Breast cancer patient‐derived scaffolds as a tool to monitor chemotherapy responses in human tumor microenvironments

Abstract Breast cancer is a heterogeneous disease where the tumor microenvironment, including extracellular components, plays a crucial role in tumor progression, potentially modulating treatment response. Different approaches have been used to develop three‐dimensional models able to recapitulate the complexity of the extracellular matrix. Here, we use cell‐free patient‐derived scaffolds (PDSs) generated from breast cancer samples that were recellularized with cancer cell lines as an in vivo‐like culture system for drug testing. We show that PDS cultured MCF7 cancer cells increased their resistance against the front‐line chemotherapy drugs 5‐fluorouracil, doxorubicin and paclitaxel in comparison to traditional two‐dimensional cell cultures. The gene expression of the environmentally adapted cancer cells was modulated in different ways depending on the drug and the concentration used. High doses of doxorubicin reduced cancer stem cell features, whereas 5‐fluorouracil increased stemness and decreased the proliferative phenotype. By using PDSs repopulated with other breast cancer cell lines, T‐47D and MDA‐MB‐231, we observed both general and cell line specific drug responses. In summary, PDSs can be used to examine the extracellular matrix influence on cancer drug responses and for testing novel compounds in in vivo‐like microenvironments.


| INTRODUCTION
Breast cancer is a heterogeneous disease with variable phenotypes and genetic features, influencing prognostic, treatment decisions and patient outcome. This heterogeneity has been observed between patients, but also within a tumor, as a response to disease progression and treatments. Furthermore, different phenotypes related to spatial distribution of cells in tissues have been observed, represented by diverse histologic and biochemical properties including variable mutations and expression of biomarkers (Ellsworth et al., 2017). The extracellular matrix (ECM) plays an important role in structural and functional organization of the tumor. The mechanochemical stimulation and biological interaction between cells and ECM influence proliferation, survival, migration, invasiveness, and drug response (Frantz et al., 2010;Rijal & Li, 2016;Senthebane et al., 2017). The tumor ECM also promotes survival of the cancer stem cells (CSC), a population that displays stem cell properties, such as self-renewal, multipotent differentiation and a high tumor-initiating capacity (Ebben et al., 2010). The CSC phenotype has been linked to poor prognostic features, drug resistance and increased risk of disease recurrences (Akrap et al., 2016;Zhao, 2016).
Traditional two-dimensional (2D) culture systems used to study breast cancer lack the complexity of three-dimensional (3D) cell to cell contacts and cell to ECM interactions that occur in vivo. The use of scaffolds is an emerging approach to generate 3D systems able to reflect the in vivo tumor microenvironment characteristics. Scaffolds can be produced using biocompatible and biodegradable synthetic materials like polymers (Balachander et al., 2018;Feng et al., 2013;Ravikrishnan et al., 2016), or using materials derived from natural sources, such as fibrous proteins, biopolymers, or mice-derived tissue ECM (Dondajewska et al., 2018;Florczyk et al., 2013;Miyauchi et al., 2017;Rijal & Li, 2017). Nevertheless, the selection of the material is crucial due to several factors related to the scaffold composition and structure, such as stiffness, porosity, dimensionality and presence of adhesive proteins, all influencing the behaviors of the attached and growing cancer cells (Rijal & Li, 2016). With the aim to develop clinically relevant scaffolds, the use of human decellularized tissues, especially patient-derived tumor material is gaining popularity (Hoshiba, 2019;Piccoli et al., 2018;Pinto et al., 2017;Tian et al., 2018). Furthermore, tissues from breast cancer patients have previously been used as growth platforms to study the ECM influence on tumor progression (Jin et al., 2018;Liu et al., 2019).
Recently, we showed that patient-derived scaffolds (PDSs) produced from decellularized breast cancer tissue retained the natural ECM architecture and composition, representing the complexity of the tumor microenvironment (Landberg et al., 2020). Breast cancer cells that repopulated PDSs were influenced by the microenvironment and were clearly mimicking the in vivo-like cell-ECM interaction. Typical PDS-induced cellular features were increased fractions of CSC and epithelial-mesenchymal transition (EMT) cells, accompanied by a decreased proliferation.
In this study, we have analyzed the influence of the PDS microenvironments in response to therapeutic approaches.
Three mechanistically different chemotherapeutic drugs were tested, all of them used in the front-line treatment of breast cancer (Senkus et al., 2015): 5-fluorouracil (5-FU), a pyrimidine analog that is transformed into three active metabolites involved in thymidylate synthase inhibition and incorporated into DNA and RNA (Longley et al., 2003); doxorubicin (DOX) which exerts its effect through inhibition of topoisomerase II, though other mechanisms as intercalation into the DNA or free radical formation have been explored (Gewirtz, 1999); paclitaxel (PTX), a taxane with activity in the stabilization of microtubules during the mitosis process, stopping cell division (Weaver, 2014

| Tumor decellularization
Breast cancer primary samples were collected directly after surgery from the Clinical Pathology Diagnostic Unit at the Sahlgrenska University Hospital. Material from 15 patients was used in this study: 13 invasive ductal carcinomas, 1 invasive lobular carcinoma, and 1 in situ carcinoma.
The histopathological characteristics of the tumors included in this study are detailed in Table S1. The use of patient material for this project was approved by the Regional Research Ethics Committee (Regionala Etikprövningsnämnden) in Gothenburg (DNR: 515-12 and T972-18). All research was performed according to ethical guidelines and informed consent was obtained from all the participants in the study. The decellularization protocol for breast cancer PDSs followed our earlier published protocols (Landberg et al., 2020). In brief, tumors were washed with a lysis buffer composed of 0.1% sodium dodecyl sulfate (SDS), 0.02% Na-Azide, 5 mM 2H 2 O-Na 2 -EDTA, and 0.4 mM phenylmethylsulfonyl fluoride (Sigma-Aldrich) for 6 h followed by a rinse step in the same buffer without SDS for an additional 15 min. This process was repeated twice, and then, decellularized tumors were washed for 72 h in distilled water which was exchanged every 12 h to remove cell debris followed by 24 h wash in phosphate-buffered saline (PBS; Medicago). After decellularization ( Figure 1), PDSs were cut in 6 mm diameter pieces using a biopsy punch needle, and then sliced in 150 µm slices using a CM3050-S cryotome (Leica) to get several slices from the same PDS (from 5 up to 60, depending on the tissue). PDS slices were sterilized in peracetic acid 0.1% (Sigma-Aldrich) for 1 h at room temperature, followed by 24 h washing in PBS + 1% Antibiotic-Antimycotic (Thermo Fisher Scientific), at 37°C and gentle agitation at 175 rpm (Incu-Shaker TM 10L, Benchmark). PDSs were stored in a buffer containing PBS, 0.02% Na-Azide, and 5 mM EDTA at 4°C.

| Patient-derived scaffolds recellularization
Before recellularization, PDSs were soaked in cell culture media for 24 h to remove residual storage buffer. PDS slices were placed in a 48 wells-plate and seeded with 3 × 10 5 cells in 500 µl of cell line specific media supplemented with Antibiotic-Antimycotic (1%). After 24 h, the PDSs were transferred to a new plate with fresh media.
The PDSs were transferred to new media once or twice a week, up to 21 days of incubation.

| Drug treatment
The drugs tested in this article were purchased from Apoteket (Sweden). 5-FU (50 mg/ml, Accord) and DOX (2 mg/ml, Actavis) were dissolved in a saline solution, whereas PTX (6 mg/ml, Fresenius Kabi) was dissolved in 50% Kolliphor/50% ethanol (K/EtOH). Kolliphor EL was purchased from Sigma Aldrich to assess solvent toxicity.  F I G U R E 1 Schematic illustration of the patient-derived scaffolds (PDSs) production. (a) Workflow for generating PDSs describing the different steps from tumor collection to decellularization, cell culture, treatment and analyses carried out to study drug response. *Indirect measurement. (b) Pictures of breast tumor tissue, a decellularized PDS and a microscopy image of MCF7 cells cultured on PDS. In the contrast-phase image of a recellularized scaffold, round-shaped cells can be observed in the border of the PDS. Scale bar = 100 µm. 5-FU, 5-fluorouracil; DOX, doxorubicin; LDH, lactate dehydrogenase; PTX, paclitaxel; qPCR, quantitative polymerase chain reaction experiments, three slices derived from three different PDSs were used for every assay as biological replicates, and in each experiment the same PDS was represented in the treatments as well as in the controls.

| RNA extraction
Cells in PDS and cells grown in adherent conditions were harvested in 350 µl of RTL buffer (Qiagen) and stored at −80°C. PDSs were homogenized using a stainless steel bead in TissueLyzer II (Qiagen) for 2 × 5 min at 25 Hrz, and centrifuged at full speed for 3 min. RNA extraction was performed using the RNeasy Micro Kit including DNase treatment in a QIAcube machine (all from Qiagen). RNA concentration was measured by NanoDrop (Thermo Fisher Scientific).

| Gene expression analysis
Reverse transcription and quantitative polymerase chain reaction (qPCR) were performed similarly to our previous study (Landberg et al., 2020).
Complementary DNA synthesis was carried out with a GrandScript cDNA synthesis kit (TATAA Biocenter) in a T100 Thermal Cycler (BioRad) in 20 μl reaction mix at 25°C for 5 min, 42°C for 30 min, 85°C for 5 min followed by cooling to 4°C until subsequent analysis. RNA Spike II (TATAA Biocenter) was previously added to every sample as an RNA stability control. After that, all samples were diluted 1:5 or 1:6 with RNAse free water (Thermo Fisher Scientific). qPCR was performed on a CFX384 Touch Real-Time PCR Detection System (Bio-Rad) using 1X SYBR GrandMaster Mix (TATAA Biocenter), 400 nM of each primer (Table S2), and 2 μl diluted complementary DNA in a final reaction volume of 6 μl. The temperature profile was 95°C for 2 min followed by 35-50 cycles of amplification at 95°C for 5 s, 60°C for 20 s, and 70°C for 20 s and a melting curve analysis at 65°C to 95°C with 0.5°C/s increments.
Cycle of quantification values were determined by the second derivative maximum method with the CFX Manager Software version 3.1 (Bio-Rad).
Gene expression was normalized using reference genes identified with the NormFinder algorithm and expressed as relative quantities (log2) to the untreated 2D cells or to untreated PDSs, depending on the analysis.
Data preprocessing was performed using GenEx (MultiD). PCA plots and heatmaps were calculated using GenEx, and autoscaled data was represented. The radar charts illustrating the average of several genes were calculated using Excel 2016. All experiments were conducted in accordance with the Minimum Information for Publication of Quantitative Real-Time PCR Experiments guidelines (Bustin et al., 2009).

| Cytotoxicity assay
Cell death was assessed using lactate dehydrogenase (LDH) assay (Roche) on conditioned media from PDS cultures. At the same time of harvesting cells and PDSs for qPCR analysis, cell media was collected and stored at 4°C, following the manufacturer's recommendations.
A volume of 100 µl of media was added to a 96-wells plate together with 100 µl of the reaction mix. After 30 min of incubation absorbance was measured at 490 nm in a Multi-mode reader (Biotek), using Gen5 software (Biotek). Reference wavelength at 680 nm was subtracted, as well as background signal from the cell media. Data was represented as relative absorbance to the untreated samples.
For cell culture, 3DPS were washed in cell culture media, placed in 24 wells-plates and seeded with 3 × 10 5 MCF7 cells in supplemented DMEM. After 24 h, 3DPS were moved to a six well-plates with fresh media, and this process was repeated every 4 days to a total culturing time of 21 days. After that, 3DPS were treated with 5-FU and DOX at the 10X concentration as the PDS treatments. For RNA extraction, 3DPS were washed twice in cell media, lysed in 700 µl QIAzol (Qiagen) and homogenized for 2 × 2.5 min at 25 Hz. Automated isolation of total RNA from lysates was performed in a QIAcube machine (Qiagen) using a RNeasy Micro Kit (Qiagen) with QIAzol extraction directives. Reverse transcription and qPCR were performed as explained above.
Student's t test corrected by the Holm-Sidak method was used for comparing two groups and two-way analysis of variance with Tukey correction was used for comparing more than two groups. p < .05 were considered significant. All experiments were carried out in triplicates unless it was specified, using slices form three different patients in the case of the PDSs, and error bars represent standard deviation of the mean.  Data is represented relative to 2D cultures of the same cell line. Mean ± SD is shown, n = 3. Data from 2 to 3 replicates of each PDS were averaged before analysis. Significant differences between cells growing in PDSs and 2D, as wells as differences between the PDSs from different patients are stated (*p < .05, **p < .01, and ***p < .001). PDS, patient-derived scaffold LEIVA ET AL.

| PDSs increase cancer cell resistance against chemotherapy compounds
Next, MCF7 cells cultured on PDSs and in 2D were treated with the chemotherapy agents 5-FU, DOX, and PTX at the established IC 50 , a concentration that reduced proliferation by 50% in cells growing in 2D cultures ( Figure S1). After 48 h of treatment, cell media was collected and LDH assay was performed to determine the cell death ratio. Cells were harvested from the PDSs and RNA was extracted for gene expression analyses by qPCR as well as for an indirect measurement of the cell count by the total RNA yields ( Figure 1a).
For the 2D cultures, increased cell death and decreased total RNA yield were observed for DOX and PTX treatments (Figure 3a,b), whereas these treatments did not induce cell death or decrease the total RNA yield when administered to cells cultured in the PDS model. There was no effect on cell death or proliferation of the PTX solvent alone (kolliphor and ethanol -K/EtOH) in 2D or PDS cultures.
In the case of 5-FU, similar responses were observed in PDS and 2D growth models, with no cell death measured by LDH but with a decrease in the total RNA yield. These results indicated an increased resistance for two out of three chemotherapeutic compounds in the PDS system compared to 2D cultures. Furthermore, all three drugs triggered more pronounced gene expression changes in the cells cultured in 2D compared to PDSs, illustrated by a general upregulation of CSC associated markers and a decrease in proliferation genes in 2D cultures (Figure 3c). DOX and PTX treatments showed larger differences between the drug response in 2D and PDS cultures than 5-FU, supporting cell death and total RNA yield results.
Similar to the basal response to PDSs without treatment (Figure 2), diverse response patterns were observed between the EMT/differentiation markers when treated with different chemotherapy compounds. Taken together, these results show that gene expression modulation after drug treatments is significantly higher in 2D compared to PDS cultures, supporting a context-dependent drug response.

| Different doxorubicin concentrations led to cell phenotype selections in PDSs growing cells
We next analyzed the cellular response by gene expression analyses of MCF7 cells growing in PDSs exposed to increasing concentrations of DOX up to 100-fold the IC 50 defined in 2D cultures. In addition, quantifications of cell death and total RNA levels were used as complementary measurement of the drug effect.
For drug concentrations up to 10-fold the IC 50 value, there was a significant decrease in the expression of proliferation markers, whereas only minor effects were observed in the CSC-related genes relative to untreated PDS cultures (Figure 4a  F I G U R E 3 Cells cultured in PDSs present a higher resistance against chemotherapy drugs than 2D cultures. Measurement of (a) cell death, (b) total RNA yield and (c) gene expression in MCF7 cells grown in PDSs and 2D culture following treatment with 5-fluorouracil (5-FU), doxorubicin (DOX), and paclitaxel (PTX). Data is normalized to untreated 2D or PDSs, respectively. Mean ± SD is shown, n = 3. Significant differences between 2D and PDS or between untreated controls and treatments are represented (*p < .05, **p < .01, and ***p < .001). PDS, patient-derived scaffold 3.4 | Treatment with 5-fluorouracil enriches for a low proliferative CSC phenotype in cells cultured in PDSs  (d) DOX effect on the total RNA yield and LDH release, as surrogate measurements for cell count and cell death, respectively, quantified at the same drug concentrations. Relative quantities to untreated PDSs are represented. Mean ± SD is shown, n = 3. Significant differences to the untreated controls are stated (*p < .05, **p < .01, and ***p < .001). DOX, doxorubicin; LDH, lactate dehydrogenase; PDS, patient-derived scaffold

| Cells cultured in PDSs show a higher resistance against paclitaxel treatment
The effect of increased concentrations of PTX was also studied in PDS growing MCF7. For these experiments, we increased the drug concentration up to 1000-fold the IC 50 , but little to no effect was detected in the cells cultured in PDSs ( Figure 6). When the modulation of specific genes was analyzed in detail, only changes in a few marker genes were observed following treatment with PTX. Differing from the other treatments, there was an increased expression of the proliferation marker CCNA2 at all the doses analyzed ( Figure 6a).
Other changes in gene expression after PTX treatment were related to the upregulation of ABCG2 at various doses ( Figure 6b) and the upregulation of VIM at the highest concentration used (Figure 6c). As illustrated in the Principle Component Analysis (PCA) plots, the PDS cultured cells treated with PTX clustered together with untreated PDSs and PDSs treated with the lowest concentrations of DOX and 5-FU (Figure 7a), corroborating the lack of effect on gene expression modulation upon PTX treatment. Furthermore, no effect on the total RNA levels or cell death ratio was observed with this drug ( Figure 6d). To exclude PTX solvent toxicity, PTX effect was normalized to the effect of K/EtOH, since high solvent toxicity was detected using the 500X concentration ( Figure S3). Importantly, the minor changes in cell death ratio, total RNA levels and gene expression modulation in PDS cultured MCF7 cells upon PTX treatment contrast the pronounced effects observed in 2D cultures ( Figure 3). Thus, the presented data clearly indicates that PDS microenvironments profoundly affect the cancer cell response to PTX treatment.

| Repopulation of PDSs with different breast cancer cell lines presents a variety of expression profiles following treatment with chemotherapy compounds
To evaluate cell line-dependent responses to treatments, PDSs repopulated with MDA-MB-231 or T-47D cells were treated with DOX or 5-FU as previously described (Figure 1, Figure S1), but PTX was omitted due to the lack of effect in the previous analyses.

| DISCUSSION
This study describes the use of breast cancer PDSs as a physiologically relevant and in vivo-based 3D model to study the influence of the tumor microenvironment in chemotherapy response. Even though most cancer studies are based on analyses of tumor cells, the microenvironment has been suggested to substantially influence tumor progression and malignancy (Lu et al., 2012). Therefore, it is relevant to include the microenvironment in growth models assessing therapy responses to optimally mimic human-like conditions and relevant modeling of why some patients respond to a treatment whereas others are unaffected.
The PDS-platform used in this article was recently developed in our laboratory in an attempt to monitor how the microenvironment will influence cancer cells adapting to a decellurarized tumor scaffold.
Earlier published data supports that PDSs keep similar characteristics to in vivo tumors with links between the composition of the scaffolds and clinical properties (Landberg et al., 2020). Interestingly, PDSs also  Table S3. 5-FU, 5-fluorouracil; CSC, cancer stem cell; DOX, doxorubicin; EMT, epithelial-mesenchymal transition; PDS, patient-derived scaffold The effect of three front-line chemotherapeutic compounds on cancer cells cultured in PDSs was assessed using gene panels to determine coordinated drug treatment effects on several tumor biological processes. In line with previous reports, the results clearly indicated a higher drug resistance in cells cultured in PDSs in comparison to the 2D cultures, demonstrating that the growth of cells in a 3D conformation increases their robustness against chemotherapeutic drugs (Miyauchi et al., 2017;Rijal & Li, 2017). As examples, multicellular tumor spheroids showed increased resistance towards PTX and DOX treatments (Imamura et al., 2015;Reynolds et al., 2017) and Hakanson et al. (2011) observed less sensitivity to PTX in MCF7 cells cultured in 3D microwells or fibronectin matrices compared to 2D cultures. Further, breast cancer cells growing in silk scaffolds required a 40-fold higher DOX dose compared to 2D cell cultures to achieve comparable effects (Dondajewska et al., 2018) and cells cultured in scaffolds from decellularized breast cancer tissues have also shown less induced toxicity in response to 5-FU treatment (Liu et al., 2019). Other studies using colorectal cancer PDSs have shown similar results, with an increased resistance to 5-FU in comparison to 2D cultures (D'Angelo et al., 2020;Sensi et al., 2020). However, contradictory drug responses have also been reported when using different 3D models (Hakanson et al., 2011;Hongisto et al., 2013). For instance, Hongisto et al. (2013)  has been associated to increased resistance to DOX (Kajita et al., 2004;W. Li et al., 2011). Besides, ECM proteins may form a barrier preventing the drug availability to the cancer cells, and some protein constituents of our PDSs as fibronectins, collagens and laminins have earlier been linked to cell adhesion-mediated drug resistance (Landberg et al., 2020;Meads et al., 2009;Senthebane et al., 2017).
It is generally accepted that CSC have inherent resistance against chemotherapies in comparison to proliferative populations, resulting in CSC enrichment during treatments (Dean et al., 2005;X. Li et al., 2008). In fact, MCF7 derived CSC have shown more resistance to DOX than MCF7 cells at concentrations up to 1 µM (Yenigun et al., 2013). Furthermore, CSC enrichment has been ob-  (Liston & Davis, 2017). It can further be hypothesized that the reversion of the antiproliferative effect at the highest doses of DOX together with the reduction in CSC markers may be a consequence of DOX targeting the specific subpopulation with CSC phenotype. In addition, the increased expression of ABCG2 following treatment with DOX, could be in accordance with the multidrug resistance mechanism of this drug transporter (Stacy et al., 2013).
In the case of 5-FU treatment using PDSs, a drop in total RNA yield was observed for the lowest concentrations, which was accompanied by a decrease in proliferation as indicated by the downregulation of proliferation genes. To actually increase cell death, 5-and 50-fold higher doses of 5-FU were needed for MDA-MB-231 and MCF7 cells, respectively. This dual effect of 5-FU has previously been reported and seems to be related to its mechanism of action, though the mechanism which switches from cytostasis to apoptosis is not completely understood (Hernández-Vargas et al., 2006). Moreover, there was a slight upregulation of CSC markers after 5-FU treatment, which is in line with other reports linking CSC to 5-FU resistance (Dean et al., 2005;Lü et al., 2011;Saha et al., 2016 The aim of this study was to establish how a human-based microenvironment may affect drug responses. To determine the influence of the microenvironment on cancer cells, we used standardized breast cancer cell lines to repopulate the scaffolds, operating as a sensor and reporter for the cell-ECM interactions. However, the adaptability to the microenvironment will be dependent on specific cell line characteristics, as well as genetic abnormalities that might limit and potentially change the drug response (Gillet et al., 2011;Neve et al., 2006). In line with this, we observed general responses as well as varied drug effects in different cell lines cultured in the PDSs.
Larger studies including more patient samples need to be performed to identify robust adaptation patterns. Moreover, it could be possible to repopulate the PDSs with the patient primary cancer cells to create a more complete model system that can be used to monitor patient-specific responses to treatments. Nevertheless, due to the primary cells intrinsic heterogeneity, this approach would hide the tumor specific microenvironment influence achieved in this study when using a standardized cancer cell line as a reporter. Our model also presents the possibility to be complemented with additional stromal cells, such as macrophages and fibroblasts, which may play an important role in the signaling modulation between ECM and tumor cells and consequently in drug response (Dittmer & Leyh, 2015;Rijal & Li, 2016;Senthebane et al., 2017). However, the ECM has been produced and remodeled through dynamic interactions with all the cellular components of the tumor, including tumor and stromal cells, allowing PDSs to be used as adequate surrogates for the tumor microenvironment (Walker et al., 2018). In fact, exosomerelated proteins have been found in the PDSs proteomic composition (Landberg et al., 2020).
One more aspect to note in the PDS model is the variability between PDSs provided by different patients, as also observed in our previous studies (Landberg et al., 2020), and with other decellularized scaffolds (Pinto et al., 2017). This inter-scaffold variability, which may be influenced by the original tumor subtype, was more evident for the PDSs cultured using T-47D cells, suggesting that different cancer cell lines present different susceptibilities to the microenvironment. In addition, further studies should be performed to address if differences in the PDS composition or microstructure could be related to clinical parameters. The environment-related information provided by the PDSs may be a complementary diagnostic tool and potentially provide prognostic information regarding clinical behavior and outcome as well as be predictive of treatment response.

| CONCLUSION
This study provides an insight into the tumor microenvironment influence on chemotherapeutic response, using a physiologically re- Ståhlberg. All authors reviewed the manuscript.

DATA AVAILABILITY STATEMENT
The datasets are available from the corresponding author on reasonable request.