Targeting AXL and RAGE to prevent geminin overexpression-induced triple-negative breast cancer metastasis

Dissemination of metastatic precursors from primaries is the primary reason for patient death. Dissemination encompasses tumor cells invasion of stroma, followed by intravasation through the endothelium barrier into the bloodstream. Here, we describe how geminin-overexpressing tumor cells acquire dissemination ability. Acetylated HMGB1 (Ac-HMGB1) secreted by geminin-overexpressing cells activates RAGE and CXCR4 expression on mesenchymal stem cells (MSCs) located in tumor stroma. Through secreting CXCL12, geminin-overexpressing cells recruit these CXCR4+-MSCs into the tumor. Within the tumor, MSCs differentiate into S100A4-secreting cancer-associated fibroblasts (CAFs). S100A4, in a reciprocal manner, activates geminin-overexpressing cells to secrete CCL2 that recruits M0-macrophages from the stroma into the tumor. Within the tumor, CCL2 polarizes M0-macrophages into Gas6-secreting M2-tumor-associated macrophages (M2-TAMs). In concert, geminin-overexpression, S100A4/RAGE and Gas6/AXL signaling promote the invasive and intravasation abilities in geminin-overexpressing cells through exacerbating their stemness and epithelial-to-mesenchymal phenotypes and enhancing expression and functional interaction of CD151 and α3β1-integrin in geminin-overexpressing cells. Tumors formed following injection of geminin-overexpressing cells admixed with MSCs/CAFs grew faster, metastasized earlier, especially to lungs, and were extremely sensitive to anti-c-Abl, anti-RAGE, and anti-AXL drugs. These data support an intrinsic ability in geminin-overexpressing tumor cells to promote their metastatic potential through recruitment and bi-directional interactions with MSCs/CAFs and M2-TAMs.

High-mobility group box 1 (HMGB1) is a ubiquitous DNA-binding protein with essential DNA metabolism functions 17 . HMGB1 can be released passively from necrotic cells or actively from activated immune cells, hypoxic, or inflamed cancer cells 18 . HMGB1 secretion requires hyper-acetylation on the chromatin 10,19 . Recently, we described how geminin helps acetylate chromatic HMGB1 and release it from GemOE tumor cells 10 , where it through binding to receptor for advanced glycation end products (RAGE) on GemOE tumor cells activates NF-κB-induced survival, especially those exposed to the harsh condition of hypoxia and inflammation 10 within the tumor core (aka "aggressiveness niche" 20 ). Binding of extracellular Ac-HMGB1 to RAGE on naïve mesenchymal stem cells (MSCs) activates NF-κB signaling-induced CXCR4 expression. CXCR4-expressing MSCs are then recruited to CXCL12/SDF1-secreting GemOE cells, in vitro, and into the aggressiveness niche, in vivo 10,20 .
Naïve MSCs layered in inserts of Boyden chambers efficiently migrated towards Dox-induced Gem197, Gem240, or Gem257 cells CM not MSCs CM or HME cells CM added in the lower chamber within 24 h (Fig. 1D). The recruitment was almost completely blocked in the presence of HMGB1 or CXCL12 NeuAbs (Fig. 1D). Together, reinforce that a primary function for RAGE/CXCR4 activation in naïve MSCs is to promote their migration towards GemOE tumor cells (step 1, Fig. 1G).
Moreover, ELISA showed that neither normal HME cells nor Dox-induced GemOE cells secrete S100A4 (not shown). In contrast, naïve MSCs exposed (24 h) to Dox-induced Gem240 or Gem257 cells CM only secrete high-levels of S100A4 (Fig. 1F, white bars). This secretion increased further when hypoxic Dox-induced Gem240 or Gem257 cells CM was used instead (Fig. 1F, red) and almost completely blocked by HMGB1 NeuAb (Fig. 1F, black). One-way ANOVA, followed by post hoc Bonferroni tests, confirmed these data (Suppl. Fig. 4). Together The levels of RAGE and TLR4 in the indicated cell lines exposed 24 h to normoxic (upper) or hypoxic (lower). The blot was repeated 3 separate times. (F) The level of S100A4 secreted from MSCs exposed 24 h to indicated cell lines CM under normoxic or hypoxic conditions in the absence or presence of HMGB1 NeuAb. Assay performed 3 separate times, each in triplicates. (G) Schematic representation showing the data discussed in the Figure. www.nature.com/scientificreports www.nature.com/scientificreports/ suggest that in the vicinity of GemOE cells, MSCs differentiate into S100A4-secreting cancer-associated fibroblasts 10,41 (i.e., CAFs, Fig. 1G, step 2).

S100A4-activated GemOE cells attract and polarize macrophages into Gas6-secreting M2-TAMs.
Recently, we showed that IRISOE cells secrete high-levels of CCL2, in vitro, and in vivo to recruit macrophages 42 . To define whether bidirectional interaction between GemOE cells and MSCs through S100A4 exists was studied next.
Many THP1s layered in inserts of Boyden chambers migrated 24 h later towards Dox-induced Gem197, Gem240, or Gem257 cells/MSCs co-cultures CM only in the lower chambers. Importantly, in the presence of CCL2 NeuAb, this migration was significantly blocked (Fig. 2E), supporting the CCL2/CCR2 role in macrophages recruitment into GemOE tumors.
Our analysis of receptors upregulated on the surface of the 1° orthotopic GemOE mammary tumors revealed many important metastasis inducing receptors. One such receptor, AXL, was chosen for further analysis in this study due to its pronounced role in intravasation 46 and breast cancer metastasis 30 . Compared to normal HME cells, Dox-induced Gem197, Gem240, and Gem257 cells express significantly higher levels of AXL on their surfaces (western done on cell membrane extracts, Fig. 2F).

imatinib-sensitive MScs and tAMs recruitment in vivo.
Previously, we injected shCtrl-, sh-gemininor shc-Abl-expressing GemOE cells into athymic female mice mammary fat pads. Geminin-or c-Abl-depleted cells developed <20% size tumors, compared to control cells 16 , supporting the intimate relationship between c-Abl and geminin and giving credence to using imatinib to treat GemOE mammary tumors (especially TNBC) see introduction and 16 .
Following the 24 h treatments, cells were sonicated, and whole-cell extracts prepared. HME cells grown in SF-medium expressed low levels of geminin, CDH2, and survivin, while detectable levels of Twist1, Slug, Oct4, and Sox2 (Fig. 6B). In contrast, in SF-media, all Dox-induced GemOE cells expressed high levels of geminin,  www.nature.com/scientificreports www.nature.com/scientificreports/ CDH2, and survivin, and even higher levels of Twist1, Slug, Oct4, and Sox2 compared to HME cells (Fig. 6B). None of the treatments significantly affected the expression of these proteins in HME cells (Fig. 6B), while in SF-media, S100A4 + Gas6 plus vehicle elevated the expression of all proteins to different degrees in the three Dox-induced GemOE cell lines (Fig. 6B). In SF-media, S100A4 + Gas6 plus FPS-ZM1 significantly decreased the expression of the majority of the proteins except for geminin to different degrees in the three Dox-induced GemOE cell lines (Fig. 6B), plus R428 treatment decreased expression of all proteins in some instances marginally, e.g., Twist and Sox2, yet in others significantly, e.g., Slug and Oct4 (Fig. 6B). In SF-media, S100A4 + Gas6 plus both drugs treatment, decreased the expression of all proteins even geminin to a level similar to that observed with R428 alone (Fig. 6B), suggesting that RAGE perhaps primes AXL effects.
Next, we layered Gem240 or Gem257 cells in Matrigel-coated inserts of Boyden chambers. In the lower well, we added SF-media containing Dox and none or S100A4 + Gas6 plus vehicle, FPS-ZM1, R428, or both. After 24 h, we counted cells invaded the Matrigel and migrated to the other side of the inserts as an in vitro confirmation of their stemness and EMT phenotype described above.
Although, many Dox-induced Gem240 and Gem257 cells invaded in SF-condition (Fig. 6C,D), a significant increase in the number of invaded cells was detected in the presence of S100A4 + Gas6 plus vehicle (Fig. 6C,D). On the other hand, in the presence of FPS-ZM1 or R428, S100A + Gas6 effect on invasion was significantly blocked (Fig. 6C,D). Again, the reduction in invasion was the same in the presence of both drugs or R428 alone (Fig. 6C,D). One-way ANOVA test followed by post hoc Bonferroni tests confirmed these data (Suppl. Figs. 13  and 14). Together, reinforces that RAGE activation perhaps primes GemOE/TNBC cells to Gas6 effects by affecting geminin and/or AXL expression and/or activity.
To test this hypothesis, we grow HME, Gem197, Gem240, or Gem257 in Dox-containing SF-media for 24 h then switched them to Dox-containing SF-media supplemented with S100A4 plus vehicle or FPS-ZM1 (Fig. 6E) or supplemented with Gas6 plus vehicle or R428 (Fig. 6F). First, in SF-media, S100A4 elevated geminin and AXL levels in all Dox-induced GemOE (not HME) cells (compare green to red in Fig. 6E). Importantly, the expression of geminin and AXL proteins were significantly decreased in the presence of S100A4 plus FPS-ZM1 (compare blue to green in Fig. 6E). It is possible that S100A4/RAGE signaling stabilizes geminin protein (exogenous geminin is expressed from a heterologous promoter) and/or induces its transcription. Geminin, then could affect AXL gene transcription and/or protein stabilization.
Second, in SF-media, Gas6 did not affect geminin expression, while increased AXL expression in all Dox-induced GemOE cells (compare green to red in Fig. 6F). R428 did not affect geminin expression in Dox-induced GemOE cells but inhibited AXL expression (compare blue to green in Fig. 6F). It is possible that S100A4/RAGE signaling separately affects geminin or AXL. Alternatively, positive feedback between geminin and AXL is also possible. We favor the latter since only in Dox-induced GemOE cells, AXL was expressed (Fig. 6E,F).
Finally, in SF-media, Dox-induced Gem240 and Gem257 cells express relatively high levels of pro-/ active-MMP2 and -MMP9. Expression of both significantly increased in the presence of S100A4 + Gas6 plus vehicle (Fig. 6G) and was blocked in the presence of FPS-ZM1, R428, or both drugs (Fig. 6G). The bidirectional interactions with MSCs/CAFs and M2-TAMs through S100A4/RAGE and Gas6/AXL signaling that activated AKT, ERK, and NF-κB in GemOE cells elevate geminin and/or AXL expression triggering TNBC cells stemness, EMT, and their invasion ability (cf. Fig. 6H).

AXL and RAGe enhance Gemoe cells' intravasation ability.
The integrin α3 and β1 (hereafter α3β1-integrin) complex with CD151 (expressed specifically on intravasating tumor cells) is essential to initiate and maintains a tight interaction with the extracellular matrices (ECMs); e.g., laminin I (LMN I), collagen IV (COL IV) and fibronectin (FN) on the basal side of endothelial cells of the vessel [61][62][63] . To study this in our system, we grew Gem240 or Gem257 for 24 h on uncoated wells or wells that were coated with LMN I, COL IV, or FN in SF-media containing Dox and supplemented with none or S100A4 + Gas6 plus vehicle, FPS-ZM1, R428 or both. The next day, wells were washed 3 times, and the cells remaining attached to each well were counted in HF and blotted.
The Gem240 cells (identical results were obtained with Gem257 cells, not shown) showed a high binding ability to uncoated wells under all conditions (white, Fig. 7A). In comparison, binding to LMN I, COL IV, and FN was significantly decreased under SF conditions (none, Fig. 7A). The binding to all matrices was restored in SF containing S100A4 + Gas6 plus vehicle (Fig. 7A), which was blocked by FPS-ZM1, R428, or both treatments (Fig. 7A). One-way ANOVA, followed by post hoc Bonferroni tests, confirmed these data (Suppl. Fig. 15). Together, suggest that AXL and/or RAGE positively affect the binding of intravasating tumor cells with the ECM on the vessels within GemOE tumors.
To study the contribution of CD151 and α3β1 in this binding, we again in SF-media grow normal HME cells, or Dox-induced Gem197, Gem240, or Gem257 cells in the presence of none, or Gas6 plus vehicle or plus R428. HME expressed very low levels of α3-integrin, β1-integrin, while a high level of β4-integrin (β4) and no CD151 (Fig. 7B). In contrast, in SF-media, all Dox-induced GemOE cell lines expressed high levels of α3-integrin and β1-integrin, low levels of β4-integrin, and very high levels of CD151 (red in Fig. 7B). Gas6 treatment did not affect the expression of these proteins in normal HME cells, but significantly increased expression of α3-integrin, β1-integrin and CD151 in all Dox-induced GemOE cells by >2fold (compare green to red in Fig. 7B). Importantly, inactivating AXL significantly blocked Gas6-induced expression of all four proteins in all Dox-induced GemOE cell lines (compare blue to green in Fig. 7B), suggesting a direct role for AXL in GemOE intravasation ability.
AXL + /β1 + cell populations. All cell lines contained ~40% of AXL + /β1 + cell populations in SF-media (red, Fig. 7C-E), that didn't change in the presence of S100A4 + Gas6 plus vehicle in any of the cell lines (red, Fig. 7C-E). Although, S100A4 + Gas6 plus FPS-ZM1 or plus R428 didn't change this population in any of the cell lines, in S100A4 + Gas6 plus both a slightly yet significantly decreased in the AXL + /β1 + -populations in all cell lines was detected (red, Fig. 7C-E).
AXL + /CD151 + cell populations. All cell lines contained >60% of AXL + /CD151 + cell populations in SF-media (white, Fig. 7C-E), that didn't change in S100A4 + Gas6 plus vehicle, plus FPS-ZM1 or plus R428 treatments The expression of the indicated proteins in the indicated cell lines following 24 h growth in Dox-containing SF-media followed by a switch to Dox-containing SF-media supplemented with none, or Gas6 plus vehicle, or plus R428 for 24 h. Note, the actin blot is the same as in Fig. 6F because it is from the same experiment. Percentage of AXL + /β1-integrin + populations (red), the AXL + / CD151 + populations (white), or β1-integrin + /CD151 + populations (black) derived from FACS analysis of nonpermeabilized Dox-induced Gem197 (C), Gem240 (D), and Gem257 (E) treated as indicated. Each assay was performed 3 separate times, each in triplicates. (F) Schematic representation of the data in the entire study as well as future directions. (2019) 9:19150 | https://doi.org/10.1038/s41598-019-55702-w www.nature.com/scientificreports www.nature.com/scientificreports/ (white, Fig. 7C-E). Importantly, a significant decrease in the percentage of the AXL + /CD151 + population was observed in all cell lines following S100A4 + Gas6 plus both drug treatment (Fig. 7C-E).
β1-integrin + /CD151 + cell populations. All cell lines contained 25-40% of β1 + /CD151 + cell populations in SF-media (black, Fig. 7C-E). Again, this percentage was not significantly changed in S100A4 + Gas6 plus vehicle, plus FPS-ZM1 or plus R428 treatments (black, Fig. 7C-E). Importantly, a significant decrease in the percentage of this β1 + /CD151 + cell populations was observed in all cell lines following treatment with S100A4 + Gas6 plus both drugs (black, Fig. 7C-E). Taken together, we propose that GemOE triggers expression and/or activation of AXL and RAGE through interactions with the microenvironment that, in turn, enhances the expression and the functional interaction between α3β1-integrin and CD151. This complex by tightly binding to the ECM on the tumor side of endothelial cells enhances the intravasation ability of GemOE/TNBC metastatic precursors through the endothelium barrier upon dissemination (cf. Fig. 7F).

Discussion
The vast majority of breast cancer deaths are due to metastatic diseases. Inflammation and hypoxia within the tumor microenvironment, especially the core (i.e., the aggressiveness niche 20 ) exacerbate metastasis. Here, we elucidated the mechanistic role of the bi-directional interactions between GemOE tumor cells and the microenvironment in promoting GemOE metastatic precursors invasion and intravasation abilities.
MSCs and TAMs are recruited into tumors through tumor secreted factors. MSCs are recruited by Ac-HMGB1/CXCL12 27,64 , while TAMs by CCL2 31,65 . We resolved an issue that was unclear in our previous study 10 , how Ac-HMGB1 induced CXCR4 expression through RAGE in the RAGE-negative naïve MSCs? We found that Ac-HMGB1 activates TLR4 first, which then upregulates RAGE expression in naïve MSCs 66-68 (cf. Fig. 1G). Earlier reports also showed that HMGB1 promotes CXCR4 expression through RAGE and TLR4 10,69 .
Interestingly, in our experiments, Ac-HMGB1 activated NF-κB rather than AKT signaling via TLR4. A recent study showed that lipopolysaccharide (LPS) signaling activates AKT via TLR4 to promote breast cancer metastasis 66 . It is possible that during breast cancer metastasis, Ac-HMGB1 and LPS each initiate a specific signaling pathway downstream of TLR4 70 .
Another important aspect of our studies is that extracellular S100A4 instigates a tumor-supportive microenvironment in GemOE/TNBC tumors, as was recently shown 71 . High S100A4 expression associates with poor outcome in early-stage cancers 24 and direct interaction between S100A4 and RAGE was recently shown to promote motility in colorectal cancer cells via ERK-dependent mechanism 72,73 , and prostate cancer cells via NF-κB-activated pathway 40 . Chronic inflammation enhances RAGE expression in many cancers 25,72,74,75 , which was directly linked to low-rate patient survival. In our current analysis of a public dataset of >1700 breast cancer samples, low RFS in patients with high geminin + HMGB1 + S100A4 (two RAGE ligands)-expressing patients was observed (Fig. 3H). This observation supports the conclusion that RAGE signaling is involved in enhancing aggressiveness in GemOE TNBC tumors (cf. Fig. 3G). We propose that Ac-HMGB1 plays a previously unappreciated role in early-stage TNBCs.
TAMs presence within TNBC tumors is recognized as a critical factor in tumor progression and could be a prognostic factor of a worse outcome 49,50,56,57 . Most human tumors exhibit TAMs with an M2-like phenotype involved in promoting EMT and contributing to tumor progression and drug resistance [49][50][51] . In keeping, the presence of a high number of CD163 + -M2 macrophages 76 in GemOE orthotopic TNBC tumors was correlated with aggressive behavior 50 . We propose that TAM secretome, including Gas6, contributes to this aggressiveness through activation of the AXL receptor, which is overexpressed on GemOE/TNBC tumor cells. Indeed, recently published reports 49 demonstrate that mesenchymal-like breast cancer cells induce the release of Gas6 selectively from M2-type macrophages 49 .
It is possible that along with c-Abl, AXL is an additional therapeutic target to prevent GemOE metastatic precursors dissemination from TNBC primaries. It is possible that in vivo selective inhibition of AXL using R428 could cooperate with imatinib to impair GemOE/TNBC cell invasion and intravasation, entrapping these GemOE/TNBC metastatic precursors within primaries to be resected during surgery reducing the metastatic burden. In support of this conclusion, we found that high geminin + AXL expression was associated with reduced RFS, DMFS, OS and LMFS in TNBC patients (cf. Fig. 5F-H, and Suppl. Fig. 12) Our study has some limitations. Although, we provide evidence for intrinsic ability in GemOE/TNBC tumor cells to recruit and activate stromal elements, such as MSCs and TAMs, the specific requirement of RAGE and AXL for the interaction with MSCs and TAMs, and the distinct biological role in different subtypes of the TNBC disease warrant further studies. Despite these limitations, our results suggest that RAGE and AXL are prognostic indicators of outcome in GemOE/TNBCs patients.
Another limitation is the fact that direct testing for S100A4 in TAMs recruitment into GemOE/TNBC tumors should be done, in vivo. Although all the drugs used in this study are specific for their cognate target, their effect in our system should be tested in future studies to provide mechanistic insight, for instance, using the same assays except in cell lines harboring knockdown/knockout of these targets. Also, in future assays, we will assess whether the recruitment of M2-macrophages only or the macrophages, in general, is affected by imatinib, for example, using FACS analysis for the CD11b/F4/80 population within GemOE/TNBC tumors.
Moreover, we demonstrated AXL-mediated elevation in CD151, α3, and β1-integrin expression and functional interaction enhance binding to ECM in vitro, which could also be the case on vessels within tumors. This observation would be consistent with previous reports suggesting AXL's role in breast cancer cells intravasation of the endothelium barrier into the bloodstream 46 . Noteworthy here, in preliminary data (not shown), we found that β1-integrin is phosphorylated in GemOE. An exciting possibility is that this phosphorylation is AXL-or AXL/ CD151-driven. In support of the former, R428 abolished this phosphorylation event in GemOE cells (not shown). CD151 may also be a target along with c-Abl and AXL for therapy of GemOE/TNBC tumors. (2019) 9:19150 | https://doi.org/10.1038/s41598-019-55702-w www.nature.com/scientificreports www.nature.com/scientificreports/ Our analysis showing the opposite effect for GemOE on β4-integrin expression is intriguing. At present, the functional significance of this observation remains unknown. It is possible since α3β4-integrin is also a receptor for ECM binding in breast cancers 77 , that β4 is involved in the extravasation rather than the intravasation step of GemOE/TNBC metastatic spread, as described recently 48 . This will be investigated further in a follow-up study. Another critical experiment that will also be reported in a soon to be published article is the importance of MSCs/ CAFs in the proposed intravasation of GemOE/TNBC cells.
These findings implicate synergistic signals simultaneously mediate multiple mechanisms in GemOE/TNBC intravasation. This study provides new understanding of the signaling pathways activated in GemOE/TNBC cells through interactions with stromal cells (MSCs/CAFs and M2-TAMs) that modulate the intravasation ability of tumor cells and underlines the importance of probing the process of tumor dissemination from an essential yet hitherto under-explored bi-directional interactions between GemOE/TNBC tumor cells and their microenvironmental cells. The dependence of tumor intravasation on the microenvironment provides yet another point of intervention to prevent tumor cell dissemination and patients' demise 78 .
Methods cell culture. Human MSCs were purchased from Texas A&M HSC COM Institute for Regenerative Medicine, expanded, and frozen down. THP-1 cells were purchased from ATCC (catalog number: TIB-202 TM ), expanded, and frozen down. For both cell types, one vial is usually propagated for ≤5 generations to perform experiments before a new vail is used. Parental HME cell line (i.e., normal HME, used as control through the studies performed here) was transfected with a retrovirus expressing doxycycline (Dox)-inducible geminin allele 14 . Several clones from these cells were generated by antibiotic-selection, expanded, and tested for geminin expression 14 . A clone named Gem9 overexpressing a level of geminin similar to that observed in breast cancer cell lines was selected to use further. Gem9 cells were orthotopically injected in Dox-supplemented female SCID mice mammary fat pads (n = 10) and the primary (1°)-geminin overexpressing tumors developed were collected 13 and used to generate the 1° geminin-overexpressing tumor cell lines we named; Gem197, Gem240, Gem256, Gem257, and Gem270. These cell lines were described earlier 10 , and are usually maintained in RPMI medium supplemented with 10% FBS, 1% antibiotics, and 2 µg/ml doxycycline to induce geminin expression will be used to perform the assays in this study. All commercial and in-house cell lines were authenticated by STR profiling and tested for mycoplasma contamination.
Growth factors, cytokines, and drugs. Recombinant (r)S100A4, rGas6, and rHMGB1, were from Sino Biological. Acetylation of rHMGB1 was described earlier 10  Serum isolation from mouse blood. Serum over plasma was chosen to study because it is more representative of inflammation. Samples were collected in a covered test tube and allowed to clot undisturbed at RT for ~30 min. The clot was then removed by centrifuging at 1,000-2,000 g for 10 min at 4 °C. The isolated serum is immediately transferred into a clean polypropylene tube aliquoted into 0.5 µl aliquots and stored in −80 °C. eLiSA analysis. Wells of a PVC microtiter plate were coat with the antigen by pipetting 50 μl of the conditioned medium (CM) or Serum (or dilution thereof in PBS) in triplicates and plate covered and incubated for overnight at 4 °C. After washing 3x with PBS, blocking of none-specific sites was done using 5% BSA in PBS incubated for ~2 h at room temperature (RT) followed by 3x washing in PBS. Diluted 1° antibody was added for 2 hours at RT, followed by 3x washing in PBS. HRP-conjugated 2° antibody diluted in blocking buffer was incubated 1 hour at RT then 3x washed with PBS. Detection using OPD (o-phenylenediamine dihydrochloride) tables and detection at 492 nm was done. cytokine array. HME and HME/geminin cells were assessed for differential cytokine secretion using human cytokine antibody array (RayBio). Conditioned media from an equal number of HME and HME/geminin cells plated in a serum-free medium for 20 h under standard conditions were used to perform according to the manufacturer's instructions and previously described 10 . co-culture experiment. Boyden chambers (BD biosciences) of 8 µm (for migration) or 0.4 µm (for secretome) pore size were used. Certain cells (or their CM) were layered in the lower chamber with or without (2019) 9:19150 | https://doi.org/10.1038/s41598-019-55702-w www.nature.com/scientificreports www.nature.com/scientificreports/ neutralizing antibodies, and test cells were layered in the transwell inserts. Cells migrated to the lower compartment of the Boyden chamber were counted and plotted 24 h later. Occasionally, hypoxia was introduced.
conditioned media transfer experiment. Conditioned medium (CM) generated for analysis of secreted cytokines and surface receptor expression under normoxia (20% O 2 for 24 h), or hypoxia (1% O 2 for 24 h) conditions from an equal number of cells were transferred onto MSCs in the presence or absence of drugs or NeuAb. MSCs re-CM medium was transferred onto HME or GemOE cells with or without drugs or NeuAb for 24 h to re-reconditioned. The resultant CM was then transferred to an equal number of THP1-macrophages for 24 h to be re-re-reconditioned. Finally, this CM was re-added onto HME or GemOE cells for 24 h. At every step, the receiving cells were seeded at equal numbers to avoid number variations discrepancies. At various points, ELISA on CM and western blot on membrane fractions or whole-cell extracts generated by sonication was done.
In vivo drug treatments. Tumor-bearing mice were treated with the drugs, concentration, and routes indicated in Results.
Preparation and injection of GFP-MSCs or -THP1s. GFP-expressing MSCs or THP1 were generated using lentivirus GFP-expressing plasmid. Antibiotics selected clones were propagated and stored frozen. Tumor-bearing mice and treated as described in Results were intracardiac injected through the left ventricle. Following tumor resection, part of the tumor was dissociated using Collagenase I and the protocol used by 94 , grown in culture for 24 h to adhere, and GFP-cells were photographed, counted, and plotted using Prism 7 ® . fluorescence iHc. IHC was performed as previously described 95 . Briefly, 5μm thick paraffin-embedded sections of tumor tissue excised from GemOE orthotopic mammary tumors were deparaffinized, rehydrated, and washed in PBS. Antigen retrieval for all antibodies was done using Sodium citrate buffer (10 mM Sodium citrate, 0.05% Tween 20, pH 6.0), 10 min at 95 °C. Slides were blocked with 10% normal serum (2° antibody species) for 1 h at RT, washed, and probed with 1° antibodies overnight at 4 °C. Slides were exposed to Alexa Fluor 568 (red), and 488 (green) conjugated secondary antibody for 1 h at RT and counterstained and mounted with VECTASHIELD mounting medium with DAPI (Vector) and imaged under the fluorescence microscope.