Hypoxia Attenuates Trastuzumab Uptake and Trastuzumab-Emtansine (T-DM1) Cytotoxicity through Redistribution of Phosphorylated Caveolin-1

This article is featured in Highlights of This Issue, p. 515

The HER2 gene is amplified in approximately 15% of all patients with breast cancer and is characterized by a relatively aggressive phenotype (1–4). However, a growing arsenal of HER2-targeting drugs has significantly improved the outcome of this patient subgroup (5). Trastuzumab-emtansine (T-DM1) is a HER2-targeted antibody–drug conjugate (ADC) where the cytotoxic microtubule-inhibitory agent DM1 is conjugated to trastuzumab antibody via a stable thioether linker (6). T-DM1 was approved as a single agent for the treatment of patients with locally advanced or metastatic HER2⁺ breast cancer by the FDA in 2013 (7), and recently showed improved survival as an adjuvant in patients with residual invasive disease after neoadjuvant therapy (8). Moreover, T-DM1 has recently shown promising results in HER2⁺ lung cancer (9). However, despite its significant clinical efficacy, intrinsic and acquired resistance is a major limiting factor for the therapeutic efficacy of T-DM1 (10–16).

Malignant tumors display regions of severe hypoxia that is associated with resistance to conventional oncologic treatments (chemo- and radiotherapy; refs. 17–19). A far less-explored area is how hypoxia may contribute to tumor cell escape from targeted therapies, such as T-DM1. HER2 is known to heterodimerize with EGFR, and hypoxia was shown to induce constitutive EGFR receptor signaling by delayed sorting to and deactivation in the endolysosomal compartment (20). Moreover, hypoxia may augment ligand-independent EGFR signaling by increased dimerization and prolonged activation in caveolin-1 (CAV1)–associated membrane domains, further leading to enhanced tumor cell proliferation and invasiveness (21). Hypoxia was shown to downregulate membrane protein internalization through a mechanism that involved CAV1 (22), and other studies suggest that HER2 homodimers codistribute with cholesterol-rich membrane raft domains and an involvement of caveolae-dependent mechanisms in the regulation of HER2 trafficking (23–28). Importantly, the possible link between hypoxia, HER2 internalization, and T-DM1 resistance remains unexplored.

Here, we were interested in investigating how adaptive responses to tumor hypoxia conditions may regulate T-DM1 treatment efficacy with potential implications in the management of HER2⁺ breast cancer.

HER2⁺ breast cancer cell lines (SKBR3 and BT474) were newly purchased from the ATCC and routinely cultured in HyClone McCoy 5a and DMEM (GE Healthcare Life Sciences), respectively, supplemented with 10% FBS, 2 mmol/L l-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin (Sigma-Aldrich; growth medium). HeLa cells were purchased from the ATCC and routinely cultured in DMEM supplemented with 10% FBS, 2 mmol/L l-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin (PEST). For cell authentication, the ATCC uses morphology, karyotyping, and PCR-based approaches to confirm the identity of cell lines and to rule out both intra- and interspecies contamination [cytochrome C oxidase I (COI) analysis and short tandem repeat profiling, respectively]. All cells were grown in a humidified 5% CO₂ incubator at 37°C and regularly (at least once per month) tested for mycoplasma by DAPI staining and confocal fluorescence microscopy. For all hypoxia-related experiments, cells were grown in hypoxia preconditioned media in a humidified Sci-tive NN Hypoxia workstation (Ruskinn Technology Ltd.) set at 5% CO₂, 1% O₂, and 37°C.

The patients are participants of The Breast Cancer and Blood Study (BC-Blood Study), which is an ongoing population-based study at the Skåne University Hospital in Lund, Sweden, since 2002 (29). The present tumor microarray (TMA) cohort is based on consecutive patients included in the study between November 2005 and June 2012. Among 789 included patients, 731 were diagnosed with invasive breast cancer. Patients with preoperative treatment (n = 20), distant metastasis within 0.3 years from inclusion (n = 6), and patients with carcinoma in situ (n = 32) were excluded from the study. In total, tumors from 635 patients were evaluated for CAV1 expression by IHC (see below). The study was approved by the Lund University Ethics Committee (Dnr LU75-02, LU37-08, LU658-09, LU58-12, LU379-12, LU227-13, LU277-15, and LU458-15) and performed according to the ethical permit guidelines. All patients signed written informed consent.

HeLa CAV1 knockdown (KD) and scrambled control cells, as previously described (22), were transfected with HER2 as per the manufacturer's instructions. Briefly, cells were transfected with 10 μg of HER2 WT plasmid DNA (#16257, Addgene) using lipofectamine 2000 (Thermo Fisher Scientific) transfection cocktail. To obtain HeLa CAV1 KD and scrambled control cells expressing equal levels of HER2, cells were sorted for HER2 expression by FACS. Following incubation with trastuzumab (50 μg/mL; Roche) at 4°C for 30 minutes, cells were washed and incubated with AlexaFluor488-conjugated anti-human antibody (25 μg/mL; #A11013, Thermo Fisher Scientific) in serum-free DMEM at 4°C for another 45 minutes. Cells were washed and resuspended in PBS containing 1% BSA at a cell density of 1 × 10⁶ cells/mL and sorted by FACSAria IIu (BD Biosciences). Sorted cells were collected in selection media (3.5 μg/mL puromycin, 2 mg/mL G418, 1% PEST, and 1% l-glutamine) initially supplemented with 50% FBS for cell recovery, and then with 10% FBS for further use.

For transient, siRNA-mediated CAV1 KD, SKBR3 cells were transfected with 25 nmol/L CAV1 siRNA ON-TARGET plus SMARTpool (L-003467-00-0005/L-003467-00-0010/L-003467-00-0020/L-003467-00-0050, Dharmacon) using DharmaFECT transfection solution. Cells were cultured for 48 hours without media change before further use. For generation of stable pCAV1 mutants, SKBR3 cells were transfected with 2 μg of wild-type CAV1-RFP (WT-CAV1-RFP, control), or a CAV1 tyrosine phosphorylation mimic (Y14D-CAV1-RFP), or phosphorylation-deficient CAV1 (Y14F-CAV1-RFP) plasmids, as previously described (30), using lipofectamine 2000 (Thermo Fisher Scientific) transfection cocktail. Cells were cultured for 24 hours without media change before selection in growth medium supplemented with 50 μg/mL kanamycin.

Normoxic or hypoxic SKBR3, BT474, MDAMB468, and HeLa cells, or the various SKBR3 and HeLa cell transfectants, were treated with T-DM1 (Roche AB), trastuzumab, DM1, MCC-DM1, or LMCC-DM1 with a range of concentrations, as indicated in the respective figure legend. Alternatively, SKBR3 cells were treated with mouse anti-HER2 antibody (1:100; clone CB11, Thermo Fisher Scientific) complexed with anti-mouse IgG Fc-monomethyl auristatin F (αMFc-CL-MMAF) secondary ADC (#AM-102-AF, Moradec LLC) or with paclitaxel-albumin nanoparticles (Abraxane), at the indicated concentrations. Following an incubation for 48 to 72 hours at normoxia or hypoxia, MTS reagent (CellTiter 96 AQ_ueous One Solution Cell Proliferation Assay Kit, #G3580, Promega) was added for 3 hours, and the absorbance was measured at 490 nm using VERSA_max tunable microplate reader with SoftMax Pro Software. The cell viability of the compounds was expressed as relative IC₅₀ after normalizing to appropriate controls (only trastuzumab, PBS, or DMSO). Control absorbance values subtracted from the respective baseline absorbance values (medium + PBS or DMSO) were defined as 100% cell viability (100% value), and cells treated with the highest concentration of test compound were defined as 0% cell viability (0% value) control. Relative IC₅₀ values were calculated from Log10 versus normalized response curves (variable slope) equation, generated using GraphPad Prism software v6 (GraphPad Software Inc.).

The Caspase Glo 3/7 cell viability assay was performed as per the manufacturer's instructions. Briefly, normoxic and hypoxic SKBR3 cells were treated with trastuzumab or T-DM1, as indicated, for 72 hours, and then transferred to room temperature (RT) before addition of 100 μL of Caspase-GloR 3/7 reagent (#G8090, Promega) for 2 hours before measuring luminescence using a FLUOstar OPTIMA (BMG LABTECH).

Normoxic or hypoxic cells were lysed with nondenaturing lysis buffer [20 mmol/L Tris-HCl, pH 7.4, 137 mmol/L NaCl, 2 mmol/L EDTA, 10% Glycerol, 1% Triton X-100, protease inhibitors (add fresh)] for 20 minutes at 4°C. Cell lysate was clarified by centrifugation at 18,000 × g for 10 minutes at 4°C, and the soluble supernatant was used for further analysis. For immunoprecipitation, cell lysates were swirled with anti-HER2 (#134182, Abcam) or anti-CAV1 antibody (#2941, Abcam) at 4°C overnight. The antibody–antigen solution was mixed with Protein G–conjugated Dynabeads (#10007D, Thermo Fisher Scientific) for 3 hours at 4°C followed by extensive washing using a DynaMag-2 magnetic separator (Thermo Fisher Scientific). Protein G–conjugated Dynabeads and an isotype-matched IgG antibody were used as negative control. Bound proteins were eluted according to the manufacturer's recommendation, mixed with NuPAGE LDS Sample Buffer 4 × (Thermo Fisher Scientific), and heated at 80°C for 10 minutes. Equal amount of proteins was loaded and separated in a NuPage 4–12% Bis-Tris gel (Thermo Fisher Scientific) at reducing conditions, and then transferred onto a polyvinylidene fluoride membrane (Immobilon-FL, Merck KGaA), followed by blocking in TBS 0.05% Tween 20 (TBST) containing 3% skim milk at RT for 1 hour. To probe for CAV1 and HER2, the membrane was incubated with the following antibodies in TBST containing 3% BSA overnight at 4°C: Rabbit anti-CAV1 (1:4,000; #ab2910, Abcam), mouse anti-HER2 (1:1,000; #OP39 Calbiochem, Sigma-Aldrich), and anti–β-actin (1:5,000; #ab8227, Abcam). After washing, the membrane was incubated with horseradish peroxidase–conjugated anti-rabbit (1:10,000; #7074, Cell Signaling Technology) or anti-mouse IgG (1:10,000; #A9044, Sigma-Aldrich) antibodies. Protein bands were visualized by Pierce Enhanced Chemiluminescence Western Blotting Substrate (Thermo Fisher Scientific).

Normoxic or hypoxic SKBR3 and BT474 cells were incubated with trastuzumab (40 μg/mL) for 30 minutes at 4°C (surface) or for the indicated time periods at 37°C (internalization). To visualize only internalized trastuzumab, cell surface–bound trastuzumab was removed with detachment solution (2 mol/L Urea/50 mmol/L Glycine/150 mmol/L NaCl, pH 2.4, 3 × 5 minutes). Subsequently, cells were fixed with 4% PFA for 10 minutes and permeabilized with 0.5% saponin for 15 minutes at RT, and then incubated with AlexaFluor488-conjugated goat anti-human secondary antibody (25 μg/mL) for 1 hour. Nuclei were counterstained with Hoechst 33342 (1:20,000; #1399, Thermo Fisher Scientific) for 10 minutes at RT, and cells were analyzed using Zeiss LSM 710 confocal scanning equipment with Plan-Apochromat 20x/0.8 or a C-Apochromat 63x/1.2 W Korr objective and Zen software (Carl Zeiss). Super resolution imaging details were acquired using the Airyscan detector system (Carl Zeiss).

TMAs were constructed as previously described (29), with duplicate cores of 1.0 mm from each primary tumor using a semiautomated tissue array instrument (Beecher Instruments). For IHC, 4 μm TMA sections were incubated at 60°C for 2 hours, deparaffinized, and then hydrated and pretreated with automatic PT-link system [DAKO, EnVision FLEX + Mouse (LINKER) # K8021] to unmask the epitopes. Sections were then stained for CAV1 with a primary rabbit polyclonal anti-CAV1 antibody (1:1,000; ab2910, Abcam) in Flex TRS low bluffer (pH 6.1) at 96°C for 20 minutes on Indalo P-block and labeled with Polymer EnVision FLEX/HRP to visualize CAV1, and then counterstained with hematoxylin (EnVision FLEX Hematoxylin). CAV1 was scored according to the intensity of cytoplasmic staining of invasive breast cancer cells across the two tumor cores for each patient. If at least 20% of the invasive cells were stained, the intensity was assigned 1 (weak staining), 2 (moderate), or 3 (strong). If less than 20% of the tumor cells were stained, the tumor was assessed as 0 (no staining). Scoring was performed by two independent readers (M. Barbachowska and V. Indira Chandran), and in case of disagreement, a senior evaluator was consulted (B. Nodin), and consensus was reached. All evaluators were blinded to data pertaining to the tumor samples. Nine patients had bilateral invasive breast cancer. For 5 of these patients, CAV1 expression was available for both tumors, and in only one case, CAV1 intensity differed and the value from the side with the highest intensity was used. The tumor characteristics from the evaluated tumor were used in all analyses where tumor characteristics were included.

Normoxic or hypoxic SKBR3 cells were incubated with trastuzumab (40 μg/mL) for 30 minutes on ice (surface) or 2 hours at 37°C (internalized). For quantifying internalized trastuzumab, cell surface–bound trastuzumab was removed with detachment solution (2 mol/L Urea/50 mmol/L Glycine/150 mmol/L NaCl, pH 2.4, 3 × 5 minutes), followed by blocking of residual surface trastuzumab with unlabeled goat anti-human antibody (50 μg/mL) for 30 minutes. Subsequently, cells were detached, fixed with 4% PFA for 10 minutes, and permeabilized with 0.5% Saponin for 15 minutes at RT. Cells were then incubated with AlexaFluor488-conjugated secondary antibody (25 μg/mL) for 1 hour at RT before FACS analysis using Accuri C6 (BD Biosciences).

Subcellular membrane fractionation was performed using a modification of a detergent-free method (31). Normoxic or hypoxic SKBR3 cells were scraped into lysis buffer (150 mmol/L Na₂CO₃, pH 11, containing 1 mmol/L EDTA, protease inhibitor mixture) and sonicated with 3 cycles of 20-second bursts (QSonica Q125 sonicator). For discontinuous sucrose gradients, equal volume of membrane homogenate was mixed with 80% sucrose in 25 mmol/L MES (2-(N-Morpholino)ethanesulfonic acid sodium salt) and 150 mmol/L NaCl (MBS, pH 6.5) to form 40% sucrose bottom layer, with the above layer of 6 mL of 35% sucrose in MBS, followed by 4 mL of 5% sucrose. The gradient mixture was centrifuged at 175,000 × g (33,000 rpm) for 3 hours at 4°C using a SW41Ti rotor (Beckman Instruments). Samples were removed in 1 mL fractions from the bottom of the tube using a fraction collector. Total protein from 1 mL of each fraction was precipitated with methanol/chloroform, and the final pellet was air-dried and resuspended in RIPA buffer for immunoblotting.

Statistical analyses of quantitative experimental methodologies were performed using unpaired Student t test with the GraphPad prism suite. Data are presented as mean ± SD. As for clinical samples, CAV1 expression was analyzed for associations with clinicopathologic characteristics by the χ² test and linear-by-linear with one degree of freedom, using SPSS Statistics version 22 (IBM Corp., 2013). All P values were two-tailed, and P < 0.05 were considered to be statistically significant.

We set out to determine the cytotoxic potency of T-DM1 in breast cancer cells grown in normoxia or hypoxia (1% O₂). Normoxic HER2⁺ cells, as expected, were highly sensitive to T-DM1 at low nanomolar concentrations (IC₅₀, 2.9 and 6.8 nmol/L in SKBR3 and BT474 cells, respectively; Fig. 1A and B), whereas HER2⁻ cells (MDAMB468) were inherently resistant even at the highest concentration tested (10 nmol/L; Fig. 1C). Interestingly, HER2⁺ cells seemed to acquire relative resistance to T-DM1 under hypoxic conditions (IC₅₀ not reached, NR; Fig. 1A and B), as further supported by decreased caspase 3/7 induction by T-DM1 in hypoxic compared with normoxic cells (Supplementary Fig. S1). To understand if the observed resistance under hypoxia is restricted to T-DM1, we next tested a different anti-HER2 antibody precomplexed with a secondary antibody–monomethyl auristatin F toxin conjugate. Remarkably, hypoxic SKBR3 cells were relatively resistant also to this ADC (IC₅₀, NR), whereas normoxic cells were sensitive at low nanomolar concentration (IC₅₀, 4.7 nmol/L; Fig. 1D). Consistent with previous studies (6), trastuzumab antibody alone caused limited cell death and did not reach IC₅₀ even at the highest concentration of 50 nmol/L, independently of the oxygenation status (Fig. 1E).

The tumor-inhibiting activity of trastuzumab has been attributed to attenuation of protumorigenic cell signaling downstream of HER2. To further explore if hypoxia-induced resistance to T-DM1 involved hypoxic regulation of HER2-dependent cell signaling activation, the phosphorylation status of multiple kinases was visualized by antibody array, revealing no major differences between normoxic and hypoxic cells under these conditions (Supplementary Fig. S2). Moreover, AKT and ERK phosphorylation inhibition by trastuzumab treatment was similar in normoxic and hypoxic cells (Fig. 1F). We conclude that hypoxia induces resistance specifically to T-DM1, whereas the effect of trastuzumab alone appears insensitive to hypoxic conditions.

We next hypothesized that the relative resistance to T-DM1 under hypoxia was due to reduced efficacy of the DM1 cytotoxin moiety. To this end, we tested active free DM1, linker attached DM1 (MCC-DM1), and lysine-linker attached DM1 (LMCC-DM1) for cytotoxic effects in normoxia and hypoxia. DM1 was toxic both in normoxia and hypoxia in SKBR3 (IC₅₀ 9.1 and 7.0 nmol/L, respectively) and BT474 cells (IC₅₀ 0.43 and 0.27 nmol/L, respectively; Supplementary Fig. S3A and S3D). MCC-DM1 and LMCC-DM1 carry a net positive charge, which compromises their ability to readily cross the plasma membrane (32, 33). Accordingly, MCC-DM1 and LMCC-DM1 were relatively nontoxic to SKBR3 and BT474 cells (i.e., IC₅₀ NR) both at hypoxic and normoxic conditions (Supplementary Fig. S3B–S3C and S3E–S3F). These findings suggest that hypoxia induces resistance to the intact T-DM1 conjugate and not against its cytotoxin catabolites.

We next investigated the possibility that hypoxic conditions inhibit T-DM1 cytotoxicity through regulation of HER2 expression. Total HER2 as well as constitutive cell surface HER2 expression was unaltered by hypoxia both in SKBR3 and BT474 cells, as determined by immunoblotting (Fig. 2A–C), confocal immunofluorescence microscopy (Fig. 2D and E), and FACS (Fig. 2F and G). However, we found significantly decreased HER2-mediated internalization of trastuzumab at hypoxia as compared with normoxia in SKBR3 as well as BT474 cells, as visualized by confocal microscopy (Fig. 2H), and quantified by FACS (approximately 50% and 65% reduction in SKBR3 and BT474 cells, respectively; Fig. 2I). Hypoxic inhibition of trastuzumab internalization was apparent already at 30 minutes and remained up to at least 4 hours (Fig. 2J). These results suggest that hypoxia can induce resistance to T-DM1 through decreased trastuzumab/HER2 internalization.

We next sought to understand the underlying mechanism of hypoxia-mediated inhibition of trastuzumab internalization and T-DM1 cytotoxicity. Our interest was focused on CAV1, i.e., a structural protein with a preference for cholesterol-rich, lipid raft membrane domains. CAV1 is generally known as a mediator of caveolar endocytosis, but has also been shown to negatively regulate extracellular ligand uptake and receptor protein internalization through plasma membrane stabilization (22, 34–37). We found no difference in total CAV1 levels between normoxia and hypoxia in both SKBR3 and BT474 cells (Fig. 3A and B). Interestingly, however, high-resolution confocal microscopy revealed that hypoxia redistributes CAV1 from intracellular vesicles to the plasma membrane in breast cancer cells (Fig. 3C and D), similarly to previous studies with HeLa and MEF cells (22). In normoxia, we found clear colocalization of internalized trastuzumab with CAV1 in cytoplasmic vesicles, whereas significantly fewer double-positive vesicles were seen in hypoxic cells (Fig. 3C and D). We next further explored the possibility that hypoxia alters CAV1 membrane microdomain distribution in breast cancer cells. Consistent with its raft association, CAV1 was found to be enriched in membrane domains with relatively low density, as shown by sucrose gradient membrane fractionation studies (Fig. 3E). However, in hypoxia, CAV1 was partly redistributed to more high-density, nonraft membrane fractions (Fig. 3E). Moreover, coimmunoprecipitation of HER2 with CAV1 appeared greater in normoxic as compared with hypoxic SKBR3 (Fig. 3F) and BT474 cells (Fig. 3G). Together, these data suggest that hypoxia redistributes CAV1 in the plasma membrane, resulting in decreased vesicular colocalization of trastuzumab and CAV1.

Based on the above findings, we hypothesized that hypoxic resistance to T-DM1 is linked to decreased HER2/trastuzumab internalization through cellular redistribution of CAV1. To explore this possibility, we next used siRNA for transient KD of CAV1 mRNA, resulting in significant reduction of total CAV1 protein as compared with a scrambled, control siRNA sequence (Fig. 4A and B). Interestingly, we found trastuzumab uptake to be significantly reduced (approximately 20% uptake as compared with control) in SKBR3 cells displaying the most complete CAV1 KD, as assessed by quantitative image analysis from confocal microscopy (Fig. 4C and D). Notably, total HER2 protein remained intact in CAV1 KD cells, excluding that the observed effect on trastuzumab uptake was simply due to altered HER2 expression by CAV1 KD (Fig. 4A and B).

To expand on this finding in a different setting, we employed HeLa cervical cancer cells that normally exhibit low endogenous HER2 expression (Fig 5A). Stable CAV1 KD HeLa cells were generated by lentiviral transduction, showing substantial reduction of CAV1 as compared with control cells transduced with scrambled shRNA (Fig. 5A). HER2 was introduced into HeLa cells, followed by FACS sorting (Fig. 5B) to obtain control and CAV1 KD cells expressing comparable levels of HER2, as verified by confocal microscopy and immunoblotting (Fig. 5C and D). First of all, we could show that trastuzumab internalization was inhibited by approximately 75% in hypoxic as compared with normoxic HER2⁺ HeLa cells, as quantified by flow cytometry (Fig. 5E, SCR CTRL). These data support our findings with SKBR3 and BT474 breast cancer cells (Fig. 2H and I), showing hypoxic downregulation of HER2/trastuzumab internalization with a comparable magnitude in HeLa cells. Also, we found a substantial reduction of trastuzumab uptake in CAV1-deficient HeLa HER2⁺ as compared with control HeLa HER2⁺ cells, both in normoxia and hypoxia (Fig. 5E, CAV1 KD), thus corroborating the results with SKBR3 cells (Fig. 4C and D). To investigate how these results translated into T-DM1 treatment effects, we performed cell viability assays, revealing that only normoxic HeLa cells with intact CAV1 expression were sensitive to T-DM1, whereas hypoxic as well as CAV1-deficient normoxic cells were relatively resistant to T-DM1 (IC₅₀ NR; Fig. 5F and G). Together, these data suggest that CAV1 KD mimics hypoxic redistribution of CAV1, resulting in decreased HER2/trastuzumab internalization and T-DM1 cytotoxicity.

To further substantiate that CAV1-dependent uptake is altered in hypoxic conditions, we performed cell viability assays with Abraxane, i.e., a paclitaxel-albumin nanoparticle drug approved in the treatment of breast cancer that is known to enter cells via a CAV1-associated pathway (38). Abraxane was highly toxic to normoxic SKBR3 cells (IC₅₀ 7.0 nmol/L), whereas hypoxic cells were relatively resistant (IC₅₀ NR; Supplementary Fig. S4). These results further support the notion that hypoxic conditions can attenuate the efficacy of macromolecular drugs that enter target cells through CAV1-mediated pathways.

Previous studies have implicated a role for CAV1 tyrosine phosphorylation in protein internalization (39–43). More specifically, del Pozo and colleagues (42) showed that redistribution of phosphorylated CAV1 (pCAV1) from focal adhesions to endosomes is associated with increased endocytosis. Moreover, cellular stress (e.g., high osmolarity) was shown to increase CAV1 phosphorylation on tyrosine 14 and the localization of pCAV1 to the plasma membrane/focal adhesions (44). We next explored the possibility that inhibition of trastuzumab internalization and T-DM1 cytotoxicity by hypoxic stress is related to altered CAV1 phosphorylation and/or pCAV1 subcellular distribution. We did not observe any difference in total pCAV1 between normoxic and hypoxic conditions in SKBR3 cells (Fig. 6A); however, membrane fractionation studies suggested a relative depletion of pCAV1 in membrane raft regions of hypoxic as compared with normoxic cell lysates (fractions 7 and 8; Fig. 6B). Moreover, confocal microscopy studies demonstrated clear colocalization of internalized trastuzumab and pCAV1 in vesicular structures of normoxic cells, whereas significantly fewer double-positive vesicles were seen in hypoxic cells (Fig. 6C–E).

Together, these data are consistent with analyses of total CAV1 (Fig. 4) and suggest a specific role of the phosphorylated fraction of CAV1 in the regulation of trastuzumab uptake. To test this idea more directly, we next generated SKBR3 cells expressing either wild-type CAV1-RFP (WT-CAV1-RFP, control), Y14D-CAV1-RFP, i.e., a CAV1 tyrosine phosphorylation mimic, or phosphorylation-deficient Y14F-CAV1-RFP that exhibits dominant-negative activity. Consistent with studies by Zimnicka and colleagues in endothelial cells (43), we could show that SKBR3 cells expressing phosphomimicking CAV1 (Y14D-CAV1-RFP) as compared with phosphorylation-deficient CAV1-mutant cells (Y14F-CAV1-RFP) display a higher number of vesicular structures (Fig. 6F and G). More importantly, trastuzumab uptake was significantly decreased in Y14F-CAV1-RFP as compared with WT-CAV1-RFP and Y14D-CAV1-RFP SKBR3 cell transfectants, as quantified by imaging analysis (Fig. 6H) as well as by FACS (approximately 30% uptake in Y14F-CAV1-RFP as compared with WT-CAV1-RFP; Fig. 6I). In cell viability studies using the same set of CAV1 SKBR3 transfectants, IC₅₀ was not reached in Y14F-CAV1-RFP, whereas in WT-CAV1-RFP and Y14D-CAV1-RFP cells, IC₅₀ was reached at 2.0 nmol/L (Fig. 6J). From these data, we conclude that pCAV1 has a role in regulating trastuzumab internalization and T-DM1 cytotoxicity, and that this function is perturbed by hypoxic conditions.

The above data suggested a dependency on CAV1 vesicular localization for the function of the internalizing HER2 fraction, which constitutes a small fraction of total surface HER2. Notably, in cell studies, CAV1 did not seem to regulate total HER2 expression (Fig. 4) and vice versa (Fig. 5). To investigate how these observations are reflected in human tumors and to further explore if there is any direct association between HER2 amplification status and CAV1 protein expression, we employed a well-annotated, population-based cohort of patients with breast cancer (Supplementary Fig. S5; ref. 29). Cytoplasmic CAV1 expression was denoted in at least 20% of all stained invasive tumor cells in 274 of 635 patient samples (approximately 43% of total). Among these, 231 tumors displayed weak, 39 moderate, and 4 strong CAV1 expression (Supplementary Fig. S6). A higher level of CAV1 was observed in adjacent carcinoma in situ (CIS) cells compared with invasive tumor cells. In benign appearing ducts as well as CIS, the CAV1 level was higher in the myoepithelial compared with the luminal cells (Supplementary Fig. S7). Due to the limited number of tumors with strong CAV1 expression, these tumors were grouped with tumors of moderate intensity for further analyses. We found no significant associations between CAV1 expression and patient age at inclusion, invasive tumor size, or axillary lymph node positivity (Supplementary Table S1). CAV1 expression showed a significant inverse association with estrogen receptor (ER) and progesterone receptor status. Importantly, however, there was no association of CAV1 expression with HER2 amplification status (Supplementary Table S1). We conclude that, in line with the in vitro data, there is no apparent coregulatory association between HER2 and cytoplasmic CAV1 expression.

Since the advent of HER2 targeted therapies, the clinical outcome for patients with breast cancer with HER2⁺ tumors has improved dramatically. Still, there is evidence of significant development of resistance to trastuzumab as well as to T-DM1, posing a clinical challenge that may be resolved by an increased understanding of HER2 function in the context of the tumor microenvironment. Here, we demonstrate that hypoxia, i.e., a specific and universal feature of aggressive tumors, confers resistance to T-DM1. It is suggested that hypoxia-induced T-DM1 resistance is the result of deficient trastuzumab/HER2 internalization due to redistribution of pCAV1 from vesicular membrane raft domains to the plasma membrane. These findings reveal a novel link between tumor hypoxic stress conditions, HER2 trafficking, and T-DM1 chemosensitivity with potential implications for improved therapeutic strategies and response prediction.

Based on the present findings together with previous studies (22), different categories of cell surface receptors in relation to hypoxia and CAV1 may be proposed. These include proteins that are independent on CAV1 for their endocytosis and that may escape the negative regulation by CAV1 redistribution in hypoxia (e.g., CAIX); proteins that are independent on CAV1 in normoxia but are negatively regulated by CAV1 membrane stabilization in hypoxia; and proteins, including HER2, that are dependent on CAV1 for their endocytosis and negatively regulated by hypoxic redistribution of CAV1. This implies that the inhibitory effect of hypoxia on receptor protein internalization may in some cases be alleviated by the loss of CAV1, whereas in others the hypoxic situation would instead be mimicked by deficient CAV1 or CAV1 phosphorylation, as shown for HER2 in the present study. Our data thus point at a scenario where hypoxia perturbs CAV1 function as an important mediator of HER2 tumor antigen internalization and ADC delivery. We show that CAV1 moves from low-density (raft) to more high-density membrane regions, which may be the result of, for example, hypoxia-induced remodeling of membrane lipid composition or altered kinase activation, perhaps most importantly Src and p38 MAP kinases that previously have been associated with stress and regulation of caveolae function (44). Another possible mechanism of how hypoxia redistributes CAV1 is by increased expression or activation of a protein interaction partner that preferentially localizes to high-density plasma membrane regions. Clearly, a key remaining question for future studies is to understand exactly how hypoxia regulates CAV1 function.

Our data illustrate the limitations of studies that extrapolate genomic data or overall protein expression to biological function, and highlight the importance of adding spatial information on protein distribution. Notably, the current status of patient selection for T-DM1 treatment does not take into account the subcellular distribution of HER2, and therefore may overestimate accessible HER2 that can engage in T-DM1 internalization. Accordingly, previous PET imaging studies demonstrated a lack of correlation between HER2 expression and tumor uptake of trastuzumab (45, 46), suggesting that assessment of HER2 alone is insufficient to predict how patients respond to T-DM1. The present study provides a potential mechanistic explanation to this notion and should motivate future studies that explore how markers of tumor hypoxia and CAV1 phosphorylation may serve to better predict therapeutic vulnerability of HER2⁺ tumors to T-DM1 treatment.

Of particular relevance in breast cancer, CAV1 may directly regulate HER2 and ER signaling activity (47–49). Indeed, previous studies have demonstrated that HER2 is localized to cholesterol-rich membrane raft regions where it may colocalize with GM1 (50) as well as CAV1 (51). Recent studies by Pereira and colleagues (52) suggested an inverse relationship between HER2 and CAV1 protein expression in a large collection of cancer cell lines and in examples of human gastric tumor specimens. In the present study, we were, however, unable to find a significant association between CAV1 expression and HER2 status in vitro as well as in a large, population-based breast cancer cohort. As the cohort was established prior to the introduction of T-DM1 treatment, we were unable to perform correlative analyses between CAV1 expression and response to T-DM1 treatment. Caveolae-mediated endocytosis is one of the major clathrin-independent raft-dependent endocytic routes (53), and studies on its role in trastuzumab uptake and T-DM1 cytotoxicity have generated conflicting data. Chung and colleagues (26) suggested that CAV1 promotes chemosensitivity to T-DM1, whereas others, on the contrary, reported that the acquisition of TDM1 resistance is due to deficient cleavage of T-DM1 in CAV1-associated endosomes (28). In favor of a role of CAV1 in promoting T-DM1 sensitivity, increased CAV1 expression by metformin was shown to induce T-DM1 internalization and subsequent cytotoxicity (27). Our data strongly support the concept of a dependence on vesicular, raft-associated CAV1 for efficient T-DM1 cytotoxic activity. In addition, we show a specific role for pCAV1 in trastuzumab internalization and T-DM1 sensitivity that, however, is disrupted in the context of tumor hypoxic conditions. According to this scenario, the most obvious strategy would be to alleviate the hypoxic situation by, for example, tumor vessel normalization or actions that normalize CAV1 localization to allow more efficient targeting also of hypoxic tumor regions and potentially other niches distinguished by CAV1 clustering at the cell periphery. Interestingly, in models of gastric, bladder, and breast cancer, increased HER2 cell surface availability and improved trastuzumab therapy by disrupting CAV1-mediated HER2 internalization using the cholesterol lowering drug lovastatin were demonstrated (52). In the case of T-DM1, however, one would like to achieve the opposite effect, i.e., increased membrane cholesterol loading and CAV1-mediated HER2 internalization.

In summary, we show that hypoxic conditions induce resistance to targeted treatment with T-DM1 in HER2⁺ breast cancer cells and HER2-overexpressing cervical cancer cells by perturbing the localization and function of CAV1. Our findings may have direct implications for improved targeting and response prediction of HER2-expressing tumors.

O.M. Saad is Senior Scientist at Genentech/Roche and has an ownership interest (including patents) in Roche Stock. O. Gluz is an advisory board member for Roche and is employed with Daiichi. S. Borgquist is clinical advisor at Pfizer, reports receiving other commercial research support from Roche, and has honoraria from the speakers' bureau of Pfizer. H. Jernström has an ownership interest (including patents) in Pfizer stocks. No potential conflicts of interest were disclosed by the other authors.

Conception and design: V. Indira Chandran, M. Belting

Development of methodology: V. Indira Chandran, M. Cerezo-Magaña, I.R. Nabi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Indira Chandran, B. Nodin, N. Koppada, O.M. Saad, S. Borgquist, H. Jernström, M. Belting

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Indira Chandran, M. Barbachowska, M. Cerezo-Magaña, N. Koppada, O.M. Saad, O. Gluz, H. Jernström, M. Belting

Writing, review, and/or revision of the manuscript: V. Indira Chandran, M. Cerezo-Magaña, O. Gluz, K. Isaksson, S. Borgquist, K. Jirström, I.R. Nabi, H. Jernström, M. Belting

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Indira Chandran, A.-S. Månsson, M. Barbachowska, B. Nodin, K. Jirström, H. Jernström

Study supervision: V. Indira Chandran, H. Jernström, M. Belting

Other (Second reader for TMA scoring and author of selected TMA images): M. Barbachowska

Other (some of the Caveolin constructs were provided): B. Joshi

We thank Maria C. Johansson for excellent technical assistance and all the patients who contributed to this study.

This study was funded by grants (to M. Belting) from the Swedish Cancer Society (CAN 2017/664); the Swedish Research Council (2018-02562); the Mrs. Berta Kamprad Foundation; the Skåne University Hospital donation funds; the Governmental funding of clinical research within the national health services, ALF; and a donation by Viveca Jeppsson.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.