Human bone marrow adipocytes drive prostate cancer bone metastasis progression via lipid-mediated induction of Angiopoietin-like 4

  1. Marine Hernandez
  2. Sauyeun Shin
  3. David Rengel
  4. Stéphanie Dauvillier
  5. Nancy Geoffre
  6. Carine Valle
  7. Alessandro Taccini
  8. Mohamed Moutahir
  9. Justine Bertrand-Michel
  10. Nicolas Reina
  11. Catherine Muller
  12. Camille Attané

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

Bone is the main metastatic site in advanced prostate cancer (PCa) and contains bone marrow adipocytes (BMAds), which account for more than 70% of adult bone marrow. However, their role in the progression of PCa metastases remains poorly understood. Herein, we developed a physiologically relevant 3D culture model using primary human BMAds from red hematopoietic rich-areas (rBMAds) and we showed that rBMAds engage in metabolic crosstalk with PCa cells. Specifically, rBMAds release free fatty acids (FFAs) through a non-canonical lipolytic pathway and these FFAs are taken up by PCa cells inducing a transcriptional reprogramming that promotes motility. Among the responsive genes, Angiopoietin like 4 ( ANGPTL4) is the most upregulated via a peroxisome proliferator-activated receptor γ-dependent mechanism, and its silencing abolishes the rBMAd-driven migratory and invasive phenotype. In bone metastases of patients with PCa, high ANGPTL4 levels correlate with poor survival. Together, our findings uncover a novel adipocyte-tumor interaction and identify ANGPTL4 as a key mediator and potential therapeutic target in bone-metastatic PCa.

Article activity feed

  1. Review Commons

    Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    Reviewers' comments:

    Reviewer #1

    Evidence, reproducibility and clarity

    The study entitled "Human bone marrow adipocytes drive prostate cancer bone metastasis progression via lipid-mediated induction of Angiopoietin-like 4" by Hernandez et al. investigates the role of the bone marrow adipocytes (BMAds) from red hematopoietic rich-areas in the progression of prostate cancer (PCa) bone metastases. To do so, the authors used an elegant three-dimensional culture of primary human rBMAds isolated from the femoral bone marrow of patients undergoing hip surgery then cultured with PCa cell lines. The authors claim that PCa cells take up FFAs released by rBMAds which enhance the epithelial-mesenchymal transition and motility through upregulation of ANGPTL4. The article is well written, the figures are very clear and the methodology is perfectly described. To further improve the understanding of this mechanism, the authors should address 1) how PCa cells take up the FFAs, 2) whether PCa cells enhance non-canonical FFAs release and 3) the specificity of the rBMAds in their ability to upregulate ANGPTL4 compared to other adipocytes.

    We thank this reviewer for their positive and constructive feedbacks. Below, we provide detailed responses addressing each of their specific comments either with additional data and/or writing.

    Major comments:

    The paper's conclusions are overall convincing but the following supplementary experiments or discussion points could strengthen their claims.

    1) Is ANGPTL4 upregulation in PCa cells specific to rBMAds or could it also be seen with other adipocytes, such as periprostatic adipocytes or SCAds (used in Fig 2)? Co-culture of PCa cells with SCAds should offer insights.

    More precisely, the article does not state if ANGPTL4 upregulation could be induced prior to bone invasion, by the periprostatic adipose tissue FFA release for example, explaining why ANGPTL4-high cells could be found in bone metastasis. In that case, increased motility and migration could be BMAds independent and trigger the first metastatic events originating from the primary prostatic tumor.

    To address this important question, we performed additional co-culture experiments using adipocytes isolated from human periprostatic adipose tissue (PPAT) and rBMAds. Although PPAT adipocytes also induced ANGPTL4 expression in PCa cells compared with non-cocultivated cells (NC), this induction was approximately 2.5-fold greater in PCa cells co-cultured with rBMAds (new Figure 6A). Consistent with these in vitro findings, analysis of publicly available PCa datasets revealed higher ANGPTL4 expression in bone metastases than in primary tumors (new Figure 6B). Together, these results suggest that although ANGPTL4 expression may be initiated within the primary tumor microenvironment, it is markedly amplified following colonization of the bone marrow niche.

    In addition, as suggested by Reviewer 3, we assessed the clinical relevance of ANGPTL4 expression at different stages of disease progression. High ANGPTL4 expression in primary PCa tumors was not associated with either overall or disease-free survival (new Figure 6C). In contrast, elevated ANGPTL4 expression in bone metastases was significantly associated with poorer survival (new Figure 6D), indicating that the clinical relevance of ANGPTL4 is primarily linked to the metastatic bone microenvironment rather than the primary tumor. We therefore propose that, although ANGPTL4 expression may be initiated at the primary tumor site, it is markedly amplified within the bone marrow adipocyte-rich niche, where it promotes PCa cell migration and metastatic outgrowth. We have revised the manuscript accordingly (see lines 559-583).

    2) The article hypothesizes an ANGPTL4-associated increase of motility and migration of PCa cells upon interaction with bone marrow adipocytes in the first metastatic bone, inducing enhanced invasion of said bone and secondary metastasis sites. However, the article lacks evidence that ANGPTL4 is also upregulated in cells from secondary metastasis sites in patients. It would be interesting to explore in the existing data set the expression of ANGPTL4 in secondary metastasis of patients who had previous bone metastasis. These points could be addressed in the concluding remarks.

    We thank the reviewer for this interesting comment. Bone represents 80% of prostate cancer metastases (Gandaglia et al, Prostate, 2014, PMID: 24132735) and the bone metastases have been shown to act as a source of further metastatic dissemination (Gundem et al, Nature, 2015, PMID: 32728210; Hong et al, Nature commun, 2015, PMID: 25827447). Unfortunately, currently available public datasets only provide the anatomical site of metastatic lesions and do not distinguish between primary and secondary metastatic sites. Nevertheless, we examined ANGPTL4 expression in metastatic lesions. ANGPTL4 expression is maintained in liver, lung and lymph nodes (new Supplementary Figure 6A) suggesting that it might contribute to metastatic dissemination beyond bone. The manuscript has been revised accordingly (lines 575–580).

    Additional experiments or qualifying the claims:

    3) Fig 1 : It would be interesting to have an insight on the mechanism by which FFAs are uptaken by PCa.

    Several evidences demonstrate that membrane-associated transport proteins facilitate and regulate this process, including CD36 (fatty acid translocase) and members of the fatty acid transport protein family (FATPs/SLC27A1–5). We evaluated the expression of these transporters implicated in the uptake of long-chain FFAs in our cellular models. CD36 expression was not detected at either the mRNA or protein level in all three prostate cancer cell lines (PC3, Du145, and LNCaP). The validity of our detection strategy was confirmed using a CD36-positive control (U937 leukemic cell line) and a corresponding negative control (U937 cells KO for CD36) (Supplementary Fig. F and G added). Among the FATP family members, FATP4 (SLC27A4 gene) was the most highly expressed transporter in all the three PCa cell lines (Supplementary Fig. 1H, added). We therefore investigated its potential contribution to FFA uptake by silencing SLC27A4 expression using siRNA in PC3 cells and knockdown efficiency was confirmed at both mRNA (Supplementary Fig. 1I added) and protein (Supplementary Fig. 1J added) level. Silencing of FATP4 reduced the uptake of exogenous fluorescent oleate (Supplementary Fig. 1K) but not palmitate (Supplementary Fig. 1L, added) showing a selectivity of FFA uptake depending on their nature. However, FATP4 knockdown did not affect lipid transfer in the co-culture model (Supplementary Fig. 1M, added), where several FFA species are present. Collectively, these findings indicate that the uptake rBMAd-derived FFAs is unlikely to be primarily mediated by FATP4 suggesting that alternative mechanisms, such as extracellular vesicle-mediated transfer or passive diffusion, may account for their uptake. These results have been included in Supplementary Figure 1F–M, and the manuscript has been revised accordingly (lines 396–411).

    4) Fig 1D : It would be interesting to add an extra control of NC PCa treated with BODIPY to measure basal BODIPY uptake of PCa

    To address this point, we performed an additional control experiment in which PC3 cells were directly exposed to 5 µM BODIPY FLC16, corresponding to the concentration used to preload rBMAds in the co-culture experiments (Fig. 1D). As expected, direct exposure resulted in higher BODIPY FL C16 uptake compared with co-culture with preloaded rBMAds.

    This difference is expected since, during direct treatment, the fluorescent FFA is immediately available at 5 µM in the culture medium. In contrast, BODIPY FLC16 represents only a tracer when rBMAds are preloaded, and the released of FFAs during co-culture consist of a mixture of fluorescent and non-fluorescent species. The figure can be included in the manuscript upon referee’s request.

    5) Fig 2 : The figure efficiently states that rBMAds undergo non-canonical lipolysis in comparison to SCAds, without PCa cells. Glycerol dosage and/or pan-lipase inhibitor treatment of rBMAds in co-culture with PCa would confirm rBMAds still show non-canonical lipolysis in bone metastasis environment.

    To address this point, we quantified glycerol and FFAs released into the culture medium from PC3 cells alone or from rBMAds co-cultured with PC3 cells, in the presence or absence of the pan-lipase inhibitors paraoxon ethyl or orlistat. As shown in the new Supplementary Figure 2B, pan-lipase inhibition markedly reduced FFA release, whereas glycerol release remained unchanged. These results confirm that, even in the presence of PCa cells, FFA release from rBMAds occurs independently of canonical lipolysis.

    6) Fig2 : Is ANGPTL4 upregulation in PCa cells specific to rBMAds or could it also be seen with other adipocytes, such as periprostatic adipocytes or SCAds ? Co-culture of PCa cells with SCAds should offer insights.

    This comment was addressed in point 1.

    7) Fig 2 : Could the authors specify if there is a possibility the lipolysis kinetics is slower in rBMAds instead of an actual incomplete lipolysis?

    To determine whether the reduced glycerol release reflected delayed rather than incomplete lipolysis, we extended isoproterenol stimulation up to 24 h in rBMAds, using SCAds as a control. As shown in the new Supplementary Figure 2A (commented in the text, lines 425–430), glycerol release did not increase over time in rBMAds, whereas it progressively increased in SCAds. These results do not support delayed lipolysis in rBMAds but rather indicate that lipolysis remains incomplete despite prolonged stimulation.

    7) Fig 3B/C : The author should provide tables to show : 1) all significantly up and down GO term 2) the top 20 genes DEG 3) any GO terms/genes related to EMT (table 1 should be associated with figure 3)

    As proposed by the reviewer, we added a Table 1 in the revised manuscript with all significant upregulated and downregulated GO terms including gene names and the proportion of genes related to each GO term and p-value. We also added the top 20 of upregulated and downregulated genes.

    8) Fig 5 : To complete experiences of KD of ANGPTL4 and PPARy inhibition, the author should also include the mRNA levels of EMT associated genes in these conditions.

    We thank the reviewer for this suggestion. The EMT-associated genes originally included in the manuscript were selected based on the literature in pancreatic cancer as descriptive markers (Gordon et al, BMC Cancer, 2023, PMID: 37291514). In the revised manuscript, to further investigate the mechanisms underlying ANGPTL4-dependent cell migration, we reconstructed an ANGPTL4-associated gene network in Cytoscape using RNA-seq-derived differentially expressed genes enriched in cell migration GO terms and their interactions with ANGPTL4. Interestingly, the gene signature identified in our conditions do not clearly overlaps with canonical EMT-associated genes described in the literature. This analysis revealed a highly interconnected network rather than a linear signaling cascade (new Supplementary Figure 5L) suggesting that ANGPTL4 promotes cell migration by orchestrating a coordinated transcriptional program involving multiple complementary signaling pathways. While this systems-level analysis reveals a coherent migration-associated network, it does not aim to establish a direct functional role for each individual gene, but rather to define the global transcriptional context of ANGPTL4-dependent migration. In line with this, we added and discussed the Cytoscape network of ANGPTL4-associated migration genes in the revised manuscript (new Supplementary Figure 5L and commented in the text lines 538-545).

    To better the clarity of the paper, the author should indicate in table S1 which patients are used in which figures. Also, what is the rationale for not using female patients in all the experiments?

    We thank the reviewer for this helpful suggestion. We have revised Table S1 to improve its clarity by indicating, for each experiment, the corresponding patient information, including age, sex, and BMI. In this study, rBMAds from both male and female donors were used to establish and validate the 3D culture model and to characterize the mechanisms of FFA release. However, for the co-culture experiments investigating the crosstalk between rBMAds and PCa cells, only rBMAds from male donors were used. Since PCa is a male-specific disease, restricting these experiments to male donors minimizes the potential confounding effects of sex-dependent differences in BMAd biology and ensures that the model remains physiologically relevant.

    The suggested additional experiment should be achievable in 6 months.

    *9) Statistical analyses: The authors state in the Materiel and Methods that they used Student's t-test, one-way ANOVA or two-way ANOVA. These tests are applied to parametric data sets, validated for their normal distribution. Usually, n To determine the appropriate statistical test and post hoc analysis, data normality was first assessed using the Shapiro-Wilk test. Parametric tests were applied when data followed a normal distribution, whereas non-parametric tests were used when normality assumptions were not met. The specific statistical tests used for each analysis have now been indicated in the corresponding figure legends.

    Minor Comments:

    10) In Fig 1F, the author should discuss the discrepancy between the significant increase of TG uptake in PCa shown in 1B and the non-significant effect seen in 1F, given that TG represents 85% of the total FFAs shown in 1E and 1F.

    Figures 1B and 1F correspond to two distinct experimental readouts and therefore should not be directly compared. Figure 1B measures the intracellular accumulation of TG in PCa cells following coculture with adipocytes. These TG are formed after the uptake of adipocyte-derived FFAs and their subsequent re-esterification into neutral lipids stored in lipid droplets. In contrast, Figure 1F quantifies the concentration of FFAs remaining in the culture medium at the end of the coculture period. Although extracellular FFA levels appear slightly higher in the presence of PCa cells than with rBMAds alone, this difference is not statistically significant. This is likely because FFAs released by rBMAds are continuously taken up by PCa cells during coculture, thereby limiting their accumulation in the medium. Therefore, the lack of a significant increase in extracellular FFAs does not contradict the marked intracellular accumulation of TG observed in Figure 1B.

    *11) Fig 3 : In Fig 3G and 4E, staining quantification would be necessary. *

    The quantification of the area of lipid droplets per nucleus depending on exogenous FFA treatment was done according to reviewer comments and added in a Figure 4E. Concerning Figure 3G, the images shown were intended as representative illustrations of the cytoskeletal and morphological changes observed following co-culture, rather than as a quantitative endpoint. The conclusions of this section are primarily supported by the quantitative functional assays measuring cell migration (Fig. 3D-E) and invasion (Fig. 3F), which directly assess the biological consequences of these morphological changes.

    12) In Fig3/Sup Fig3, why is the invasion assay not shown for LNCaP cell line ?

    According to reviewer comments, invasion assay was performed on LNCaP and added in the manuscript (new Supplementary Fig. 3D) on the revised manuscript. Coculture of rBMAds with LNCaP stimulate invasion as other PCa cell lines (PC3 and Du145).

    *13) 5F: Is there a reason why the sum of FFA has a stronger effect than the individual ones ? *

    We observed that the combination of palmitate, oleate, and linoleate induced a stronger increase in lipid accumulation (Figure 4E, only statistically significant when compared palmitate alone with the mix) and target gene expression (Figure 5F) than each FA alone. Although we did not investigate the underlying mechanism in the present study, we speculate that the combination is likely to better mimic the FA environment of rBMAds and may promote more efficient downstream signaling than individual FA. This aspect is also discussed in response to the point 4 of the referee 2.

    14) 5E/F : Rationale for changing FFA treatment duration between 5E and 5F (72h then 24h) ?

    While most FFA treatment and co-culture experiments with rBMAds were performed for 72 h, experiments involving ANGPTL4 siRNA or the PPARγ inverse agonist T0070907 were limited to 24 h. This shorter duration was chosen to preserve the efficacy of T0070907 and to capture the optimal window of siRNA-mediated knockdown. Importantly, we observed that FFA treatment was already sufficient to induce ANGPTL4 expression after 24 h, allowing us to investigate the contribution of ANGPTL4 and PPARγ signaling under conditions where the pharmacological inhibition and gene silencing remained effective. We have clarified this rationale in the revised manuscript (lines 255-256).

    *15) For the discussion: the authors clearly describe the two different BMAds subtypes (cBMAds vs rBMAds), would it be interesting to question the role of cBMAds in PCa progression as well ? *

    We thank the reviewer for this suggestion and agree that this is an important point to address. Because PCa bone metastases predominantly localize to red bone marrow areas, our study focused on the role of rBMAds in PCa progression. Nevertheless, we cannot exclude a potential contribution of constitutive bone marrow adipocytes, which reside primarily in yellow bone marrow, and whose effects on PCa cells may be similar to or distinct from those of rBMAds. We have now included this point in the concluding remarks of the Discussion (lines 599–601).

    The studies are referenced appropriately.

    *16) Please, could the authors indicate the co-culture duration with rBMAds in the legends of Figure 5. Excepted this point, the text and figures are clear and accurate. *

    We added in the revised manuscript the duration time of coculture for the different experiments in Figure 5.

    Significance

    General assessment:

    the authors used an elegant three-dimensional culture of primary human rBMAds isolated from the femoral bone marrow of patients undergoing hip surgery then co-cultured with PCa cell lines. The authors claim that PCa cells take up FFAs released by rBMAds which enhance the epithelial-mesenchymal transition and motility through upregulation of ANGPTL4. The most important aspects of this study are 1) the description of the non-canonical FFAs release from rBMAds, 2) the induction of EMT/motility of PCa cells induced by FFAs uptake. To further improve the understanding of this mechanism, the authors should address 1) how PCa cells take up the FFAs, 2) whether PCa cells enhance non-canonical FFAs release and 3) the specificity of the rBMAds in their ability to upregulate ANGPTL4 compared to other adipocytes.

    This study extends the knowledge in the field of prostate cancer metastasis.

    The authors have developed an elegant 3D culture model of primary human rBMAds, which is highly relevant for investigating the role of adipocytes in cancer cell behavior. Indeed, most models used in the literature rely on adipocytes differentiated from bone marrow mesenchemal stem cells. Although adipocytes differentiation models have recently been improved by incorporating 3D structure, the model developed by the authors is most physiological, as it is based on the direct use of primary bone marrow adipocytes. In this way, this model represents a valuable tool to advance our understanding of mechanisms involved in bone metastasis in breast/prostate cancers.

    The type of audience that will be interested by this research include both basic and specialized researchers. Moreover, this study should improve 1) the understanding of the impact of rBMAds on prostate cancer behavior and 2) the robustness and physiological relevance of adipocyte-based studies.

    Our expertise lies in the interaction between normal or pathological hematopoiesis with the components of bone marrow microenvironment.

    We sincerely thank the reviewer for the constructive comments and suggestions, which have significantly contributed to improving the mechanistic depth and clarity of our manuscript. We believe that the additional data and revisions provided in this response address the main points raised and further strengthen the conclusions of our study.

    Reviewer #2

    Evidence, reproducibility and clarity

    Summary: This manuscript focuses on evaluation of effects of bone adipocytes on prostate cancer and their potential role in prostate cancer bone metastases progression. Mechanistically, the results show that fatty acid (FFA) secreted by bone marrow adipocytes increased expression of ANGPTL4 in prostate cancer cell lines in vitro, and this increase was dependent on PPARgamma signaling, resulting in increased migration and invasiveness of the cells. Evaluation of data from SU2C data set, was also included to support the conclusions, showing that increased levels of ANGPTL4 expression in bone metastases are negatively correlated with patients' survival, suggesting that ANGPTL4 could be a therapeutic target for prostate cancer bone metastasis.

    The authors previously investigated and published on crosstalk/interaction between periprostatic adipocytes and prostate cancer cells, where FFAs induced released by adipocytes elicited increased aggressiveness and dissemination of prostate cancer cells, with mechanisms involving increase in NOX5 expression. However, specifically bone marrow adipocytes effect on prostate cancer cell have not been investigated in details. The experiments are logically sequenced and data clearly presented, in general supporting the conclusions that FFAs secreted by bone marrow adipocytes affect prostate cancer cells migration and invasiveness in vitro.

    *Major comments: *

    1) The authors state: "These data collectively demonstrate that PCa cells can trigger the release of FFAs from rBMAds cultured in 3D (Fig 1), which are then rapidly taken up and re-esterified into TGs by cancer cells." There are no data presented to show that prostate cells trigger the FFAs release from adipocytes- the data only show that FFA from adipocytes are taken up by prostate cancer cells. Use of condition media in combination with the coculture experiments would address this point. It is not clear how the authors came to the following conclusions: " In co-culture, the levels of these FFAs tended to increase (Fig. 1E), a trend also observed in total FFA content (Fig. 1F). The data presented in this figure have no statistical significance to support this statement, and moreover, the relative abundance in the PC3 COC appears to be simply addition of the abundance of PC NC and rBMAds NC, no alterations of secretion.

    We agree that our data do not directly demonstrate that prostate cancer (PCa) cells induce the release of FFAs from rBMAds. Rather, our results show that FFAs present in the co-culture system are efficiently taken up and re-esterified into triglycerides by PCa cells (Fig. 1B-D). Accordingly, we have revised the manuscript to attenuate our interpretation and avoid implying a causal effect of PCa cells on induction of FFA release by rBMAds. We have also modified the corresponding text describing Figures 1E and 1F to underline that there are no changes in FFA present in the culture medium between rBMAds alone or cocultivated with cancer cells (lines 392-395).

    2) GO analysis presented in this manuscript indicates that prostate cancer cells grown in bone marrow adipocytes condition media exhibit increased expression of gene sets associated with cell migration, cell adhesion, etc. were significantly induced under these conditions (Fig 3C). However, it seems contradictory that while locomotion and invasion were increased so were the gene associated with adhesion, as one would speculate that with increased invasion and mobility you will not have increased adhesion. This analysis also show that downregulated gene sets were mainly associated with proliferation regulation. However, the data included in Fig 3 H and I and supp figure show no effects on proliferation of the prostate cancer cells. This discrepancy should be discussed.

    We agree with the reviewer that the concomitant upregulation of the “cell adhesion” and the “cell migration” GO terms may appear contradictory. However, genes annotated within the “cell adhesion” category are not exclusively involved in stable cell–cell adhesion; many of them like integrins subunits (ITGA3, 5, 6, ITGB1, 5) or actin cytoskeleton genes (ACTB1, ACTG1, ACTN1, ACTN4) also participate in the dynamic remodeling of cell–matrix and cell–cell interactions that are required for efficient cell migration. In this context, increased expression of adhesion-related genes can be consistent with enhanced migratory capacity. Similarly, the “regulation of cell population proliferation” GO term includes both pro- (ID2, LRP5) and anti-proliferative genes such as cell cycle inhibitor genes (CDKN2A, CDKN2B, CDKN2C, CDKN3, TGFB3, TGFBR2) reflecting a complex and balanced transcriptional response. This may explain why no significant changes in prostate cancer cell proliferation were observed in functional assays under co-culture conditions (Fig. 3H–I). As requested by the Reviewer 1, the list of genes associated with each GO term has now been added in Table 1 and we modify the manuscript accordingly (line 456-460).

    3) Some of the genes associated with effects of adipocytes condition media or treatment with FFAs were confirmed by PCR and western blot in PC3 cells, but not in DU145 or LNCaP. These experiments should be done, and results in the other two cell lines should be included to demonstrated the generality of the mechanism of the observed effects. This is important specifically, as LNCaP cells is the only cell line used that expresses androgen receptor that is the hallmark of prostate cancer, and majority of bone metastasis express androgen receptor (PC3 and DU145 do not). Similarly, no migration results under coculture of LNCaP and rBMAds were included.

    We would like to clarify that several of the points raised are supported by data already included in the manuscript. Specifically, we showed that both coculture with rBMAds and treatment with exogenous FFAs increase ANGPTL4 expression in PC3 cells (Fig. 5B–F) as well as in Du145 cells (Supplementary Fig. 5A–E). In contrast, although LNCaP cells exhibited increased migration (Supplementary Fig. 3B) and invasion (new Supplementary Fig. 3D) following coculture with rBMAds, ANGPTL4 expression was undetectable in this cell line under both basal conditions and after coculture (data not shown). These findings indicate that the enhanced migratory and invasive properties of LNCaP cells are mediated through an ANGPTL4-independent mechanism, highlighting the biological heterogeneity of prostate cancer and indicating that not all cell lines recapitulate the same molecular pathway. The manuscript was revised accordingly (lines 514-518). Importantly, the relevance of ANGPTL4 is further supported by our analysis of clinical prostate cancer datasets, which showed increased ANGPTL4 expression in bone metastases in comparison to primary sites and its association with poor prognosis in metastatic disease (see response to Reviewer 1 points 1 and 2 and new Fig. 6).

    4) An important question concerns the clinical relevance of the observed effects of FFAs. In the in vitro experiments mainly migration and invasion, FFAs were used at 10 µM, including conditions in which three different FFAs were combined, each at a concentration of 10 µM. Notably, the magnitude of the effects was similar when FFAs were applied individually or in combination, suggesting possible saturation of the signaling pathway. Therefore, lower FFA concentrations, such as those found in the bone marrow microenvironment should be used to better reflect physiological conditions. Moreover, the FFAs were measured in adiposities condition media but only relative abundance was included in figure 1E, not actual levels.

    We agree that the clinical relevance of the FFA concentrations used should be carefully discussed. We would first like to clarify that all experiments with exogenous FFAs were performed using 100µM of each FFA (palmitate, oleate, or linoleate), including the mixture in which each FFA was added at 100µM. The 10 µM concentration mentioned in the review appears to result from a misunderstanding. As the reviewer points out, no additive effect was observed when the three FFAs were combined compared with individual treatments in the migration (Fig. 4F and Supplementary Fig. 4A–B) and invasion assays (Fig. 4H), suggesting that the signaling pathways involved may already be saturated under these experimental conditions.

    To address the reviewer's concern regarding physiological relevance, we estimated the amount of FFAs released by rBMAds after 3 days of culture to be approximately 100µM total FFAs. Lipidomic analysis (Fig. 1E) further showed that palmitate, oleate, and linoleate account for approximately 20%, 30%, and 20% of the total FFAs, corresponding to physiological concentrations of approximately 20µM palmitate, 30µM oleate and 20µM linoleate, respectively__. __We therefore performed additional migration experiments using these physiological concentrations. Under these conditions, treatment with each individual FFA did not significantly affect PC3 cell migration. In contrast, the combination of the three FFAs at their physiological concentrations significantly increased migration, suggesting that although individual FFAs are insufficient at physiological levels, their combined presence reaches a threshold required to promote prostate cancer cell migration. We are ready to include theses results in the manuscript upon referee request.

    *5) The authors used SU2C data sets to evaluate ANGPTL4 expression in bone metastases and associated increased expression worse better survival. However, in this data set, it is clear that ANGPTL4 exhibits significantly higher expression in liver metastases than in bone metastasis. Some discussion about this aspect would increase the potential translational indication. *

    We thank the reviewer for this important comment (also raised by Reviewer 1, point 2). While approximately 80% of prostate cancer metastases are located in bone, these metastases can acquire the capacity to generate secondary metastatic sites and contribute to further metastatic dissemination (Gundem et al, Nature, 2015, PMID: 32728210; Hong et al, Nature Communications, 2015, PMID: 25827447). Analysis of prostate cancer metastatic datasets, in which ANGPTL4 expression is assessed across different metastatic sites, revealed that ANGPTL4 expression is maintained and even increased in metastases located at secondary sites, such as liver and lung, compared with bone metastases (new Supplementary Fig. 6A). These findings suggest that high ANGPTL4 expression is not restricted to the bone microenvironment but may persist or become amplified during metastatic progression and contribute to late-stage metastatic dissemination. These results have been added to the revised manuscript and discussed accordingly (lines 575-580).

    Minor comments:

    6) Preparation of the gel description stipulates "One hundred microliters of SCAds or rBMAds were added and homogenized quickly with 100µL of thrombin....." Number of cells added for normalization should be added to the methods.

    We have now included in the method part (lines 137-138), the estimated number of adipocytes within the gel, based on cell counts performed in a subset of samples (Shin et al, Cell Reports, 2026, PMID: 42348417).

    7) In this sentence- is it correct to assume that oleate is also uM and not uL? "One day after seeding, cells were treated with 100μM palmitate (Cayman #29558), 100μL oleate (Sigma Aldrich #O3008), 100μM linoleate (Sigma Aldrich #L9530), or with the mix of 3 FFAs (100μM palmitate/100μM oleate/100μM linoleate) during 3 days.

    We agree with the reviewer, it is a mistake. It is a concentration of oleate and not a volume. It was corrected in the manuscript by 100µM oleate (line 215).

    8) There is a description of an experiment: "PC3 and Du145 were transiently transfected with ANGPTL4 siRNA pool (final concentration 25nmol/L) using ON-TARGETplus SMARTpool (Thermo Scientific Dharmacon) or ON-TARGETplus nontargeting pool used as a control (Thermo Scientific Dharmacon). Transfection was done according to the manufacturer's instructions with Lipofectamine RNAiMAX (Invitrogen Life Technologies) followed by 24h of coculture with rBMAds. After coculture, a second transfection was done and gene extinction, migration and invasion were evaluated." The rationale for second transfection (assuming still with siRNA) has not been provided and it is not clear.

    We agree with the reviewer that clarification is needed. A second siRNA transfection was performed 24 h after coculture with rBMAds to maintain ANGPTL4 knockdown during the subsequent 24 h migration and invasion assays. For more clarity, the section materials and methods was corrected accordingly (lines 255-256).

    9) Figure 6. Survival: HR and number of patients at each time point should be added to the graph. Some English editing is needed

    The number of patients in each section “low ANGPTL4 expression” (n=8) and “high ANGPTL4 expression” (n=8) was added in the legend of the figure 6D (line 705).

    Significance

    There are multiple studies focusing on adipocytes and FFAs effects on cancer, in general, as well as on prostate cancer specifically. Effects on migration and invasiveness are well documented in literature. The novel aspect of the current manuscript is specifically looking at red bone marrow adipocytes, but it is not clear that the effects of FFAs would be different than when other adipocytes are used.

    We agree that adipocyte-derived fatty acids have been implicated in cancer progression in a large number of study as we recently reviewed (Attané and Muller Trends in Cancer, 2020, PMID: 32610069). However, the novelty of our study is that it demonstrates that the origin and metabolic identity of adipocytes determine the prostate cancer (PCa) response. Most previous studies have focused on adipocytes from primary tumor sites, while studies investigating bone metastasis have mainly relied on in vitro-differentiated bone marrow adipocytes (like OP9) or mesenchymal stem cell-derived models, which do not fully reproduce the properties of mature human bone marrow adipocytes (Shin et al, Cell Reports, 2026, PMID: 42348417).

    Using primary human rBMAds, we show that bone marrow adipocytes display specific biological features, including a non-canonical lipolytic pathway, and induce a distinct molecular response in PCa cells compared with periprostatic adipocytes. Specifically, rBMAds induce stronger ANGPTL4 expression, whereas NOX5, previously identified by our team as a mediator of PPAT adipocyte-induced effects (Laurent et al, Mol Cancer Res, 2019, PMID: 30606769), is not increased following rBMAd coculture. This demonstrates that adipocytes from different anatomical niches are not functionally equivalent and activate distinct tumor signaling pathways. The relevance of ANGPTL4 is further supported by our analysis of clinical PCa datasets, which showed increased ANGPTL4 expression in bone metastases in comparison to primary sites and its association with poor prognosis in metastatic disease.

    Reviewer #3

    (Evidence, reproducibility and clarity (Required)):

    The study identifies a role for free fatty acids released by bone marrow adipocytes from red hematopoietic rich-areas in promoting prostate cancer cell migration and viability by inducing the expression of ANGPTL4. To achieve this the authors use to cancer cell-lines (LNCaP, PC3 and DU145) and a coculture model for their phenotypic assays and perform RNA-seq on one line (n=5 per condition) subjected to co-culture activation or not. The selection of ANGPTL4 is supported by its significant association with shorter survival times in clinical RNA-seq data from a metastasis cohort (SU2C) when highly expressed.

    *1) The paper is concise and clear. The authors also quantify free fatty acid release and neutral lipid accumulation. It would be interesting to have a more comprehensive analysis (pathway and coexpression) of the RNA-seq data with RT-PCR validation of a number of significant transcripts that are up and downregulated. *

    In the revised manuscript, we further strengthened the analysis of the RNA-seq data by reconstructing an ANGPTL4-associated gene network in Cytoscape using RNA-seq-derived differentially expressed genes enriched in cell migration GO terms and their interactions with ANGPTL4. This analysis revealed a highly interconnected network rather than a linear signaling cascade (Supplementary Fig. 5L) suggesting that ANGPTL4 promotes cell migration by orchestrating a coordinated transcriptional program involving multiple complementary signaling pathways. While this systems-level analysis reveals a coherent migration-associated network, it does not aim to establish a direct functional role for each individual gene, but rather to define the global transcriptional context of ANGPTL4-dependent migration. In line with this, we added and discussed the Cytoscape network of ANGPTL4-associated migration genes in the revised manuscript (lines 538-545).

    2) Also a survival analysis in a localised PCa cohort using publicly available RNA-seq data.

    As suggested, we assessed the clinical relevance of ANGPTL4 expression at different stages of disease progression. High ANGPTL4 expression in primary PCa tumors was not associated with either overall or disease-free survival (new Figure 6C). In contrast, elevated ANGPTL4 expression in bone metastases was significantly associated with poorer survival (Figure 6D), indicating that the clinical relevance of ANGPTL4 is primarily linked to the metastatic bone microenvironment rather than the primary tumor. We have revised the Discussion accordingly (lines 569–575).

    *(Significance (Required)): * Explores an interesting biological question but in a limited number of pre-clinical models. Interesting as a startpoint for further studies.

    Prostate cancer research is inherently limited by the availability of representative experimental models, with only a limited number of well-characterized prostate cancer cell lines compared with other cancer types (Cunningham et al, J Biol Methods, 2015, PMID: 26146646). The major prostate cancer models used in the field, PC3, Du145, and LNCaP, were included in our study. Importantly, our study goes beyond classical cancer cell line models by incorporating ____primary human rBMAds____, which remain rarely used due to their limited accessibility and the technical challenges associated with their culture. Using this physiologically relevant model, we demonstrate that adipocyte origin and metabolic identity critically influence the tumor response. Indeed, rBMAds display specific biological properties, including a non-canonical lipolytic pathway, and induce a distinct molecular response in prostate cancer cells compared with periprostatic adipocytes. Specifically, rBMAds induce stronger ANGPTL4 expression and elevated ANGPTL4 expression in bone metastases was significantly associated with poorer survival (Figure 6D). In contrast, high ANGPTL4 expression in primary PCa tumors was not associated with either overall or disease-free survival (new Figure 6C) indicating that the clinical relevance of ANGPTL4 is primarily linked to the metastatic bone microenvironment rather than the primary tumor. Together, these results support the biological and translational relevance of our model and identify a bone marrow adipocyte-specific signaling axis involved in prostate cancer progression.

  2. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    The study identifies a role for free fatty acids released by bone marrow adipocytes from red hematopoietic rich-areas in promoting prostate cancer cell migration and viability by inducing the expression of ANGPTL4. To achieve this the authors use to cancer cell-lines (LNCaP, PC3 and DU145) and a coculture model for their phenotypic assays and perform RNA-seq on one line (n=5 per condition) subjected to co-culture activation or not. The selection of ANGPTL4 is supported by its significant association with shorter survival times in clinical RNA-seq data from a metastasis cohort (SU2C) when highly expressed.

    The paper is concise and clear. The authors also quantify free fatty acid release and neutral lipid accumulation. It would be interesting to have a more comprehensive analysis (pathway and coexpression) of the RNA-seq data with RT-PCR validation of a number of significant transcripts that are up and downregulated. Also a survival analysis in a localised PCa cohort using publicly available RNA-seq data.

    Significance

    Explores an interesting biological question but in a limited number of pre-clinical models. Interesting as a startpoint for further studies.

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    Summary: This manuscript focuses on evaluation of effects of bone adipocytes on prostate cancer and their potential role in prostate cancer bone metastases progression. Mechanistically, the results show that fatty acid (FFA) secreted by bone marrow adipocytes increased expression of ANGPTL4 in prostate cancer cell lines in vitro, and this increase was dependent on PPARgamma signaling, resulting in increased migration and invasiveness of the cells. Evaluation of data from SU2C data set, was also included to support the conclusions, showing that increased levels of ANGPTL4 expression in bone metastases are negatively correlated with patients' survival, suggesting that ANGPTL4 could be a therapeutic target for prostate cancer bone metastasis. The authors previously investigated and published on crosstalk/interaction between periprostatic adipocytes and prostate cancer cells, where FFAs induced released by adipocytes elicited increased aggressiveness and dissemination of prostate cancer cells, with mechanisms involving increase in NOX5 expression. However, specifically bone marrow adipocytes effect on prostate cancer cell have not been investigated in details. The experiments are logically sequenced and data clearly presented, in general supporting the conclusions that FFAs secreted by bone marrow adipocytes affect prostate cancer cells migration and invasiveness in vitro.

    Major comments: The authors state: "These data collectively demonstrate that PCa cells can trigger the release of FFAs from rBMAds cultured in 3D (Fig 1), which are then rapidly taken up and re-esterified into TGs by cancer cells." There are no data presented to show that prostate cells trigger the FFAs release from adipocytes- the data only show that FFA from adipocytes are taken up by prostate cancer cells. Use of condition media in combination with the coculture experiments would address this point.

    It is not clear how the authors came to the following conclusions: " In co-culture, the levels of these FFAs tended to increase (Fig. 1E), a trend also observed in total FFA content (Fig. 1F). The data presented in this figure have no statistical significance to support this statement, and moreover, the relative abundance in the PC3 COC appears to be simply addition of the abundance of PC NC and rBMAds NC, no alterations of secretion.

    GO analysis presented in this manuscript indicates that prostate cancer cells grown in bone marrow adipocytes condition media exhibit increased expression of gene sets associated with cell migration, cell adhesion, etc. were significantly induced under these conditions (Fig 3C). However, it seems contradictory that while locomotion and invasion were increased so were the gene associated with adhesion, as one would speculate that with increased invasion and mobility you will not have increased adhesion. This analysis also show that downregulated gene sets were mainly associated with proliferation regulation. However, the data included in Fig 3 H and I and supp figure show no effects on proliferation of the prostate cancer cells. This discrepancy should be discussed.

    Some of the genes associated with effects of adipocytes condition media or treatment with FFAs were confirmed by PCR and western blot in PC3 cells, but not in DU145 or LNCaP. These experiments should be done, and results in the other two cell lines should be included to demonstrated the generality of the mechanism of the observed effects. This is important specifically, as LNCaP cells is the only cell line used that expresses androgen receptor that is the hallmark of prostate cancer, and majority of bone metastasis express androgen receptor (PC3 and DU145 do not). Similarly, no migration results under coculture of LNCaP and rBMAds were included.

    An important question concerns the clinical relevance of the observed effects of FFAs. In the in vitro experiments mainly migration and invasion, FFAs were used at 10 µM, including conditions in which three different FFAs were combined, each at a concentration of 10 µM. Notably, the magnitude of the effects was similar when FFAs were applied individually or in combination, suggesting possible saturation of the signaling pathway. Therefore, lower FFA concentrations, such as those found in the bone marrow microenvironment should be used to better reflect physiological conditions. Moreover, the FFAs were measured in adiposities condition media but only relative abundance was included in figure 1E, not actual levels. The authors used SU2C data sets to evaluate ANGPTL4 expression in bone metastases and associated increased expression worse better survival. However, in this data set, it is clear that ANGPTL4 exhibits significantly higher expression in liver metastases than in bone metastasis. Some discussion about this aspect would increase the potential translational indication.

    Minor comments:

    Preparation of the gel description stipulates "One hundred microliters of SCAds or rBMAds were added and homogenized quickly with 100µL of thrombin....." Number of cells added for normalization should eb added to the methods. In this sentence- is it correct to assume that oleate is also uM and not uL? "One day after seeding, cells were treated with 100μM palmitate (Cayman #29558), 100μL oleate (Sigma Aldrich #O3008), 100μM linoleate (Sigma Aldrich #L9530), or with the mix of 3 FFAs (100μM palmitate/100μM oleate/100μM linoleate) during 3 days.

    There is a description of an experiment: "PC3 and Du145 were transiently transfected with ANGPTL4 siRNA pool (final concentration 25nmol/L) using ON-TARGETplus SMARTpool (Thermo Scientific Dharmacon) or ON-TARGETplus nontargeting pool used as a control (Thermo Scientific Dharmacon). Transfection was done according to the manufacturer's instructions with Lipofectamine RNAiMAX (Invitrogen Life Technologies) followed by 24h of coculture with rBMAds. After coculture, a second transfection was done and gene extinction, migration and invasion were evaluated." The rationale for second transfection (assuming still with siRNA) has not been provided and it is not clear.

    Figure 6. Survival: HR and number of patients at each time point should be added to the graph. Some English editing is needed

    Significance

    There are multiple studies focusing on adipocytes and FFAs effects on cancer, in general, as well as on prostate cancer specifically. Effects on migration and invasiveness are well documented in literature. The novel aspect of the current manuscript is specifically looking at red bone marrow adipocytes, but it is not clear that the effects of FFAs would be different than when other adipocytes are used.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Summary:

    The study entitled "Human bone marrow adipocytes drive prostate cancer bone metastasis progression via lipid-mediated induction of Angiopoietin-like 4" by Hernandez et al. investigates the role of the bone marrow adipocytes (BMAds) from red hematopoietic rich-areas in the progression of prostate cancer (PCa) bone metastases. To do so, the authors used an elegant three-dimensional culture of primary human rBMAds isolated from the femoral bone marrow of patients undergoing hip surgery then cultured with PCa cell lines. The authors claim that PCa cells take up FFAs released by rBMAds which enhance the epithelial-mesenchymal transition and motility through upregulation of ANGPTL4. The article is well written, the figures are very clear and the methodology is perfectly described. To further improve the understanding of this mechanism, the authors should address 1) how PCa cells take up the FFAs, 2) whether PCa cells enhance non-canonical FFAs release and 3) the specificity of the rBMAds in their ability to upregulate ANGPTL4 compared to other adipocytes.

    Major comments:

    The paper's conclusions are overall convincing but the following supplementary experiments or discussion points could strengthen their claims. Is ANGPTL4 upregulation in PCa cells specific to rBMAds or could it also be seen with other adipocytes, such as periprostatic adipocytes or SCAds (used in Fig 2) ? Co-culture of PCa cells with SCAds should offer insights. More precisely, the article does not state if ANGPTL4 upregulation could be induced prior to bone invasion, by the periprostatic adipose tissue FFA release for example, explaining why ANGPTL4-high cells could be found in bone metastasis. In that case, increased motility and migration could be BMAds independent and trigger the first metastatic events originating from the primary prostatic tumor. The article hypothesizes an ANGPTL4-associated increase of motility and migration of PCa cells upon interaction with bone marrow adipocytes in the first metastatic bone, inducing enhanced invasion of said bone and secondary metastasis sites. However the article lacks evidence that ANGPTL4 is also upregulated in cells from secondary metastasis sites in patients. It would be interesting to explore in the existing data set the expression of ANGPTL4 in secondary metastasis of patients who had previous bone metastasis. These points could be addressed in the concluding remarks.

    Additional experiments or qualifying the claims:

    Fig 1 : It would be interesting to have an insight on the mechanism by which FFAs are uptaken by PCa. Fig 1D : It would be interesting to add an extra control of NC PCa treated with BODIPY to measure basal BODIPY uptake of PCa
    Fig 2 : The figure efficiently states that rBMAds undergo non-canonical lipolysis in comparison to SCAds, without PCa cells. Glycerol dosage and/or pan-lipase inhibitor treatment of rBMAds in co-culture with PCa would confirm rBMAds still show non-canonical lipolysis in bone metastasis environment. Fig2 : Is ANGPTL4 upregulation in PCa cells specific to rBMAds or could it also be seen with other adipocytes, such as periprostatic adipocytes or SCAds ? Co-culture of PCa cells with SCAds should offer insights. Fig 2 : Could the authors specify if there is a possibility the lipolysis kinetics is slower in rBMAds instead of an actual incomplete lipolysis ? Fig 3B/C : The author should provide tables to show : 1) all significantly up and down GO term 2) the top 20 genes DEG 3) any GO terms/genes related to EMT (table 1 should be associated with figure 3) Fig 5 : To complete experiences of KD of ANGPTL4 and PPARy inhibition, the author should also include the mRNA levels of EMT associated genes in these conditions. To better the clarity of the paper, the author should indicate in table S1 which patients are used in which figures. Also, what is the rationale for not using female patients in all the experiments ?

    the suggested additional experiment should be achievable in 6 months.

    Satistical analyses: The authors state in the Materiel and Methods that they used Student's t-test, one-way ANOVA or two-way ANOVA. These tests are applied to parametric data sets, validated for their normal distribution. Usually, n < 30 implies the use of non-parametric tests. The authors should better justify the rationale for their choices. Also, the details of tests for each figure is not indicated in the legends as stated in the Mat&Med.

    Minor Comments:

    In Fig 1F, the author should discuss the discrepancy between the significant increase of TG uptake in PCa shown in 1B and the non-significant effect seen in 1F, given that TG represents 85% of the total FFAs shown in 1E and 1F. Fig 3 : In Fig 3G and 4E, staining quantification would be necessary. In Fig3/Sup Fig3, why is the invasion assay not shown for LNCaP cell line ? 5F: Is there a reason why the sum of FFA has a stronger effect than the individual ones ? 5E/F : Rationale for changing FFA treatment duration between 5E and 5F (72h then 24h) ? For the discussion : the authors clearly describe the two different BMAds subtypes (cBMAds vs rBMAds), would it be interesting to question the role of cBMAds in PCa progression as well ?

    The studies are referenced appropriately.

    Please, could the authors indicate the co-culture duration with rBMAds in the legends of Figure 5. Excepted this point, the text and figures are clear and accurate.

    Significance

    General assessment:

    the authors used an elegant three-dimensional culture of primary human rBMAds isolated from the femoral bone marrow of patients undergoing hip surgery then co-cultured with PCa cell lines. The authors claim that PCa cells take up FFAs released by rBMAds which enhance the epithelial-mesenchymal transition and motility through upregulation of ANGPTL4. The most important aspects of this study are 1) the description of the non-canonical FFAs release from rBMAds, 2) the induction of EMT/motility of PCa cells induced by FFAs uptake. To further improve the understanding of this mechanism, the authors should address 1) how PCa cells take up the FFAs, 2) whether PCa cells enhance non-canonical FFAs release and 3) the specificity of the rBMAds in their ability to upregulate ANGPTL4 compared to other adipocytes.

    This study extends the knwoledge in the field of protaste cancer metastasis. The authors have developed an elegant 3D culture model of primary human rBMAds, which is highly relevant for investigating the role of adipocytes in cancer cell behavior. Indeed, most models used in the literature rely on adipocytes differentiated from bone marrow mesenchemal stem cells. Although adipocytes differentiation models have recently been improved by incorporating 3D structure, the model developed by the authors is most physiological, as it is based on the direct use of primary bone marrow adipocytes. In this way, this model represents a valuable tool to advance our understanding of mechanisms involved in bone metastasis in breast/prostate cancers.

    The type of audience that will be interested by this research include both basic and specialized researchers. Moreover, this study should improve 1) the understanding of the impact of rBMAds on prostate cancer behavior and 2) the robustness and physiological relevance of adipocyte-based studies.

    Our expertise lies in the interaction between normal or pathological hematopoiesis with the components of bone marrow microenvironment.

  5. Version published to 10.64898/2025.12.16.694397 on bioRxiv