A feed-forward loop between SorLA and HER3 determines heregulin response and neratinib resistance

Abstract

Current evidence indicates that resistance to the tyrosine kinase-type cell surface receptor (HER2)-targeted therapies is frequently associated with HER3 and active signaling via HER2-HER3 dimers, particularly in the context of breast cancer. Thus, understanding the response to HER2-HER3 signaling and the regulation of the dimer is essential to decipher therapy relapse mechanisms. Here, we investigate a bidirectional relationship between HER2-HER3 signaling and a type-1 transmembrane sorting receptor, sortilin-related receptor (SorLA; SORL1). We demonstrate that heregulin-mediated signaling supports SorLA transcription downstream of the mitogen-activated protein kinase pathway. In addition, we demonstrate that SorLA interacts directly with HER3, forming a trimeric complex with HER2 and HER3 to attenuate lysosomal degradation of the dimer in a Ras-related protein Rab4-dependent manner. In line with a role for SorLA in supporting the stability of the HER2 and HER3 receptors, loss of SorLA compromised heregulin-induced cell proliferation and sensitized metastatic anti-HER2 therapy-resistant breast cancer cells to neratinib in cancer spheroids in vitro and in vivo in a zebrafish brain xenograft model.

Fig. 1. HER3 signaling regulates SORL1 expression.

A SORL1 expression is significantly higher in breast cancer cell lines with high ERBB3 expression (CCLE; N = 59). RMA: robust multi-array average. Data are mean ± SD; statistical analysis: Mann–Whitney U. B SORL1 expression is higher in tumors with high ERBB3 and ERBB2 expression (cBioPortal; N = 476). Violin plot boxes represent median and 25th and 75th percentiles (interquartile range), and whiskers extend to maximum and minimum values. See methods for details on how the groups were defined. C Hrg β-1 increases SorLA levels. BT-474 cells were stimulated with 20 ng.mL⁻¹ Hrg β-1 for the indicated times. Representative immunoblotting of SorLA, AKT(p)S473, total AKT, ERK1/2(p)T202/Y204, total ERK1/2, with α-tubulin as a loading control. D Quantification of SorLA levels normalized to loading control and relative to non-stimulated (0 h) cells. E Hrg β-1 increases SORL1 expression. Quantification of SORL1 mRNA levels, relative to HPRT1, determined with RT-qPCR in BT-474 cells stimulated with 20 ng.mL⁻¹ Hrg β-1 for the indicated time points relative to non-stimulated (0 h) cells. F Representative immunoblotting of SorLA, AKT(p)S473, total AKT, ERK1/2(p)T202/Y204, total ERK1/2, with α-tubulin as a loading control from control (mCherry)- or Hrg β-1-overexpressing BT-474 cells. G Quantification of SorLA levels normalized to loading control and relative to control cells. H SMDF increases SorLA expression. BT-474 cells were cultured on a monolayer of mCherry-positive control or SMDF-overexpressing fibroblasts (TIFF). BT-474 cells were FACS sorted (see “Methods”) and cell lysates were analyzed for SorLA, AKT(p)S473, total AKT, ERK1/2(p)T202/Y204, total ERK1/2, with α-tubulin as a loading control. I Quantification of SorLA levels normalized to loading control and relative to control TIFF cocultured cells. J Representative immunoblotting of SorLA, HER2, and HER3, with α-tubulin as a loading control from parental BT-474 and brain-tropic metastasis variant BT-474-Br cells. K Quantification of indicated protein levels normalized to loading control and relative to BT-474 cells. L Quantification of SORL1 mRNA levels, normalized to HPRT1, determined with RT-qPCR in parental BT-474 and BT-474-Br cells relative to BT-474 cells. Unless otherwise indicated, data are mean ± SD from three independent biological experiments; statistical analysis: Student’s t test (unpaired, two-tailed, unequal variance).

Fig. 2. Heregulin-induced SORL1 regulation requires HER3 signaling through ERK1/2.

A Representation of the different SORL1 proximal promoter constructs (P1-7) used in luciferase assays with their corresponding molecular lengths (bp). B P3 is a responsive promoter sequence to Hrg β-1 stimulation. A control empty vector (EV) and P1-7 were individually expressed in BT-474 cells together with pRL-TK Renilla luciferase transfection control and cells were stimulated or not with 20 ng.mL^–1 Hrg β-1 for 24 h. Shown are luciferase activities relative to the control sample (EV, PBS-treated). C Representative confocal microscopy images of P3-GFP promoter reporter construct-expressing BT-474 cells treated or not for 24 h with 20 ng.mL⁻¹ Hrg β-1 in the presence or absence of 100 nM trametinib. Scale bars: 10 µm. D Quantification of the mean intensity of GFP signal per cell (whole-cell area). N = 36 cells per group. Three biological replicates. E Trametinib inhibits Hrg β-1-induced upregulation of SorLA. BT-474 cells were cotreated with 20 ng.mL⁻¹ Hrg β-1 and either the pan-AKT inhibitor MK-2206 (2 µM) or the ERK pathway inhibitor trametinib (100 nM) for 48 h. Representative immunoblotting of SorLA, AKT(p)S473, total AKT, ERK1/2(p)T202/Y204, total ERK1/2, with GAPDH as a loading control. F Quantification of SorLA levels normalized to loading control and relative to non-treated cells. G SORL1 mRNA levels, relative to HPRT1, determined with RT-qPCR in BT-474 cells stimulated or not with 20 ng.mL⁻¹ Hrg β-1 in the presence or absence of trametinib relative to non-treated cells. Data are mean ± SD from four (B) and three (F and G) independent biological replicates; statistical analysis: (B) Two-way ANOVA, Dunnett’s multiple comparisons test. (F and G) Student’s t test (unpaired, two-tailed, unequal variance). (D) Box plots represent median and interquartile range, and whiskers extend to maximum and minimum values; One-way ANOVA, Dunn’s multiple comparisons test.

Fig. 3. SorLA regulates HER3 stability.

A SorLA and HER3 protein levels correlate positively in breast cancer cell lines (DepMap portal; N = 29). B–E HER3 and HER2 expression correlates with SorLA in HER2-positive breast cancer. B, C SorLA-GFP transfection in JIMT-1 cells increases HER2 and HER3 levels compared to GFP transfected cells. D, E. SorLA silencing in BT-474 cells decreases HER2 and HER3. B, D Representative immunoblotting of total HER2, HER3, and SorLA with β-actin as a loading control. C, E represent the respective quantifications of immunoblots in (B, D) with HER2/HER3 levels normalized to loading control and relative to control-silenced cells. F SorLA silencing decreases HER3 stability. RNAi transfected BT-474 cells were treated with 25 µg.mL⁻¹ of CHX for the indicated time points and HER3 protein levels were determined by immunoblotting (see Fig. S3E). Shown are HER3 levels normalized to α-tubulin and relative to 0 h timepoint. Half-lives (T_1/2) represent the time required for HER3 to decrease to 50% of its initial level. The least squares fitting method and extra-sum-of-squares F test were used to assess the statistical difference between curves from control and SorLA-silenced cells (P = 0.0002). A representative western blot validating SorLA silencing is shown. G HER3 is primarily degraded through the lysosomal pathway. BT-474 cells were treated with 1 µM of bortezomib or 50 nM of bafilomycin A1 for 4 h to inhibit proteasome and lysosome activities, respectively. HER3 expression was analyzed by immunoblotting, with α-tubulin as a loading control. H Quantification of HER3 levels normalized to loading control and relative to DMSO-treated control cells. I SorLA silencing triggers HER3 lysosomal degradation. SorLA-silenced BT-474 cells were cotreated for 4 h with CHX and bafilomycin A1. HER3 expression was analyzed by immunoblotting, with α-tubulin as a loading control. J Quantification of HER3 levels normalized to loading control and relative to CHX-treated control-silenced cells. Data are mean ± SD from three (C, F, H, J) or four (E) independent biological replicates. Statistical analyses: Student’s t test (unpaired, two-tailed, unequal variance) unless indicated otherwise. Scr: control non-targeting siRNA.

Fig. 4. SorLA interacts with HER2-HER3 dimers.

A A scheme depicting BiFC between SorLA-v1 and HER2/3-v2. B The indicated BiFC dimers were expressed in BT-474 cells and their interaction was assessed through imaging of reconstituted Venus. Cells are outlined with white dotted lines. Shown are representative confocal microscopy images and a control Venus-expressing cell showing diffuse fluorescence localization. Scale bars (main): 10 µm. Scale bars (insets): 2 µm. C Venus or SorLA-v1/HER3-v2 were transiently expressed in BT-474 cells. Cell lysates were subjected to nanobody pulldown (specific for the reconstituted Venus v1 + v2 dimer) and pulldowns and total cell lysates (input) were immunoblotted with the indicated antibodies. D–F. SorLA interacts with HER3 in a pH-dependent manner. SPR analysis on immobilized SorLA and a 20–200 nM concentration series of HER3 at pH 5.0 (D), 6.0 (E) and 7.4 (F). KD: equilibrium dissociation constant. Data are representative of three independent biological replicates.

Fig. 5. SorLA regulation of HER2 and HER3 requires functional Rab4.

A Representative Airyscan confocal microscopy images of BT-474 cells coexpressing GFP-SorLA with the indicated endosomal markers. SiR-Actin was used for counterstaining the actin cytoskeleton. White arrows depict colocalizing signals. Scale bars: 10 µm. Scale bars (insets): 2 µm. B GFP-SorLA strongly colocalizes with mCherry-Rab4 in BT-474 cells. Colocalization was calculated (see “Methods”) from BT-474 cells transfected and imaged as in (A). N = 36 cells per group. C JIMT-1 cells were cotransfected with GFP-SorLA and either GFP control or GFP-Rab4^S22N dominant-negative mutant. Representative immunoblotting of SorLA, HER2, HER3, GFP, and Rab4, with α-tubulin as a loading control. GFP immunoblot detects GFP control and GFP-Rab4^S22N proteins. The higher molecular weight GFP-SorLA was probed with anti-SorLA primary antibody. In the Rab4 immunoblot, the upper arrowhead indicates GFP-Rab4^S22N and the lower arrowhead the endogenous protein. D Quantification of HER2 and HER3 levels normalized to loading control and relative to control GFP-transfected cells. E Representative confocal microscopy images of BT-474 cells co-overexpressing mCherry-Rab4 with the indicated BiFC dimers. SiR-Actin was used for counterstaining the actin cytoskeleton. White arrows depict colocalizing signals. Scale bars: 10 µm. Scale bars (insets): 2 µm. F Colocalization analysis between BiFC and mCherry-Rab4. N = 30 cells per group. D Data are mean ± SD from three independent biological experiments; statistical analysis: Student’s t test (unpaired, two-tailed, unequal variance). B and F Box plots represent median and 25th and 75th percentiles (interquartile range), and whiskers extend to maximum and minimum values; three biological replicates. Statistical analysis: One-way ANOVA, Dunn’s multiple comparisons test.

Fig. 6. SorLA expression determines heregulin response.

A SorLA silencing decreases cell viability in basal and Hrg β-1-enriched cell culture conditions. Control and SorLA-silenced BT-474 cells were stimulated or not with 20 ng.mL⁻¹ Hrg β-1 for 48 h and cell viability was assessed using a WST-8-based method. Cell viability values are represented as fold change relative to unstimulated control BT-474 cells. A representative western blot validating SorLA silencing is shown. B Hrg β-1-expressing BT-474 cells were silenced for SorLA and cell viability was analyzed as in (A). A representative western blot validating SorLA silencing is shown. C, D SorLA silencing inhibits spheroid growth in Hrg β-1-enriched extracellular matrix. SorLA-silenced mCherry BT-474 cells were cocultured with Hrg β-1-overexpressing fibroblasts (Hrg β-1 TIFF) in matrigel for 12 days. Shown are representative live-cell images (C) and quantification of mCherry fluorescence reflecting spheroid growth (D). Data are mean ± SD from three independent biological replicates; statistical analysis: (A and D), One-way ANOVA, Dunn’s multiple comparisons test and (B), Student’s t test (unpaired, two-tailed, unequal variance). Scr = control non-targeting siRNA.

Fig. 7. SorLA silencing specifically reverts resistance to neratinib in vitro and in vivo.

A, B SorLA silencing and neratinib show synergistic effects in inhibiting MDA-MB-361 cell growth. Control- and SorLA-silenced MDA-MB-361 cells were treated with either the pan-PI3K inhibitor buparlisib (500 nM), the pan-AKT inhibitor MK-2206 (500 nM) or the dual HER2/EGFR tyrosine kinase inhibitor neratinib (300 nM) for 48 h. Cell viability was measured using WST-8-based method. Values are represented as fold change relative to DMSO-treated control MDA-MB-361 cells. A representative western blot validating SorLA silencing is shown. C, D SorLA-silencing exhibits a synergistic effect with neratinib in inhibiting anchorage-independent spheroid growth of MDA-MB-361 cells in 3D low-attachment cell culture conditions. Control- and SorLA-silenced MDA-MB-361 cells were grown in low-attachment cell culture conditions for 7 days in the presence of the indicated treatments. Scale bar 100 µm. Spheroid sizes are quantified in and a representative western blot validating SorLA silencing are shown in (D). N > 50 spheroids per group. E Control and SorLA-silenced GFP-MDA-MB-361 cells were engrafted in the brain of zebrafish embryos and allowed to grow for 4 days in the presence of DMSO control or neratinib (400 nM). Representative GFP + Bright Field images of brain tumors are shown. Scale bars: 1 mm. Scale bars (insets): 100 µm. F Control- and SorLA-silenced GFP-MDA-MB-361 cells were engrafted in zebrafish brain and allowed to grow for 4 days in the presence of DMSO, buparlisib (2 µM), neratinib (400 nM) or MK-2206 (400 nM). Tumor growth is represented as fold change in GFP intensity relative to day 1 post-engraftment. A representative western blot validating SorLA silencing is shown. G A representative scheme of a neratinib resistance mechanism driven by a feed-forward loop supporting SorLA-HER2-HER3 expression. Key elements of this loop are highlighted in green boxes. HER2-HER3 signaling increases SORL1 expression through the ERK pathway. Increased SorLA levels prevent HER2-HER3 lysosomal degradation (red arrow with lined arrowhead) presumably by stimulating receptor recycling from Rab4-positive endosomes (green arrows). Data are (A and B), mean ± SD from three 3 independent biological replicates; (D), Box plots representing median and interquartile range; whiskers extend to maximum and minimum values, and (F), median with interquartile range. Statistical analysis: (A, B), two-way ANOVA, Dunnett’s multiple comparisons test. (D), One-way ANOVA, Dunn’s multiple comparisons test. (F), two-way ANOVA, Dunnett’s multiple comparisons test (main SorLA silencing effect P < 0.0001). Scr: control non-targeting siRNA.