GRP94 promotes brain metastasis by engaging pro-survival autophagy

Abstract

Background

GRP94 is a glucose-regulated protein critical for survival in endoplasmic reticulum stress. Expression of GRP94 is associated with cellular transformation and increased tumorigenicity in breast cancer. Specifically, overexpression of GRP94 predicts brain metastasis (BM) in breast carcinoma patients with either triple negative or ErbB2 positive tumors. The aim of this study was to understand if microenvironmental regulation of GRP94 expression might be a hinge orchestrating BM progression.

Methods

GRP94 ablation was performed in a BM model BR-eGFP-CMV/Luc-V5CA1 (BRV5CA1) of breast cancer. In vitro results were validated in a dataset of 29 metastases in diverse organs from human breast carcinomas and in BM tissue from tumors of different primary origin. BM patient-derived xenografts (PDXs) were used to test sensitivity to the therapeutic approach.

Results

BMs that overexpress GRP94 as well as tumor necrosis factor receptor-associated factor 2 are more resistant to glucose deprivation by induction of anti-apoptotic proteins (B-cell lymphoma 2 and inhibitors of apoptosis proteins) and engagement of pro-survival autophagy. GRP94 ablation downregulated autophagy in tumor cells, resulting in increased BM survival in vivo. These results were validated in a metastasis dataset from human patients, suggesting that targeting autophagy might be strategic for BM prevention. Indeed, hydroxychloroquine treatment of preclinical models of BM from PDX exerts preventive inhibition of tumor growth (P < 0.001).

Conclusions

We show that GRP94 is directly implicated in BM establishment by activating pro-survival autophagy. Disruption of this compensatory fueling route might prevent metastatic growth.

Key Points.

1. GRP94 promotes BM cell survival to hypoglycemic stress by activating anti-apoptotic proteins and autophagy.

2. GRP94 deletion decreases cell autophagy and reduces BM efficiency in vivo.

3. Preventive treatment with hydroxychloroquine reduces BM growth in a patient-derived xenograft.

Importance of the Study.

Brain metastasis develops in 30% of patients with breast cancer. After surgery alone, the median survival time ranges from 4 to 6 months. With surgery and radiation, the median survival time may exceed 6 months. Despite the fatal prognosis of BM, our knowledge about metastatic biology and mechanisms required to adapt to the tissue environment in the central nervous system remains limited. An understanding of the pathogenic mechanisms leading to metastatic growth may have prognostic and therapeutic value. Previous published data suggested that GRP94 could be used as a BM biomarker to predict progression-free survival of breast cancer patients, but the mechanisms by which GRP94 promotes metabolic adaptation to the microenvironment are unexplored. Using in vivo models of BM, as well as tissue from patients, we show that GRP94 promotes survival under glucose deprivation by activation of anti-apoptotic proteins as well as autophagy. The results of this study suggest that autophagy could be a new therapeutic target in BM useful for the design of new therapies.

Despite extraordinary advances in our understanding of the biology underlying breast cancer progression and potential molecular targets for its treatment,^1,2 30% of patients develop brain metastasis (BM). Since less than 10% of patients have detectable distant metastasis at the time of diagnosis,³ it remains a challenge to understand the pathogenic mechanisms of tumor cell brain tropism for the early identification of patients with BM risk. Differentiating among breast cancer subtypes for likelihood of metastasis to the brain would allow a strict patient follow-up and early treatment before clinical symptoms appear.⁴

Computational modeling⁵ provided driver functions in breast cancer BM where glucose-regulated protein GRP94 (94-kDa glycoprotein) appeared as a central protein in a cluster of chaperones and co-chaperones differentially expressed in BM cells,⁶ being a useful biomarker to predict progression-free survival.^7,8 GRP94 is an ATP-dependent chaperone that often functions as a dimmer.⁹ It provides a platform for the assembly or oligomerization of loaded protein cargo, which is required for the conformational maturation of proteins directed to cell surface display or export.¹⁰ GRP94 serves as a pro-survival component that protects cells against death induced by endoplasmic reticulum (ER) stress (ERS), connecting a set of pathways known collectively as the ERS resistance phenotype (ERSRP).⁷ Expression of GRP94 might counteract the exposure of cells to reactive oxygen species promoting anti-oxidant molecules.¹¹ GRP94’s overexpression has been associated with cellular transformation and increased tumorigenicity in a variety of cancer cell lines and human cancer biopsies, which may correlate with increased metabolic stress resistance.^12,13 When bound to the cell surface, GRP94 can act as a danger-associated molecular pattern,^14,15 presenting antigens to the immune system and acquiring new functions unrelated to its classical ER role. GRP94 activation promotes many critical functions required for metastasis progression, like chaperoning HER2 dimerization on the cell membrane, a process critical for activation of human epidermal growth factor receptor 2 (HER2) signaling in a subtype of HER2 positive breast carcinomas.¹⁶

The establishment of metastatic cells in a new microenvironment requires preferential selection of those metabolic functions that allow survival.¹⁷ The brain is one of the organs with higher glucose dependence, where astrocytes consume glucose and shunt lactate to neurons.¹⁸ As a result, the glucose levels in the brain interstitial space are lower than in blood.^19–21 Here we report that GRP94 over-expression is critical for BM progression, inducing a metabolic switch by coordinating the unfolded protein response and pro-survival autophagy, which relieves metabolic stress in low glucose conditions. These findings highlight the role of autophagy modulators to prevent BM progression.

Methods

Cell Culture and Treatments

MDA-MB-435 (435-P) breast cancer cells and their metastatic variants to brain (435-Br1), lung (435-L3), bone (435-B1), liver (435-Lv1), and lymph node (435-N1) were obtained from successive in vivo/in vitro passages.²² MDA-MB-361 cells were obtained from Laboratoire d’Oncogénetique in Saint-Cloud, France. SA52 cells were obtained from the Laboratory of Experimental Cancerology at the University of Ghent, Belgium. BRV5 and BRV5CA1 were obtained after several rounds of in vitro/in vivo injection of BR-eGFP-CMV/Luc cells described elsewhere.²³ All cell lines were maintained under standard conditions unless noted (see Supplementary Methods).

Cell Viability and Cell Death Measurement

MTT tetrazolium assay was used in cells plated at 10⁴ cells per well for 72 h in the presence or absence of the mentioned drugs or different glucose concentrations (from 4.5 to 0 mg/mL). For crystal violet staining, cells were fixed in 4% paraformaldehyde and stained with crystal violet (0.2% in 2% ethanol) for 20 min. Dye was extracted with 10% sodium dodecyl sulfate and relative proliferation was determined by measuring optical density at 595 nm.

GRP94 Protein Knockdown

For transient GRP94 knockdown, 435-Br1 cells were transfected with Lipofectamine TM2000 (Invitrogen) at 3 µg/mL and 50 nM of small interfering (si)GRP94 (Invitrogen), using the specified knockdown sequences (see the Supplementary Methods section).

Protein Expression

Protein expression was assessed by fluorescence microscopy, western blotting, and immunohistochemistry (IHC) following standard protocols with some modifications that are detailed in the Supplementary Methods section.

Animal Models

Orthotopic establishment of BM and bioluminescence was performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals, with the approval of the animal care committee (reference 9703) as previously reported.³⁰ Mice were euthanized when they showed signs of declining health and visible body weight loss. Intracarotid establishment of BM was induced by intracarotid artery inoculation, described elsewhere (see the Supplementary Methods).

Human Transcriptomic Data

Gene expression data from the GSE14017 dataset (ref. 19573813) were used. Normalized transcriptomic data (series matrix) on 29 human metastatic samples (15 brain, 10 bone, 4 lung) were downloaded from the Gene Expression Omnibus repository. Expression differences among each metastatic location in genes tumor necrosis factor receptor-associated factor 2 (TRAF2) and activating transcriptional factor 6 (ATF6) were plotted and statistical significance was calculated using a Wilcoxon test. Hierarchical clustering was performed to display the classification ability among metastatic locations of genes with a role in autophagy (list extracted from autophagy database) and fatty acid metabolism (list extracted from gene set enrichment analysis database).

Statistical Analysis

For survival analysis of groups, we used the nonparametric Mann–Whitney test. Bioluminescence data were transformed using the log(1 + x) function (where x = AvR) to obtain a more regular and positive distribution, and normalized by subtracting the first observation (day 4) from each of the following ones. Student’s t-test was used to compare groups. Survival curves for each treatment were estimated by the Kaplan–Meier method, and the log-rank test was used to assess if they were significantly different. P-values lower than 0.05 were considered significant. Microsoft Excel and GraphPad Prism were used to plot graphs.

Results

GRP94 Orchestrates the Metabolic ERS Resistance Phenotype

435-Br1 cells overexpressed GRP94 compared with 435-P cells by immunofluorescence (Supplementary Fig. 1A) and western blot (Fig. 1A). Overexpression was observed in BM variants 435-Br1 and BRV5 and in 2 more breast cancer BM cells (MDA-MB-361 and SA52) compared with metastatic cells in bone (435-B1) and lung (435-L3). GRP94 was also stably overexpressed under the influence of the brain microenvironment in experimental models (Fig. 1B). GRP94 expression has been previously reported as being co-regulated with GRP78/binding immunoglobulin protein (BiP) under a variety of stress conditions.²⁴ In contrast to GRP94, GRP78 was similarly expressed in all metastatic variants except in 435-L3 (Supplementary Fig. 1A). GRP94 expression in nonmetastatic breast cancer cell lines was comparable to 435-P cells, whereas GRP78 was overexpressed in SK-BR3 with regard to other cell lines (Supplementary Fig. 1B). These results were consistent with a transcriptomic analysis performed in BM from breast cancer patients⁷ and in previous studies where GRP94 functioned independently of other co-chaperones,²⁵ suggesting a specific role in BM seeding.

Fig. 1.

BM cells overexpress GRP94 and present a survival advantage in hypoglycemic conditions. (A) Immunoblotting of indicated cells (top). Semiquantitative analysis shows a significant increase in GRP94 expression in BM cell lines (bottom). Bars represent relative fold change vs 435-P; error bars represent SD of independent experiments (435-P, BrV5, and L3 n = 5, B1 and SA52 n = 4, 361 and Br1 n = 3). (B) Hematoxylin & eosin (10x) staining of BM (upper) and IHC expression of GRP94 (bottom) in experimental BM (20x) induced by intracranial injection of BRV5CA1 cells (scale bars 100 μm). (C) MTT analysis of 435-P, 435-Br1, and 435-L3 metastatic cells challenged with hypoglycemic medium (72 h). (D) MTT analysis of 435-Br1 cells, 435-Br1 siC, and 435-Br1 siGRP94 to assess cell viability (48 h) under normal or hypoglycemic conditions (left). Immunoblotting of 435-Br1 after GRP94 siRNA knockdown (siGRP94) and siRNA control (siC) for 48 h (right). (E) Immunoblotting of 435-P, 435-Br1, 435-Br1 siC, 435-Br1 siGRP94 or cells treated with the HSP inhibitor 17AAG (435-Br1 + 17AAG) that antagonizes GRP94. TRAF2 and ATF6 expression were quantified vs actin and normalized to 435-P cells. For (C) and (D), error bars represent SD of technical replicates (one representative of 3 experiments). For all panels, significance was determined with t-test. *P < 0.05, **P < 0.01, ***P < 0.001.

Given the well-known regulation of GRP94 by glucose levels, we evaluated cell survival after hypoglycemic stress according to metastatic organ specificity (Fig. 1C). In extreme hypoglycemic conditions, 435-Br1 cells survived significantly more (62%) than parental 435-P (42%) and 435-L3 cells (48%). These results were confirmed by MTT and crystal violet, suggesting that altered mitochondrial function is not responsible for the observed phenotype (Supplementary Fig. 1C). Interestingly, surviving populations after glucose deprivation expressed similar levels of GRP94 (Supplementary Fig. 1D), suggesting that GRP94 is required to overcome starvation. Depletion of GRP94 by siRNA (Fig. 1D, right) reduced survival of siGRP94-Br1 cells (25–30%) with regard to control siC-Br1 cells (Fig. 1D, left), confirming a role for GRP94 in adaptation and survival in low glucose conditions not due to off-target effects (Supplementary Figures 1 E–F). Moreover, we analyzed unfolded protein response (UPR) activation and found TRAF2 over-expressed in 435-Br1 cells with regard to 435-P while it remained unchanged in 435-Br1 siGRP94 cells versus controls Fig. 1E. ATF6 expression was not differential among these cell lines, suggesting a minor contribution of this pathway. Since Ire1 is rapidly degraded when cells are treated with geldamicin,²⁶ we found lower TRAF2 expression in cells challenged with 17AAG, a synthetic derivative of geldanamycin that inhibits HSP90, an analog of GRP94 (Fig. 1E).

We further analyzed the expression of heat shock proteins (HSPs) and other glucose-regulated proteins (GRPs), which have been previously associated with metastatic progression.²⁷ A moderate overexpression of GRP58/ERp57 was observed in 435-Br1 cells at 4.5 mg/mL glucose with regard to with 435-P cells, while HSPs like HSP60 or HSP70 were underexpressed compared with 435-P cells (Supplementary Fig. 1G). Activation of HSP60/70 only occurred in 435-Br1 at 0 mg/mL glucose to similar levels as 435-P basally, suggesting that 435-Br1 cells are able to delay activation of HSP response while maintaining an increased GRP activity. We challenged 435-P and 435-Br1 cells with other ER stressors such as thapsigargin (inhibitor of ER Ca²⁺ATPases), tunicamycin (inhibitor of early-step protein glycosylation), cycloheximide (inhibitor of protein synthesis) and CoCl₂ (cobalt chloride), which mimics hypoxia, but cell viability was not differential between 435-P and 435-Br1 cells (Supplementary Fig. 1H).

All together, these results suggested that the increased basal levels of GRP94 may act as a buffer system in BM cells, conferring a survival advantage in nutrient deprivation conditions by activating alarm genes associated with metabolic ERS and delaying cell damage.

GRP94 Engages ERSRP in BM Cells by Induction of Anti-Apoptotic Proteins and Pro-Survival Autophagy

Uncontrolled ERS can activate general regulators of apoptosis leading to cell death.²⁸ To confirm that engagement of the ERSR through GRP94 activation conferred resistance to metabolic stress, we analyzed cell death by flow cytometry and annexin/propidium iodide staining (Supplementary Fig. 2A, B). In extreme hypoglycemic conditions, 435-P cells showed increased annexin/propidium iodide positive staining (late apoptotic) in contrast to 435-Br1 (Supplementary Fig. 2A, B), suggesting that BM cells are more resistant to cell death induced by metabolic stress. Cell cycle analysis in glucose deprivation conditions showed no significant differences between parental and 435-Br1 cells (Supplementary Fig. 2C). To get insight into the pro-survival phenotype of 435-Br1 cells under glucose deprivation we checked the differential expression of 35 genes involved in cell death–related pathways relative to basal glucose conditions (4.5 mg/mL). In contrast to metastatic variants in lymph node (435-N1) and lung (435-L3), 435-Br1 cells overexpressed genes involved in anti-apoptotic pathways such as B-cell lymphoma 2 (BCL-2), cellular inhibitor of apoptosis protein 1 (HIAP1), and X-linked inhibitor of apoptosis protein (XIAP) (Supplementary Fig. 3A) under glucose deprivation. Furthermore, the autophagy-mitophagy genes BNIP3 and BNIP3L, and the pro-apoptotic member of the BCL-2 family BID, were also overexpressed in 435-Br1 cells. Comparative expression of HIAP1 was confirmed by western blot analysis, either in basal or glucose deprivation conditions (Supplementary Fig. 3B), suggesting that the pro-survival advantage observed in BM cells resided on an upregulation of anti-apoptotic pathways. Accordingly, BCL-2 was overexpressed in 435-Br1 versus 435-P cells at 4.5 mg/mL of glucose (Fig. 2A) and these differences were exacerbated in glucose-starved conditions (Fig. 2A).

Fig. 2.

BM cells engage expression of anti-apoptotic proteins and pro-survival autophagy. (A) Immunoblotting of 435-P and 435-Br1 cells at different glucose concentrations (0, 0.5, and 4.5 mg/mL). Semi-quantitative analysis shows significant increase in BCL-2 expression in BM cell lines (BCL-2 vs α-tubulin and normalized to 435-P cells in 4.5 mg/mL glucose). (B) Survival by MTT of 435-P and 435-Br1 cells challenged with 3-methyladenine (10 mM) in hypoglycemic conditions. Error bars represent SD of technical replicates (one representative of 3 experiments). (C) Immunoblotting of 435-P and BM cells (435-Br1 and 435-BrV5) with or without bafilomycin (20 nM, 9 h) and with (4.5 mg/mL) or without glucose. LC3B-II expression quantified vs tubulin and normalized to 435-P cells at 4.5 mg/mL glucose without bafilomycin (one representative of 3 experiments). (D) Survival of metastatic cells treated with bafilomycin (20 nM) with (top) or without glucose (bottom). Bars represent relative survival of each cell line vs its control (nontreated) either with (top) or without (bottom) glucose; error bars represent SD of independent experiments (435-P, Br1 and BrV5 n = 6, L3 n = 5, B1 n = 4). Statistical comparisons vs 435-P cells in each condition (top). *Shows comparison vs 435-P; ^#shows comparison for no glucose+bafilomycin vs no glucose conditions for each cell line. For all panels, significance determined with t-test. *P < 0.05, ***P < 0.001.

Autophagy is activated in response to starvation as an adaptive mechanism that supplies nutrients supporting the metabolic needs of tumor growth in multiple tumor types.²⁹ Our data suggest that overexpression of proteins regulating anti-apoptotic pathways as well as autophagy/mitophagy genes might be mediators of the ERS resistance phenotype. Therefore, we challenged cells with 3-methyladenine, an autophagy inhibitor (Fig. 2B) that blocks phosphatidylinositol-3 kinase. Survival of 435-Br1 cells, especially in hypoglycemic conditions, was decreased with regard to parental cells (60% survival). Furthermore, we tracked autophagic flux by treating cells with bafilomycin (a specific inhibitor of vacuolar H+-ATPase) and assessing accumulation of LC3B-II, a microtubule-associated protein 1 light chain 3 (LC3), essential for autophagosome formation and fusion with the lysosome.³⁰ BM cells expressed more LC3B-II than 435-P cells in basal and hypoglycemic conditions, especially after bafilomycin (Fig. 2C). Indeed, BrV5 cells induced higher levels of autophagy compared with 435-Br1 cells, suggesting that autophagy-mediated metabolic stress resistance is a trait selected in highly metastatic brain cells.

To assess the pro-survival role of autophagy in BM, we treated metastatic cells with bafilomycin in the presence of glucose and measured growth by crystal violet (Fig. 2D, top). Only 20–30% of 435-Br1 and BRV5 cells remained, compared with nontreated conditions, suggesting that autophagy contributes to growth of BM cells basally. In contrast, bone and lung metastatic cells showed a 35‒40% growth, while 50% of 435-P cells survived in contrast to nontreated cells (Fig. 2D, top). These differences increased in glucose-starved conditions (Fig. 2D, bottom). While bafilomycin minimally impaired growth of 435-P cells, it decreased growth of 435-Br1 cells, suggesting a higher reliance on autophagy. In contrast, 30–35% of lung and bone metastatic cells survived, suggesting partial reliance on autophagy.

To explore the importance of GRP94 in the metabolic stress resistant phenotype of BM cells, we used BRV5CA1-eGFP-CMV/Luc cells²³ to stably knock down GRP94 (Fig. 3A, B, Supplementary Fig. 4A, B). We selected short hairpin (sh)GRP94-2 (clone 2) and shGRP94-8 (clone 8), which showed downregulations of 60% and 70%, respectively, in GRP94 and a pool of control clones transfected with the scrambled version of GRP94 (BRV5CA1 CTRL). Consistent with the anti-apoptotic advantage observed in BM cells, GRP94 ablation led to downregulation of BCL-2 with regard to BRV5CA1 CTRL (Fig. 3C). Moreover, TRAF2 expression was slightly diminished in GRP94 ablated cells (Fig. 3D). These results suggested that GRP94 mediates activation of UPR and BCL-2 expression in BM cells. GRP94 deletion decreased LC3B-II expression after bafilomycin treatment with regard to control cells (Fig. 3E), especially in shGRP94-8 cells, which presented the lowest expression of GRP94 (Fig. 3E, right panel). Other functions important for tumorigenesis, like invasion, were not substantially modified by GRP94 (Supplementary Fig. 4C, D). In fact, 435Br1 cells presented lower motility than 435-P cells (Supplementary Fig. 4C), and GRP94 ablation moderately decreased migration with regard to control cells (Supplementary Fig. 4D). Overall, our data show that GRP94 engages BM cells in the ERSR phenotype, mainly by upregulating pro-survival autophagy.

Fig. 3.

GRP94 ablation induces apoptotic proteins and decreases autophagy. Stable GRP94 knockdown clones of BRV5CA1 cells. (A) Immunoblotting of control (BRV5CA1 CTRL) and shGRP94 cells (shGRP94-2 and shGRP94-8). (B) Immunofluorescence analysis of GRP94 expression (yellow) in BRV5CA1 CTRL and shGRP94 cells shows downregulation of GRP94 in clones. 4′,6′-Diamidino-2-phenylindole was used for nucleus visualization (scale bars 100 μm). (C, D) Immunoblotting for BCL2 (C) and TRAF2 (D) after GRP94 deletion. Quantified vs tubulin (C) or actin (D). (E) Immunoblotting for LC3B-II expression in BRV5CA1 CTRL and shGRP94 (quantified vs actin, right). Cells were treated with bafilomycin (20 nM) for the indicated periods of time.

GRP94 Has a Cause-Effect Role in BM Development

To analyze the role of GRP94 in BM settlement in vivo, shGRP94 cells and BRV5CA1 CTRL (control) were orthotopically injected in mice brains by stereotaxis. Tumor burden was periodically assessed by luciferase activity from day 4 after injection until mice showed clinical symptoms (Fig. 4A). Mice injected with BRV5CA1 CTRL cells developed bigger brain masses at a faster rate than mice injected with shGRP94 clones (Fig. 4B), with significant differences in cell burden. BRV5CA1 CTRL–injected mice presented decreased survival and succumbed to the disease in a couple of months (Fig. 4C). In contrast, survival of shGRP94 mice was significantly increased with regard to controls, with 2 mice from each group (2/5) surviving over 120 days (Fig. 4C). Complete remission was observed in 2 mice (2/5) from the shGRP94-8 group, where metastatic cells disappeared from the brain 2 months after stereotaxis (Fig. 4D). Outcomes showed that mice injected with GRP94-ablated cells presented significantly increased survival (Supplementary Fig. 4E). Indeed, the BRV5CA1 CTRL group survived an average of 63 days versus 91 days in shGRP94-8 or 70 days in shGRP94-2 groups. From these experiments we concluded that GRP94 is critical in allowing metastatic cell growth in the brain. Since cell dormancy is an autophagy-dependent process, we hypothesize it might be crucial in BM cells that require metabolic reprogramming toward metastatic progression.

Fig. 4.

GRP94 deletion reduces BM efficiency and LC3B expression in an orthotopic BM model. (A) and (B) Nude mice were intracranially injected with control (BRV5CA1 CTRL) or shGRP94 (clone 2 and 8) cells. Tumor volume was assessed by luciferase activity (total flux, p/s). Error bars represent SEM for n = 5 mice per group. t-Tests were performed at each time point comparing each clone to the control. (C) Mouse survival increases in those injected with shGRP94-2 (P = 0.0105) and shGRP94-8 (P = 0.0033) vs BRV5CA1 CTRL cells. Survival was estimated by Kaplan–Meier method, and log-rank test was used to assess significance. (D) Quantification of total flux (p/s) in a representative shGRP94-8 mouse shows progressive decrease in tumor burden (2 out of 5 mice) correlating with mice turning asymptomatic. (E) H&E staining (10x) and LC3B immunostaining (20x) of brain tissues from mice injected with BRV5CA1 CTRL or shGRP94-8 by stereotaxy (scale bars: 10x and 20x, 100 μm). For all panels, significance determined with t-test. *P < 0.05, **P < 0.01, ***P < 0.001.

Finally, we analyzed LC3B in brain tissues from controls and GRP94-ablated tumors ex vivo (Fig. 4E). Histological analysis showed LC3B underexpression in shGRP94-BM with regard to control animals supporting the mechanistic relationship between GRP94 and autophagy in BM progression.

Pro-Survival Autophagy Is a Common Phenotype in BM

We used a dataset comprising 29 metastases in diverse organs from human breast primary tumors (GSE14017) to determine if autophagy genes were also involved in human breast cancer BM progression. BM tended to cluster together and exhibited a differential autophagy profile from lung and bone metastasis (Supplementary Fig. 5A, Supplementary Table 1). ATF6 was significantly underexpressed in BM tumors with regard to bone and lung, while TRAF2 expression showed a trend toward overexpression that was not significant, probably due to the reduced number of samples (Fig. 5A). These results were partially in agreement with the previous in vitro results that showed underexpression of ATF6 and overexpression of TRAF2 in brain metastatic cells with regard to parental cells (Fig. 1E).

Fig. 5.

Autophagy induction is a common event in BM patients. (A) Plots compare quantification of ATF6 and TRAF2 between patients with brain, lung or bone metastatic progression showing lower ATF6 and trend towards increased TRAF2 expression in BM tumors. (B) LC3B expression by IHC of primary tumors and BM: (a and b) breast carcinoma (a, 10x and b, 20x), arrows indicate necrotic areas (N), (c and d) BM biopsy from breast carcinoma (c, 20x and d, 40x), arrows indicate positive cells (P) in the invasive front of BM, (e) colon BM (20x) with LC3B positive cells in sheltering (S) subpopulations of necrotic areas (N), (f) LC3B expression in endothelial cells from blood vessels (BV) surrounding the tumor in colon adenocarcinoma (20x), (g) sheltering subpopulations (S) in a colon BM sample (20x) and in nerves from primary tumor (h). Representative fields from 3 to 5 pairs analyzed. Scale bars: 10x and 20x, 100 μm; 40x, 25 μm. (C) LC3B and GRP94 expression analysis by IHC of lung primary tumor and its BM: (a) LC3B expression in BM with scattered positive neoplastic cells (P) close to necrotic areas (N) with positive endothelial cells from blood vessels (BV) (10x), (b) GRP94 diffuse staining in the same BM (10x) (c) negative LC3B expression in primary lung carcinoma from the same patient (10x) and (d) negative GRP94 expression in the primary tumor (10x). Scale bars, 100 μm.

To analyze the association of pro-survival autophagy with BM progression from different carcinomas, we analyzed LC3B expression by IHC in breast carcinomas (Fig. 5B, a and b); BM from breast tumors (Fig. 5B, c and d) and colon carcinomas (Fig. 5B, e and f); and BM from colon or clear cell kidney carcinomas (Fig. 5B, g and h, respectively). The results showed: first, that cells expressing LC3B were close to necrotic areas (N) in breast primary tumor (Fig. 5B, a and b) or BM from breast carcinoma (Fig. 5B, c and d), as well as in BM of colon adenocarcinoma (e); second, LC3B positive cells were found in sheltering subpopulations in the necrotic microenvironment (Fig. 5B, e), but also in the invasive front of breast, colon, and kidney BM (Fig. 5B, d, g, and h; respectively); and third, LC3B was expressed in nerves and endothelial cells of blood vessels surrounding the tumor/metastatic subpopulation and some stromal cells, as shown in colon adenocarcinoma (Fig. 5B, f). Similar LC3B expression was found in non–small cell lung carcinomas, with scattered positive cells mainly on BM rather than in primary tumors (Fig. 5C). In contrast to breast carcinomas, GRP94 showed a more diffuse staining pattern, being positive in 6 BM (4 cases from adenocarcinomas and 2 from squamous cell carcinomas), and only in 1 primary lung adenocarcinoma, without association with the treatment. These results indicate that autophagy is a general process required for BM progression of several carcinomas.

Finally, we questioned whether targeting autophagy could be strategic for BM prevention or treatment. Chloroquine and its analogue hydroxychloroquine (HCQ) are inhibitors of lysosomal acidification and autophagosomal degradation.³¹ To evaluate the ability of HCQ treatment to prevent growth, we used a previously characterized⁸ PDX (PDX1110) from a woman with non–small cell lung carcinoma who developed BM. To avoid the effects of the blood–brain barrier (BBB) on drug delivery, we injected these cells subcutaneously and performed a preventive protocol by daily administration of HCQ (50 mg/kg/day), starting 6 days before PDX engraftment (Fig. 6A). The volume of the treated tumors was significantly reduced at day 37, indicating that BM growth might be prevented with HCQ. IHC analysis of GRP94 and LC3B showed decreased LC3B expression in treated PDX with regard to untreated mice (Fig. 6B), suggesting an on-target inhibition of autophagy by HCQ. Moreover, GRP94 levels tended to be unaffected (Fig. 6B). These results suggest a preventive effect of HCQ on BM growth, inhibiting the pro-survival autophagy phenotype that seems to be the conundrum of BM progression.

Fig. 6.

HCQ prevents BM growth in a patient-derived xenograft (PDX) model. (A) Hydroxychloroquine (HCQ, 50 mg/kg/day) or vehicle (saline) were administered daily starting treatment 6 days before PDX back engraftment in left and right sites (n = 5). Graphs show tumor volume (mm³) and SEM, significance was determined with t-test. **P < 0.01, ***P < 0.001. (B) Hematoxylin & eosin (10x, scale bars 100 μm) staining of PDX tissue and IHC analysis of GRP94 (red) and LC3B (green) expression in treated and control mice (20x, scale bars 20 μm). (C) BM cells engage the ERSRP as a consequence of low glucose levels in the brain microenvironment, overexpressing GRP94 and TRAF2, together with anti-apoptotic proteins and pro-survival autophagy. GRP94 depletion inhibits TRAF2 and decreases expression of anti-apoptotic proteins as well as autophagy, impairing BM progression.

Discussion

GRP94 is critical for BM success, since GRP94-ablated cells were unable to progress to BM and in some cases succumbed after a latency period. Therefore, among the different classes of metastasis genes, we conclude that GRP94 might determine the organ-selective nature of BM in part by activating pro-survival autophagy and rendering micro metastases dormant for years. This is in accordance with studies suggesting that BM cells have an enhanced ability to cope with hypoglycemic stress by using alternative fuel sources obtained by increased gluconeogenesis and oxidation of glutamine/branched chain amino acids.³²

Cell populations that are experimentally enriched for metastasis-initiating cells suffer extreme attrition in the organs that they invade.³³ We previously described the ERSRP, where GRP94 was a linker molecule between 34-LMR (laminin receptor) and HSP 27 (heat shock protein 27) activation in response to environmental stress in the brain parenchyma.⁶ We hypothesized that GRP94 may facilitate the metabolic switch from dormancy to growth of BM cells in the low glucose microenvironment, giving rise to organ-specific populations.

Low levels of glucose in the tumor microenvironment have been identified as a selective pressure that drives migration of metastatic cancer cells from the primary site in response to nutrient and growth factor availability.²⁶ The normoglycemic conditions in serum might favor suppression of GRP94 in circulation.³⁴ Further upregulation of GRP94 in cells sustaining ERSRP^12,35 emerges as a hub toward BM by inhibiting apoptosis and activating pro-survival autophagy in hypoglycemic conditions, two major cellular functions¹³ that are coupled to mediate microenvironmental adaptation and growth in the CNS. In fact, BM cells exerted a high ability to oxidize fatty acids in hypoglycemic conditions by overexpressing proteins involved in lipid synthesis and degradation (SREBP-1, LXRα, ACOT7), and we found that GRP94 ablation restored glucose dependence and modulated lipid metabolism to favor BM progression.³⁶

Autophagy is a catabolic process that targets cellular contents to the lysosomal compartment for degradation and is an important cytoprotective mechanism against metabolic or redox stress in cancer cells.²⁹ Autophagy-mediated stress adaptation is crucial in apoptosis-defective cells,³⁷ a characteristic of BM cells. Indeed, we found an upregulation of TRAF2 downstream of GRP94, which correlates with UPR activation (Fig. 6C). The degradation and remodeling of the cellular proteome by Ire1 through TRAF2 during ERS is associated with a cytoprotective attempt to relieve proteotoxic stress by delaying cell death.³⁸ Emerging evidence shows that autophagy promotes not only the survival of dormant cells and dissemination of tumor cells in the circulation, but also adaptation to new stromal interactions at the metastatic foci, activation of dormant cells,³⁹ and growth at metastatic sites.⁴⁰ Interestingly, BM from patients showed some cell populations surrounding central necrotic areas with activated autophagy, which suggests the presence of resistant niches of cells exerting this cytoprotective mechanism. These results are in accordance with others showing that cell dormancy might be autophagy dependent, since autophagy may promote the survival of cells undergoing stress-induced growth.³⁶ This has been shown in a model of lung metastasis where autophagy inhibition prevented establishment of metastasis but did not alter tumor growth once tumors were already formed, suggesting that autophagy dependence differs at different stages of tumor formation.⁴² Moreover, we found expression of LC3B in the active front of the tumor, which is in agreement with the reported role of autophagy in breast cancer progression and reduced survival of patients.⁴³

Further studies are required to understand the relationship between GRP94 overexpression and autophagy. LC3B expression in endothelial capillary cells at the front of BM lesions highlights the contribution of autophagy to angiogenic processes. Autophagy appears critical in neo-angiogenesis mechanisms by acting as a paracrine activator of vascular endothelial growth factor,⁴⁴ which promotes endothelial cell angiogenesis to sustain delivery of nutrients to metastatic cells.⁴⁵ In addition, LC3-positive stromal cells surrounding the tumor suggest a metabolic crosstalk between cancer cells and stromal cells.⁴⁶ Given its role in tumor cell growth, efforts to target autophagy are already under way. In agreement with these observations, genetic and pharmacological autophagy inhibitors overcome molecularly distinct resistance mechanisms, including vasculature normalization that restrains invasion and metastasis.⁴⁷ In fact, autophagy blockade significantly increases the anti-angiogenic therapeutic effect of bevacizumab in glioblastomas.⁴⁸

There is a window where metastatic cell growth in the brain could be restricted with preventive therapies. Prompt disruption of autophagy pathways at the time of primary tumor diagnosis might be an efficient preventive strategy to extend therapeutic benefits and reduce the incidence of BM in cancer patients. In light of these results, and bearing in mind the need for further studies, preventive BM treatments involving autophagy inhibitors or vaccines (or both) might be indicated.

Funding

This study was supported by grants from the Spanish Ministry of Health and Consumer Affairs, FIS-PI10/00057, FIS-PI14/00336, and FIS-PI18/00916, from the I+D+I National Plan with the financial support from ISCIII-Subdirección General de Evaluación and the Fondo Europeo de Desarrollo Regional (FEDER), and by grants from Generalitat de Catalunya 2017_SGR_1305, Fundació Privada Cellex, Fundació Privada Froi, and the AECC (Spanish Association Against Cancer).

Conflict of interest statement. A.S. and R.S. are inventors of the patent US 9,645,150 B2 and EP 2440931 named “Methods for determining the risk for developing brain metastasis, and a kit to carry out said method,” which refers to GRP94’s potential as a biomarker to predict brain metastasis and to target therapies. All other authors have declared that no conflicts of interest exist.

Authorship statement. N.S.C., L.M., R.S-P., and A.S. conceived the study and designed the experiments. R.S-P, N.S.C., A.A., L.P., and A.M.P. performed cell-based experiments. R.S-P. analyzed the transcriptomic human data. L.M. generated the shGRP94 clones and performed in vivo experiments. R.F. and M.B-V. performed in vivo experiments and histology. A.M.C. and J.G. performed and analyzed MLPA experiments. J.G., I.A., I.Al. provided tumor samples. I.A. and I.Al. performed pathological evaluation of IHC techniques. N.S.C. and A.S. analyzed the data and wrote the manuscript with input from all authors. All authors read and approved the manuscript. All authors consent to publication. The authors declare that all the data supporting the findings of this study are available within the article and its Supplementary Material and from the corresponding author upon reasonable request.