Lipidome-based Targeting of STAT3-driven Breast Cancer Cells Using Poly-L-glutamic acid-coated Layer-by-Layer Nanoparticles

Cell lines.

MDA-MB-231 and MDA-MB-468 cells (kindly provided by Myles Brown, Dana-Farber Cancer Institute) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum. SUM159PT cells (received from Kornelia Polyak, Dana-Farber Cancer Institute) were maintained in Ham’s F-12 medium supplemented with 5% heat inactivated fetal bovine serum, 10 mM HEPES (Gibco), 1 µg/mL hydrocortisone (Sigma) and 5 µg/mL insulin (Sigma). MCF-10A cells (ATCC, crl-10317) were stably engineered to express STAT3C (with a FLAG epitope) under an inducible promoter, and were maintained in DMEM/F12 containing 5% horse serum, 20 ng/mL EGF (Peprotech, Rocky Hill, NJ), 0.5 µg/mL hydrocortisone (Sigma), 100 ng/mL cholera toxin (Sigma) and 10 µg/mL insulin (Sigma). To induce STAT3C expression, cells were incubated with 2 µg/mL doxycycline (Sigma) for 24 hours prior to treatment with nanoparticles. Cells were maintained in a humidified incubator at 37℃ with 5% CO2 and passaged for less than three months after thawing. All cells were authenticated by short tandem repeat DNA profiling, and were routinely tested for the presence of Mycoplasma by PCR.

Immunoblot analyses.

Cells for immunoblot analyses were lysed on ice for 15 minutes in RIPA lysis buffer (Boston BioProducts) with phosphatase inhibitor cocktail (Cell Signaling Technology) and complete protease inhibitors (Roche). Immunoblots were probed with antibodies to phospho-STAT3 (Tyr705) (9131, Cell Signaling Technology), STAT3 (sc-482, Santa Cruz), FLAG (F1804, Sigma), tubulin (T5168, Sigma-Aldrich).

mRNA Expression Analyses (RT-PCR).

Total cellular RNA was isolated using Qiagen RNeasy Mini kits. RNA quality was evaluated on a NanoDrop 8000 spectrophotometer (Thermo Fisher Scientific), and reverse transcribed with TaqMan (Applied Biosystems) to generate cDNA. Quantitative polymerase chain reaction (qPCR) was performed in quadruplicate using Power SYBR master mix (Applied Biosystems) on a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems). Specificity of amplification was confirmed by melt curve analysis. Cycle threshold (Cτ) values for target genes were normalized to the endogenous reference gene GAPDH, and the fold change was determined by dividing the expression in each sample by that of the untreated control sample. Primer sequences (Supplementary Table 1) were designed from the UCSC genome browser reference mRNA sequences using Primer3.

Nanoparticles.

The LbL nanoparticle library was synthesized and characterized as previously described (7,8). Passive loading was employed for cisplatin (CDDP) encapsulation as described in the Supplementary Materials and Methods (9,10).

RNA interference.

Cells were transfected using Lipofectamine RNAiMAX with 10 nM of siRNA targeting STAT3 #2 or STAT3 #3 (D-003544–02-0010 and D-003544–03-0010, respectively; Dharmacon), or non-targeting siRNA control (sc-37007, Santa Cruz). Cells were transfected with siRNA for 48 hours, after which the media was changed to remove lipofectamine, and the cells were then incubated for an additional 24 hours (MDA-MB-231) or 72 hours (SUM159PT) prior to nanoparticle treatment.

Drug treatment.

MCF-10A cells were pre-incubated with 5 µM pyrimethamine (46706, Sigma Aldrich) or 2.5 µM ruxolitinib (ab141356, Abcam) for 4.5 hours. They were then induced with doxycycline (or vehicle control) for 24 hours, prior to immunoblot and qRT-PCR analyses or nanoparticle treatment. When evaluating the mode of cellular death in response to drug-loaded NP treatment, the cells were treated as previously described to induce or inhibit STAT3, then incubated for 1 hour with 3 µM ferrostatin-1 (SML0583, Sigma Aldrich) or 20 µM necrostatin-1 (N9037, Sigma Aldrich). The cells were then treated with the indicated concentration of CDDP-loaded NPs and analyzed as described below.

Mass spectrometry-based lipidomic profiling.

Prior to mass spectrometry analysis, MDA-MB-468 cells were transfected with siRNA targeting STAT3 #3 or non-targeting siRNA control as previously described, or treated with 10 µM pyrimethamine (or DMSO control) for 24 hours, and MCF-10A cells were treated for 24 hours with 2 µg/mL DOX (or DMSO control). Liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) was performed with five replicates per condition using an electrospray ionization (ESI) source on an Agilent 6430 or 6460 QQQ LC-MS/MS (Agilent Technologies), as described in detail in Supplementary Materials and Methods (11,12).

Nanoparticle treatment.

After the indicated treatments, cells were incubated with 100 ng/mL fluorescently-labeled nanoparticles for 24 hours prior to flow cytometric analysis or microscopic imaging. Following treatment with CDDP-loaded nanoparticles or free CDDP, the cells were washed with PBS, replenished with fresh media and incubated for a total of three days, after which viability was detected by flow cytometry.

Nanoparticle library screening.

MCF-10A cells were incubated with doxycycline (to induce STAT3C) or vehicle control for 24 hours, after which the membrane was stained red using Wheat Germ Agglutinin (W11262, Life Technologies) and nuclei were stained blue using Hoechst stain (4082S, Cell Signaling Technology). The stains were washed with phosphate-buffered saline (PBS) and the cells were treated with 100 ng/mL of different coating nanoparticles and imaged on CellObserver (Zeiss) every 30 minutes for a total of 8 hours. The membrane and nuclear binding were analyzed with custom pipelines (Supplemental Materials and Methods) created in CellProfiler 2.2.0 software (Broad Institute of Harvard and MIT).

Flow cytometry.

Cells were induced with doxycycline (MCF-10A) or transfected with siRNA (SUM159PT and MDA-MB-231), and treated with fluorescently-labeled non-loaded or CDDP-loaded nanoparticles as described above. After incubation with nanoparticles, the cells were washed with PBS, trypsinized and neutralized with corresponding media, and washed again with PBS.

Three dimensional organoid culture.

Cells were cultured for three to ten days in Nanoculture plates (MBL International) using the recommended culturing protocol. To preserve STAT3C induction, the MCF-10A cells were replenished with 2 µg/mL doxycycline (or DMSO control) every three days of culturing. Organoids were stained with 2 µM Lox1 hypoxia stain (#Lox1, MBL International) and treated with 100 ng/mL nanoparticles for 24 hours prior to confocal microscopy imaging using Zeiss LSM 880+ Airyscan microscope. Yellow-green (NP) and red (Lox1) fluorescence were quantitated with a custom macro (Supplemental Materials and Methods) written in ImageJ software (National Institutes of Health). Organoid images, Z-stack videos, and XYZ projection were prepared using ImageJ software, and video presentations of the NP distribution across three dimensions were prepared using Vision4D 3.1 software (arivis AG, Germany).

Deconvolution microscopy.

The preparation of samples is described in detail in the Supplementary Materials and Methods. Briefly, MCF-10A cells were seeded in microscope slide chamber with DOX- or DMSO-containing media for 24 hours, followed by PLE-NP treatment for an additional 2 or 24 hours (n=3). They were fixed and stained for nucleus and membrane using Hoechst stain (blue) and Wheat Germ Agglutinin (red), respectively. The cells were imaged on an Inverted Olympus X71 microscope and processed using OMX softWoRx software (Applied Precision/GE).

Gamma irradiation.

Cells were irradiated with a ¹³⁷Cs source as a single administration of the indicated dose. Relative viable cell number was determined after 1 to 3 days of incubation by measuring ATP-dependent bioluminescence (Cell Titer-Glo; Promega) on a Luminoskan Ascent luminometer. For NP-cell binding experiments, the cells were treated for 24 hours with 100 ng/mL PLE-NP after irradiation, after which NP-cell association was quantified with flow cytometry, as described above.

Gene set enrichment analysis.

GSE5460 breast cancer and GSE3526 normal tissue microarray datasets (13) were downloaded from the Gene Expression Omnibus. Gene-set enrichment analysis was performed using xapps.gsea from the Broad Institute of MIT and Harvard (http://software.broadinstitute.org), based on STAT3 gene expression signatures (14).

Statistical Analyses.

Results are presented as mean±SE unless otherwise indicated. Two-tailed Student t tests for paired samples were performed with GraphPad Prism 8.4 software (La Jolla, CA). Values of p<0.05 were considered significant (*, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001).

STAT3 transcriptional activity modulates cellular lipid profile in breast cancer cells.

STAT3 is aberrantly activated in the majority of breast cancers and represents a crucial mediator of the pathogenesis of this disease. Whereas STAT3 protein is ubiquitously expressed in normal tissue, its transcriptional activity differs greatly between normal and breast cancer tissue (Figure 1A), indicating that STAT3-directed treatment approaches would be associated with a favorable therapeutic index. Since STAT3 is difficult to target directly, we investigated whether metabolic changes driven by constitutive activation of STAT3 could provide a vulnerability in breast cancer cells that can be exploited therapeutically. Recognizing that malignant transformation is associated with changes in a variety of cellular metabolites, we examined how STAT3 qualitatively and quantitatively affects lipid distribution in mammary epithelial cells using two complementary systems. First, we used non-transformed MCF-10A cells in which an activated form of STAT3 can be expressed with an inducible doxycycline (DOX)-responsive promoter (Figure 1A), and is sufficient to cause tumorigenic properties (15). Induction of the promoter increased STAT3 protein expression and its required tyrosine phosphorylation, followed by enhanced mRNA expression of target genes including SOCS3, BCL3, BCLAF1, and a positive autoregulation of its own expression (Figure 1B). We also used a triple-negative breast cancer cell line, MDA-MB-468, characterized by constitutive STAT3 tyrosine phosphorylation, in which we inhibited STAT3 activity genetically (with RNA interference) or pharmacologically (with the small molecule inhibitor pyrimethamine (PYR) (Figure 1C) (16)). We then performed liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS)-based profiling to evaluate a total of 220 lipid molecules (Supplementary table 2). We found several classes of lipid molecules that were significantly changed by activated STAT3 (Figures 1D and E). Four distinct but similar N-acyl-taurine (NAT) metabolites were decreased in MDA-MB-468 cells with activated STAT3. Similarly, the presence of N-arachidonoyl taurine (C20:4 NAT), which has been reported to induce apoptosis of prostate cancer cells (17,18), was diminished when STAT3C was induced in MCF-10A cells. Two other taurine derivates, N-palmitoyl (C16:0) and N-stearoyl (C18:0) taurine, were strongly reduced when STAT3 is constitutively active in MDA-MB-468 cells, and showed an analogous trend with STAT3C induction. Interestingly, several lipid metabolites that were negatively modulated by STAT3, including taurine, were conjugated with arachidonic acid (AA, C20:4), the precursor for eicosanoid synthesis (Figures 1D and E). Both taurine and AA metabolites are actively involved in plasma membrane remodeling (19,20), providing a potential molecular distinction of STAT3-driven cancer cells.

LbL nanoparticle library screen to identify STAT3-targeting drug carriers.

Given these prominent changes in lipid composition driven by activated STAT3, we investigated targeting STAT3-driven cancer cells using novel nanoparticles that can exploit these differences. Layer-by-layer (LbL) nanoparticles (NP) consist of a carboxy-modified latex colloidal core (CML), surrounded by multiple ultrathin alternatively charged layers. These layers are composed of cationic poly-L-arginine (PLR) with polyanions as the terminal layer, containing either sulfated or carboxylated functional groups. The average hydrodynamic diameter of all NPs range from 100 to 155 nm (7,8). Different charge layers allow packaging of a variety of therapeutic molecules with controlled release of the active substance (21,22). Most importantly, the surface layer is responsible for the NP interaction with the plasma membrane, which enables the design of LbL NPs with tumor-targeting properties (7,23). Therefore, we screened a library of 12 LbL NPs differing in their external layer to determine the surface composition that allowed the highest binding to cells characterized by the STAT3-induced lipid profile. MCF-10A cells with and without STAT3C induction were treated with distinct NPs, after which their membrane and nuclear co-localization were recorded every 30 minutes for a total of 8 hours (Figure 2A, Supplementary Figure 1 and Supplementary Figure 2). We found that the poly-L-glutamic acid (PLE) coated NPs showed significantly greater binding to STAT3-driven cells compared to non-transformed counterparts (Figures 2A–C). This did not reflect a non-specific effect of neoplastic transformation, as LbL NPs coated with other modifications showed no difference in cell binding affinity (Figure 2A and Supplementary Figure 1). The differential PLE-NP-cell interaction was detected within the plasma membrane, and co-localization with nuclei was minimal for all investigated NPs in both conditions (Supplementary Figure 2). To further characterize the PLE-NP interactions with MCF10A cells with and without STAT3C induction we performed deconvolution microscopy (Figure 2D and Supplementary Figure 3). Whereas PLE-NPs associated to a low extent with non-transformed MCF-10A cells, induction of STAT3 led to recruitment of considerably greater numbers of these NPs (Figure 2D and Supplementary Figure 3), with targeting affinity being larger than projected from the initial screening data. This variability may have arisen from differences in cell culture density, as well as alterations in quantitation from single cell analysis compared with mean cell binding over a population (as in Figures 2A, 2B, and 2C). To address these quantitative differences in cellular binding, we further conducted flow cytometric analyses. As PLE-NPs have previously been shown to accumulate in tumor tissue in vivo, while having attenuated distribution in vital organs (7), we evaluated targeting breast cancer cells based on their abnormal STAT3 activity using PLE-coated nanoparticles.

PLE-NPs preferentially target STAT3-driven breast cancer cells.

To quantify the differential cell binding of PLE-NPs in breast cancer cells, we first used MCF-10A cells with doxycycline-inducible expression of STAT3C (Figure 3A). Using quantitative flow cytometry, we found that these nanoparticles preferentially associate with the cells with activated STAT3, displaying 50% greater cell-associated nanoparticle fluorescence compared to their non-transformed counterparts (Figure 3B). This effect was specific to PLE-NPs, as NPs with other modifications, such as dextran sulfate (DXS), did not reveal a similar effect (Figure 3B). To independently assess whether these differences were STAT3-dependent, we inhibited STAT3 phosphorylation with the Jak kinase inhibitor ruxolitinib (RUX) or STAT3 transcriptional activity with pyrimethamine (Figure 3C). Both treatments led to a 50% reduction in cell-associated nanoparticle fluorescence.

Since STAT3 is activated under basal conditions in most TNBCs, we investigated whether PLE-NPs could be used as a STAT3-targeted approach for this subset of breast cancers. We used the TNBC cell lines SUM159PT and MDA-MB-231 cells, as they both show constitutive tyrosine phosphorylation of STAT3 (Figure 3D). Furthermore, these cells are dependent on STAT3 for long term survival, as it was not possible to derive STAT3 null clones of these lines using CRISPR/Cas9. Therefore, we performed RNA interference-mediated silencing in these systems to determine STAT3-dependent interactions with NPs (Figures 3E and F). We found that the TNBC cell lines MDA-MB-231 and SUM159PT showed rapid binding of PLE-NPs, which was attenuated by depletion of STAT3 using two different siRNAs. This mechanism was specific for PLE-NPs, as DXS-coated NPs displayed either no difference (or showed increased accumulation) on cells lacking activated STAT3 (Figures 3G and H).

PLE-NP penetrate three-dimensional mammary cell organoids.

To better model behavior in tumor systems, we characterized PLE-NP accumulation in three-dimensional MCF-10A cell organoids, which more closely recapitulate the architecture of a tumor in vivo (24,25). STAT3 activation promoted cellular growth and proliferation, resulting in the formation of a markedly greater number of large colonies (Figure 4A). STAT3-transformed organoids also grew at much higher cellular density, which further resulted in greater hypoxia within the organoids (Figures 4B–D). To determine the penetration and distribution of NP within the organoid structure, we imaged the NP fluorescence across three dimensions (Figure 4C). In addition to occupying the surface of the cellular structures, PLE-NPs were clearly able to penetrate deep inside of the organoids, as seen in Z-stack and 3D organoid videos (Supplemental Figures 4 and 5) and XYZ organoid projection (Figure 4E). Furthermore, we quantified the NP fluorescence across three dimensions and found that the mean organoid fluorescence was substantially greater in organoids with activated STAT3 than in those without (Figures 4C and F). Recognizing that more voluminous organoids have greater surface area in direct contact with the NP-containing media, we also assessed the mean fluorescence per unit volume of the organoid. Despite the substantially higher cellular density of STAT3-transformed organoids, which might have been expected to impede the penetration of NPs, these organoids showed greater quantitative NP fluorescence per organoid (by 40%) as well as per organoid volume (by 73%) (Figure 4F). Thus, even in tightly packed three-dimensional organoids, PLE-NPs show preferential penetration and intra-organoid accumulation in the presence of activated STAT3.

STAT3-driven breast cancer cells show preferential sensitivity to cisplatin delivered via PLE-NPs.

To further evaluate the translational utility of these findings, we tested the effect of PLE-NPs incorporating the cytotoxic agent cisplatin (CDDP), as platinum agents are clinically used in TNBC treatment (26,27). Induction of STAT3 rendered MCF-10A cells more vulnerable to CDDP-loaded PLE-NPs and significantly increased cell death, whereas MCF-10A cells without activated STAT3 showed greater viability after such treatment (Figure 5A and Supplemental Figure 6). Similarly, the TNBC cell line SUM159PT showed increased apoptotic sensitivity to CDDP delivered via PLE-NPs, and this effect was attenuated when STAT3 was depleted with siRNA (Figure 5B and Supplemental Figures 7A and B). This did not reflect a general increased sensitivity to cell killing in the cells with activated STAT3, as these cells were overall less sensitive or generally resistant to free CDDP treatment (Supplementary Figure 6 and Figures 5A and B). Similarly, CDDP-loaded NPs that lacked LbL coatings or were coated with non-targeting DXS showed less cytotoxic effect on the cells with activated STAT3 (Figure 5A). Triple-negative SUM159PT cells displayed particular resistance to non-targeted forms of CDDP, as each of the tested conditions induced little apoptotic effect (Figure 5B), even with both prolonged incubation time and high treatment concentrations (Supplemental Figure 7A–D). Overall, these data suggest that drug-loaded PLE-NPs could be an effective targeted therapy for STAT3-driven breast cancers, including TNBC, while minimizing toxic effects to normal cells characterized by low and transient STAT3 activation.

Since cell death induced by CDDP-loaded NPs leads to early positivity of annexin V staining (without loss of membrane integrity), cell death induced by these agents is clearly, at least in part, mediated by apoptosis. However, given the rapid cell death induced by these NPs, as well as the possible clinical implications of the particular mode of cell death induced, we examined this aspect in more detail. STAT3 function has been linked with both ferroptosis and necroptosis (28,29). Therefore, we tested the effects of ferroptosis and necroptosis inhibitors (ferrostatin-1 and necrostatin-1, respectively) on cell death induced by CDDP-loaded PLE-NP. We did not observe any significant difference in cell killing in any system when incubating the cells with these inhibitors prior to CDDP-loaded NP treatment (Supplemental Figure 8). While it is possible that other modes of cell death may contribute to killing mediated by CDDP-loaded PLE-NPs, it appears that apoptosis is a prominent component of the mechanism.

Gamma radiation increases targeting properties of PLE-NPs.

Gamma radiation is commonly used in the treatment of breast cancer, including TNBC, and is known to alter the properties of the plasma membrane (30). Therefore, we tested whether radiation could impact PLE-NP-cell interaction in a STAT3-dependent manner. To exclude effects of widespread cell death, we first assessed viable cell number of STAT3-transformed and untransformed MCF-10A cells treated over a range of radiation doses (Supplementary Figure 9). We found that doses below 6 Gy were effective at avoiding a complete proliferation arrest. Next, we evaluated the effect and dose dependency of acute gamma irradiation on PLE-NP cell binding (Figure 5C). Interestingly, gamma radiation reduced the association of PLE-NPs with non-transformed MCF-10A cells in a dose dependent manner. However, binding to STAT3-transformed MCF-10A cells was minimally affected. Consequently, the difference in cell-associated PLE-NP fluorescence between non-transformed and STAT3-transformed cells increased from approximately 40% to 80% following 5 Gy irradiation (Figure 5C). These findings suggest a potential synergistic effect between radiation and nanoparticle-based targeting of breast cancer cells.