografts. Cell Rep. 4, 1116–1130 (2013).
Article CAS PubMed Google Scholar
Maia, A. et al. IFNbeta1 secreted by breast cancer cells undergoing chemotherapy reprograms stromal fibroblasts to support tumour growth after treatment. Mol. Oncol. 15, 1308–1329 (2021).
Article CAS PubMed PubMed Central Google Scholar
Bott, A. et al. miRNA-1246 induces pro-inflammatory responses in mesenchymal stem/stromal cells by regulating PKA and PP2A. Oncotarget 8, 43897–43914 (2017).
Article PubMed PubMed Central Google Scholar
Garcia-Alonso, L., Holland, C. H., Ibrahim, M. M., Turei, D. & Saez-Rodriguez, J. Benchmark and integration of resources for the estimation of human transcription factor activities. Genome Res. 29, 1363–1375 (2019).
Article CAS PubMed PubMed Central Google Scholar
Badia-i-Mompel, P. et al. decoupleR: ensemble of computational methods to infer biological activities from omics data. Bioinform Adv. 2, vbac016 (2022).
Article PubMed PubMed Central Google Scholar
Turei, D. et al. Integrated intra- and intercellular signaling knowledge for multicellular omics analysis. Mol. Syst. Biol. 17, e9923 (2021).
Article PubMed PubMed Central Google Scholar
Taniguchi, K. & Karin, M. NF-kappaB, inflammation, immunity and cancer: coming of age. Nat. Rev. Immunol. 18, 309–324 (2018).
Article CAS PubMed Google Scholar
Paramore, A. & Frantz, S. Bortezomib. Nat. Rev. Drug Discov. 2, 611–612 (2003).
Article CAS PubMed Google Scholar
Shapiro, D. J., Livezey, M., Yu, L., Zheng, X. & Andruska, N. Anticipatory UPR activation: a protective pathway and target in cancer. Trends Endocrinol. Metab. 27, 731–741 (2016).
Article CAS PubMed PubMed Central Google Scholar
Aftimos, P. et al. Genomic and transcriptomic analyses of breast cancer primaries and matched metastases in AURORA, the Breast International Group (BIG) molecular screening initiative. Cancer Discov. 11, 2796–2811 (2021).
Article CAS PubMed PubMed Central Google Scholar
Pan, H. et al. 20-year risks of breast-cancer recurrence after stopping endocrine therapy at 5 years. N. Engl. J. Med. 377, 1836–1846 (2017).
Article PubMed PubMed Central Google Scholar
Dagogo-Jack, I. & Shaw, A. T. Tumour heterogeneity and resistance to cancer therapies. Nat. Rev. Clin. Oncol. 15, 81–94 (2018).
Article CAS PubMed Google Scholar
Caswell, D. R. & Swanton, C. The role of tumour heterogeneity and clonal cooperativity in metastasis, immune evasion and clinical outcome. BMC Med. 15, 133 (2017).
Article PubMed PubMed Central Google Scholar
Andruska, N., Zheng, X., Yang, X., Helferich, W. G. & Shapiro, D. J. Anticipatory estrogen activation of the unfolded protein response is linked to cell proliferation and poor survival in estrogen receptor alpha-positive breast cancer. Oncogene 34, 3760–3769 (2015).
Article CAS PubMed Google Scholar
Assender, J. W. et al. Protein kinase C isoform expression as a predictor of disease outcome on endocrine therapy in breast cancer. J. Clin. Pathol. 60, 1216–1221 (2007).
Article CAS PubMed PubMed Central Google Scholar
Pan, Q. et al. Protein kinase C epsilon is a predictive biomarker of aggressive breast cancer and a validated target for RNA interference anticancer therapy. Cancer Res. 65, 8366–8371 (2005).
Article CAS PubMed Google Scholar
Albert, V. et al. Protein kinase C targeting of luminal (T-47D), luminal/HER2-positive (BT474), and triple negative (HCC1806) breast cancer cells in-vitro with AEB071 (Sotrastaurin) is efficient but mediated by subtype specific molecular effects. Arch. Gynecol. Obstet. 306, 1197–1210 (2022).
Article CAS PubMed PubMed Central Google Scholar
Wahjudi, L. W. et al. Integrating proteomics into precision oncology. Int J. Cancer 148, 1438–1451 (2021).
Article CAS PubMed Google Scholar
Kuhl, H. Pharmacology of estrogens and progestogens: influence of different routes of administration. Climacteric 8, 3–63 (2005).
Article CAS PubMed Google Scholar
R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. (2020).
RStudio: Integrated Development Environment for R} (RStudio, PBC., 2020).
Morgan, M. et al. ShortRead: a bioconductor package for input, quality assessment and exploration of high-throughput sequence data. Bioinformatics 25, 2607–2608 (2009).
Article CAS PubMed PubMed Central Google Scholar
Manasanch, E. E. & Orlowski, R. Z. Proteasome inhibitors in cancer therapy. Nat. Rev. Clin. Oncol. 14, 417–433 (2017).
Article CAS PubMed PubMed Central Google Scholar
Wagner, J. et al. Discovery of 3-(1H-indol-3-yl)-4-[2-(4-methylpiperazin-1-yl)quinazolin-4-yl]pyrrole-2,5-dione (AEB071), a potent and selective inhibitor of protein kinase C isotypes. J. Med. Chem. 52, 6193–6196 (2009).
Liao, Y., Smyth, G. K. & Shi, W. The R package Rsubread is easier, faster, cheaper and better for alignment and quantification of RNA sequencing reads. Nucleic Acids Res. 47, e47 (2019).
Article CAS PubMed PubMed Central Google Scholar
Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30, 923–930 (2014).
Article CAS PubMed Google Scholar
Huber, W. et al. Orchestrating high-throughput genomic analysis with Bioconductor. Nat. Methods 12, 115–121 (2015).
Article CAS PubMed PubMed Central Google Scholar
Robinson, M. D., McCarthy, D. J. & Smyth, G. K. K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139–140 (2010).
Article CAS PubMed Google Scholar
Law, C. W., Chen, Y., Shi, W. & Smyth, G. K. voom: precision weights unlock linear model analysis tools for RNA-seq read counts. Genome Biol. 15, R29 (2014).
Article PubMed PubMed Central Google Scholar
Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015).
Article PubMed PubMed Central Google Scholar
Wu, T. et al. clusterProfiler 4.0: a universal enrichment tool for interpreting omics data. Innovation 2, 100141 (2021).
CAS PubMed PubMed Central Google Scholar
Livak, K. J. & Schmittgen, T. D. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25, 402–408 (2001).
Article CAS PubMed Google Scholar
Wessel, D. & Flugge, U. I. A method for the quantitative recovery of protein in dilute solution in the presence of detergents and lipids. Anal. Biochem. 138, 141–143 (1984).
Article CAS PubMed Google Scholar
Ruprecht, B. et al. Comprehensive and reproducible phosphopeptide enrichment using iron immobilized metal ion affinity chromatography (Fe-IMAC) columns. Mol. Cell Proteom. 14, 205–215 (2015).
Article CAS Google Scholar
Rappsilber, J., Mann, M. & Ishihama, Y. Protocol for micro-purification, enrichment, pre-fractionation and storage of peptides for proteomics using StageTips. Nat. Protoc. 2, 1896–1906 (2007).
Article CAS PubMed Google Scholar
Bekker-Jensen, D. B. et al. Rapid and site-specific deep phosphoproteome profiling by data-independent acquisition without the need for spectral libraries. Nat. Commun. 11, 787 (2020).
Article CAS PubMed PubMed Central Google Scholar
Tyanova, S. et al. The Perseus computational platform for comprehensive analysis of (prote)omics data. Nat. Methods 13, 731–740 (2016).
Article CAS PubMed Google Scholar
Kim, H. J. et al. PhosR enables processing and functional analysis of phosphoproteomic data. Cell Rep. 34, 108771 (2021).
Article CAS PubMed Google Scholar
Huber, W., von Heydebreck, A., Sultmann, H., Poustka, A. & Vingron, M. Variance stabilization applied to microarray data calibration and to the quantification of differential expression. Bioinformatics 18, S96–104, (2002).
Article PubMed Google Scholar
Laemmli, U. K. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227, 680–685 (1970).
Article CAS PubMed Google Scholar
Cerami, E. et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2, 401–404 (2012).
Article PubMed Google Scholar
Gao, J. et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci. Signal 6, pl1 (2013).
Article PubMed PubMed Central Google Scholar
Barbie, D. A. et al. Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature 462, 108–112 (2009).
Article CAS PubMed PubMed Central Google Scholar
Perez-Riverol, Y. et al. The PRIDE database resources in 2022: a hub for mass spectrometry-based proteomics evidences. Nucleic Acids Res. 50, D543–D552 (2022).
Article CAS PubMed Google Scholar
Gu, Z. Complex heatmap visualization. iMeta 1, e43 (2022).
Article Google Scholar

Download references

Acknowledgements

We thank the DKFZ Single-Cell Open Lab (scOpenLab) and the Cellular Tools Core Facility, Solentim and Cytena for assistance with the spotting of single cells. We thank the DKFZ NGS and Mass Spectrometry Core Facilities for providing excellent services and technical support for the RNA-Seq, NGS of barcode region and (phospho-)proteomics analysis. This work was supported by the DKFZ International PhD Program and German-Israeli Helmholtz International Research School “Cancer-TRAX”, the European Union Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement (EpiPredict–642691), and the Helmholtz European Partnering Program for the cooperation between DKFZ and NHRF to build the Athens Comprehensive Cancer Center (ACCC).

Funding

Open Access funding enabled and organized by Projekt DEAL.

Author information

Authors and Affiliations

Division of Molecular Genome Analysis, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 580, 69120, Heidelberg, Germany
Lukas Beumers, Efstathios-Iason Vlachavas, Simone Borgoni, Luisa Schwarzmüller, Birgitta E. Michels, Emre Sofyali, Sara Burmester, Daniela Heiss, Heike Wilhelm & Stefan Wiemann
Faculty of Biosciences, University of Heidelberg, Im Neuenheimer Feld 234, 69120, Heidelberg, Germany
Lukas Beumers, Luisa Schwarzmüller & Stefan Wiemann
Division of Somatic Evolution and Early Detection, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 580, 69120, Heidelberg, Germany
Luca Penso-Dolfin & Angela Goncalves
Department of Immunology and Regenerative Biology, Weizmann Institute of Science, Rehovot, 76100, Israel
Yosef Yarden
Proteomics Core Facility, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 580, 69120, Heidelberg, Germany
Dominic Helm
Cellular Tools Core Facility, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 580, 69120, Heidelberg, Germany
Rainer Will

Authors

Lukas Beumers
View author publications
You can also search for this author in PubMed Google Scholar
Efstathios-Iason Vlachavas
View author publications
You can also search for this author in PubMed Google Scholar
Simone Borgoni
View author publications
You can also search for this author in PubMed Google Scholar
Luisa Schwarzmüller
View author publications
You can also search for this author in PubMed Google Scholar
Luca Penso-Dolfin
View author publications
You can also search for this author in PubMed Google Scholar
Birgitta E. Michels
View author publications
You can also search for this author in PubMed Google Scholar
Emre Sofyali
View author publications
You can also search for this author in PubMed Google Scholar
Sara Burmester
View author publications
You can also search for this author in PubMed Google Scholar
Daniela Heiss
View author publications
You can also search for this author in PubMed Google Scholar
Heike Wilhelm
View author publications
You can also search for this author in PubMed Google Scholar
Yosef Yarden
View author publications
You can also search for this author in PubMed Google Scholar
Dominic Helm
View author publications
You can also search for this author in PubMed Google Scholar
Rainer Will
View author publications
You can also search for this author in PubMed Google Scholar
Angela Goncalves
View author publications
You can also search for this author in PubMed Google Scholar
Stefan Wiemann
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

L.B., S.Bo. and S.W. conceived the study. L.B., S.Bo., L.S., E.S., S.Bu., Da.H., H.W., Y.Y., and R.W. designed and/or performed experiments. E.-I.V., B.M., L.P.-D., L.S., and A.G. designed and/or performed bioinformatic data analysis. L.B., S.Bo., L.P.-D., L.S., and E.-I.V. generated figures and tables. L.B., E-I.V., and S.W. wrote the manuscript. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Lukas Beumers or Stefan Wiemann.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

nr-reporting-summary

Rights and permissions

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 license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license 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 license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Beumers, L., Vlachavas, EI., Borgoni, S. et al. Clonal heterogeneity in ER+ breast cancer reveals the proteasome and PKC as potential therapeutic targets. npj Breast Cancer 9, 97 (2023). https://doi.org/10.1038/s41523-023-00604-4

Download citation

Received: 15 May 2023
Accepted: 17 November 2023
Published: 02 December 2023
DOI: https://doi.org/10.1038/s41523-023-00604-4