Tumor Cell Associated Hyaluronan-CD44 Signaling Promotes Pro-Tumor Inflammation in Breast Cancer

Cancer has been conceptualized as a chronic wound with a predominance of tumor promoting inflammation. Given the accumulating evidence that the microenvironment supports tumor growth, we investigated hyaluronan (HA)-CD44 interactions within breast cancer cells, to determine whether this axis directly impacts the formation of an inflammatory microenvironment. Our results demonstrate that breast cancer cells synthesize and fragment HA and express CD44 on the cell surface. Using RNA sequencing approaches, we found that loss of CD44 in breast cancer cells altered the expression of cytokine-related genes. Specifically, we found that production of the chemokine CCL2 by breast cancer cells was significantly decreased after depletion of either CD44 or HA. In vivo, we found that CD44 deletion in breast cancer cells resulted in a delay in tumor formation and localized progression. This finding was accompanied by a decrease in infiltrating CD206+ macrophages, which are typically associated with tumor promoting functions. Importantly, our laboratory results were supported by human breast cancer patient data, where increased HAS2 expression was significantly associated with a tumor promoting inflammatory gene signature. Because high levels of HA deposition within many tumor types yields a poorer prognosis, our results emphasize that HA-CD44 interactions potentially have broad implications across multiple cancers.


Introduction
Breast cancer is the second leading cause of cancer-related death among women in the United States [1]. The triple negative subtype, lacking estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) expression, represents approximately 12% of breast cancer cases [2]. Notably, this subtype has a poor prognosis, and because standard targeted therapies against ER and HER2 are ineffective in triple negative cases, chemotherapy is the main Figure 1. Hyaluronan synthase 2 expression (HAS2) in human breast cancer cell lines is associated with the triple negative breast cancer (TNBC) subtype. Analysis of HAS2 transcript expression using the NanoString nCounter platform to assess gene expression levels within a panel of breast cancer cell lines that include estrogen receptor ER+, progesterone receptor PR+, human epidermal growth factor receptor 2 HER2+ and triple negative (TNBC) subtypes. Gene expression levels were compared between cell line subtypes using an analysis of variance (ANOVA) test using R software. The ANOVA indicated significant differences between groups (p-value = 0.0395), prompting pairwise significance testing using the Tukey HSD post-test. A significant difference in HAS2 expression was found between TNBC vs. HER2+ subtypes (p-value = 0.0307702). Note, HAS2 expression was elevated in 11/17 TNBC cell lines. Data are summarized in the horizontal box plots (median, first and third quartiles, and 1.5 * interquartile range values are displayed).  UACC893  BT20  HCC70  HCC1954  HCC2218  MDA468  MDA453  MCF10A  CAMA1  T47D  AU565  SUM225  HCC1569  HCC1937  MDA415  MDA361  HCC202  BT474  SKBR3  MCF12A  MCF7ATCC  HCC1428  SUM149  ZR75−1  MDA436  HCC38  DU4475  MDA231  HCC1500  HCC1187  BT549  UACC812  Hs578T  HCC1395  HCC1806   − 1  0  1  2 Gene Expression (normalized, median−centered, scaled) Cell Line Subtype TNBC ER+ HER2+ Figure 1. Hyaluronan synthase 2 expression (HAS2) in human breast cancer cell lines is associated with the triple negative breast cancer (TNBC) subtype. Analysis of HAS2 transcript expression using the NanoString nCounter platform to assess gene expression levels within a panel of breast cancer cell lines that include estrogen receptor ER+, progesterone receptor PR+, human epidermal growth factor receptor 2 HER2+ and triple negative (TNBC) subtypes. Gene expression levels were compared between cell line subtypes using an analysis of variance (ANOVA) test using R software. The ANOVA indicated significant differences between groups (p-value = 0.0395), prompting pairwise significance testing using the Tukey HSD post-test. A significant difference in HAS2 expression was found between TNBC vs. HER2+ subtypes (p-value = 0.0307702). Note, HAS2 expression was elevated in 11/17 TNBC cell lines. Data are summarized in the horizontal box plots (median, first and third quartiles, and 1.5 * interquartile range values are displayed). HA was isolated from cell supernatants, protein was removed via proteinase K, and samples were precipitated using 100% ethanol. A portion of each sample was treated with hyaluronidase as a control to ensure degradation of HA fragments (+HAase). (C) Morphology (hematoxylin and eosin stain) of triple negative breast cancer xenografts in vivo. Representative 50× and 100× magnification images are shown. (D) Immunofluorescence microscopy for hyaluronic acid binding protein (HABP; green) and DAPI nuclear stain in the triple negative xenograft models. Inserts identify regions of heterogeneous HA staining, with both HA-high and HA-low/absent regions present within animal models of disease. Tumor nests surrounded by hyaluronan are outlined in white. White arrows call out interspersed stromal cells embedded in the HA-rich stroma surrounding the tumor nests, which are likely fibroblasts or monocyte/macrophages, based on the small, slightly elongated, and smoothly contoured nuclear morphology (specific stains to further elucidate were not performed). As a control, each section was treated with hyaluronidase prior to staining (+HAase). Each image was taken at 200× and 400× magnification. For large format 400× images depicting tumor nests and stromal cells, refer to Figure S1. HA was isolated from cell supernatants, protein was removed via proteinase K, and samples were precipitated using 100% ethanol. A portion of each sample was treated with hyaluronidase as a control to ensure degradation of HA fragments (+HAase). (C) Morphology (hematoxylin and eosin stain) of triple negative breast cancer xenografts in vivo. Representative 50× and 100× magnification images are shown. (D) Immunofluorescence microscopy for hyaluronic acid binding protein (HABP; green) and DAPI nuclear stain in the triple negative xenograft models. Inserts identify regions of heterogeneous HA staining, with both HA-high and HA-low/absent regions present within animal models of disease. Tumor nests surrounded by hyaluronan are outlined in white. White arrows call out interspersed stromal cells embedded in the HA-rich stroma surrounding the tumor nests, which are likely fibroblasts or monocyte/macrophages, based on the small, slightly elongated, and smoothly contoured nuclear morphology (specific stains to further elucidate were not performed). As a control, each section was treated with hyaluronidase prior to staining (+HAase). Each image was taken at 200× and 400× magnification. For large format 400× images depicting tumor nests and stromal cells, refer to Figure S1.
investigated further within this cell line. Interestingly, 4MU-mediated decreases in HA synthesis also caused a significant decrease in IL-8 production in Hs578T cells ( Figure S3A,B).

CD44 Knockout or Hyaluronan Depletion from Breast Cancer Cells Decreases Tumor Cell Production of Inflammatory Cytokines
Due to the association between HA-CD44 interactions, we investigated CD44 as a novel regulator of inflammatory cytokines within breast cancer cells. Studies were initially performed to validate RNA sequencing results using an inflammatory cytokine array. Inflammatory cytokine arrays were used as a screening tool to detect changes within tumor cell conditioned media collected from CD44 WT and KO Hs578T and MDA-MB-231 cells. As shown in Figure 4A ( Figure S14), CD44 KO globally affected the expression of key pro-inflammatory cytokines produced by breast cancer cells. Specifically, CCL2 production was impacted by the loss of CD44 in both the Hs578T and MDA-MB-231 cells, which is relevant to breast cancer, as CCL2 has been shown to recruit monocytes into the primary tumor and promote metastasis [35][36][37][38]. Therefore, a decrease in CCL2 production upon CD44 deletion was validated via an ELISA ( Figure 4B). Further studies were performed to investigate whether CCL2 was subsequently affected when HA synthesis was inhibited. Tumor cell HA synthesis was inhibited in WT Hs578T and MDA-MB-231 cells in vitro, via treatment with 600 µM [39,40] 4-methylumbelliferone (4MU) for 24 h. Conditioned medium was collected from cells and a decrease in HA production was confirmed using an HA ELISA. As was observed by limiting CD44 expression, 4MU-mediated decreases in HA synthesis caused a significant decrease in CCL2 production in both TNBC cell lines ( Figure 4C,D). While inconsistent between cell lines via cytokine array analysis, IL-8 production decreased within Hs578T CD44 KO cells. Because IL-8 has a known role in angiogenesis and cancer-associated inflammation [41][42][43], the importance of IL-8 in HA-CD44 signaling was investigated further within this cell line. Interestingly, 4MU-mediated decreases in HA synthesis also caused a significant decrease in IL-8 production in Hs578T cells ( Figure S3A,B).

CD44 Deletion in Breast Cancer Cells Delays Early Tumor Formation
Further studies were performed to determine the effects of CD44 deletion on tumor cell survival in vitro and tumor growth in vivo. To test cell survival in vitro, CD44 WT and KO cells were plated in a 96-well plate and an MTT assay was performed at 24, 48, and 72 h. Absorbances for CD44 KO cells were compared to their parent cell lines. CD44 KO had little effect on cell survival in both breast cancer cell lines ( Figure S4A). Next, studies were performed to determine whether CD44 had an impact on tumor cell growth rates in vivo. Therefore, Hs578T and MDA-MB-231 cells with and without CD44 were injected into the mammary fat pads of athymic nude mice. As shown in Figure S4B, CD44 did not affect survival based on tumor burden in Hs578T or MDA-MB-231 models, although onset was delayed within the Hs578T model. These findings demonstrate that there were not significant impacts of CD44 deletion on tumor cell growth and survival, in vitro or in vivo.
While time to tumor endpoint (2 cm 3 ) was unaffected by CD44 KO in vivo, we found that Hs578T cells took a significantly longer amount of time to form palpable tumors ( Figure 5A), consistent with delayed onset noted in the overall survival based on tumor burden ( Figure S4B). We investigated this phenomenon further, since early tumor growth following injection was not impacted by CD44 expression in the MDA-MB-231 mouse model ( Figure S4C). To accomplish this, Hs578T cells with and without CD44 were injected into the mammary fat pad of athymic nude mice for analysis at an earlier timepoint. After four weeks (when CD44 WT cells formed palpable tumors), mice were sacrificed, and mammary glands were collected for histology. Glands were embedded in paraffin, sectioned, and stained using H&E and immunofluorescence to detect HA (HABP), nuclei (DAPI), and macrophages (F4/80). As shown in Figure 5B, Hs578T CD44 WT tumors were associated with an invasive front that was found to infiltrate the adjacent mammary fat pad. In contrast, the Hs578T CD44 KO tumors were smaller and encapsulated within a dense HA matrix (which was not present in the CD44 WT tumors). Furthermore, in vivo findings were supported by an in vitro model of migration, where a scratch was created in tumor cell monolayers and wound closure was measured at 6, 12, and 24 h. Hs578T CD44 KO cells showed a significant decrease in cell migration, as shown in Figure 5C. Additionally, CD44 KO led to an increase in HA production within the Hs578T cells, consistent with the increased HA matrix observed in vivo, suggesting a feedback mechanism between CD44 and its ligand ( Figure 5D). Therefore, in vitro findings were consistent with the fact that Hs578T CD44 KO cells generated smaller tumors in vivo, lacking an invasive front and encapsulated by a dense HA matrix.

CD44 Deletion in Breast Cancer Cells Delays Early Tumor Formation
Further studies were performed to determine the effects of CD44 deletion on tumor cell survival in vitro and tumor growth in vivo. To test cell survival in vitro, CD44 WT and KO cells were plated in a 96-well plate and an MTT assay was performed at 24, 48, and 72 h. Absorbances for CD44 KO cells were compared to their parent cell lines. CD44 KO had little effect on cell survival in both breast cancer cell lines ( Figure S4A). Next, studies were performed to determine whether CD44 had an impact on tumor cell growth rates in vivo. Therefore, Hs578T and MDA-MB-231 cells with and without CD44 were injected into the mammary fat pads of athymic nude mice. As shown in Figure  S4B, CD44 did not affect survival based on tumor burden in Hs578T or MDA-MB-231 models, although onset was delayed within the Hs578T model. These findings demonstrate that there were not significant impacts of CD44 deletion on tumor cell growth and survival, in vitro or in vivo.
While time to tumor endpoint (2 cm 3 ) was unaffected by CD44 KO in vivo, we found that Hs578T cells took a significantly longer amount of time to form palpable tumors ( Figure 5A), consistent with delayed onset noted in the overall survival based on tumor burden ( Figure S4B). We investigated this phenomenon further, since early tumor growth following injection was not impacted by CD44 expression in the MDA-MB-231 mouse model ( Figure S4C). To accomplish this, Hs578T cells with Statistical analysis was performed using Student's unpaired, two-tailed t-test. Error bars represent standard error of the mean. P values * p < 0.05; ** p < 0.01; **** p < 0.0001. While Hs578T cells were derived from a primary tumor, the MDA-MB-231 cell line was derived from a distant, metastatic site [44]. Therefore, we hypothesized that loss of CD44 expression may impact metastatic lesion formation in vivo. To test this, CD44 WT and KO cells were injected into the tail vein of athymic nude mice. Seven weeks post injection, mice were sacrificed and lungs were collected for histology. As shown in Figure S5, loss of CD44 in MDA-MB-231 cells led to a decrease in the total number of lung lesions ( Figure S4A,C), as well as the percent area colonized by tumor cells ( Figure S4B). These findings are consistent with a previously published study [34] and confirm the importance for CD44 in promoting tumor progression in vivo. Thus, CD44 KO within MDA-MB-231 cells decreased tumor cell colonization within the lungs.

Figure 5. CD44 deletion in Hs578T cells delays early tumor formation. (A) Number of days that
Hs578T CD44 WT tumors (n = 10) and CD44 KO tumors (n = 14) took to form palpable lesions within the mammary fat pad of athymic nude mice. (B) Histological analysis of early tumors using H&E and immunofluorescence to identify HA in green (using hyaluronic acid binding protein), and nuclei in blue (using DAPI). Images were acquired on a Leica DM400B microscope, at either 100× or 400× magnification. (C) Scratch wound assay analyzing wound closure in vitro in Hs578T CD44 WT compared to CD44 KO cells at 6, 12, and 24 h time points. (D) HA production in vitro quantified via ELISA within CD44 WT and KO Hs578T cells. Experiments were repeated at least three times, represented by each data point. Statistical analysis was performed using Student's unpaired, twotailed t-test. Error bars represent standard error of the mean. P values * p < 0.05; ** p < 0.01; *** p < 0.001. Hs578T CD44 WT tumors (n = 10) and CD44 KO tumors (n = 14) took to form palpable lesions within the mammary fat pad of athymic nude mice. (B) Histological analysis of early tumors using H&E and immunofluorescence to identify HA in green (using hyaluronic acid binding protein), and nuclei in blue (using DAPI). Images were acquired on a Leica DM400B microscope, at either 100× or 400× magnification. (C) Scratch wound assay analyzing wound closure in vitro in Hs578T CD44 WT compared to CD44 KO cells at 6, 12, and 24 h time points. (D) HA production in vitro quantified via ELISA within CD44 WT and KO Hs578T cells. Experiments were repeated at least three times, represented by each data point. Statistical analysis was performed using Student's unpaired, two-tailed t-test. Error bars represent standard error of the mean. P values * p < 0.05; ** p < 0.01; *** p < 0.001.

CD44 Deletion in Hs578T Tumors Decreases the Number of Infiltrating CD206+ Macrophages
Because CCL2 (a known monocyte/macrophage chemokine) was reduced in CD44 KO cells, we sought to determine whether infiltrating macrophages differed between early tumors derived from CD44 WT and CD44 KO Hs578T cells. As illustrated by Figure 6A, there was no impact on the number of infiltrating F4/80+ cells; however, there was a reduction in CD206+ macrophages within the TME ( Figure 6B). Thus, our results suggest that CD44 signaling within tumor cells impacts the phenotype of infiltrating tumor-associated macrophages (TAMs), potentially by altering pro-inflammatory mediators. Images were acquired on a Leica DM400B microscope at 400× magnification. Statistical analysis was performed using Student's unpaired, two-tailed t-test. Error bars represent standard error of the mean. P values * p < 0.05, ns = not significant.

Increased HAS2 Gene Expression is Associated with Inflammatory and Stromal Biology Gene Signatures in Human Cases of Breast Cancer
Our findings thus far suggest that the HA-CD44 axis directly impacts the formation of an inflammatory microenvironment. To determine whether HAS2 expression is related to inflammation within human breast cancer samples, we examined the expression levels of inflammatory and other stromal-related genes in a cohort (n = 94) of human breast cancer samples ( Figure S7), using a custom 356 gene codeset on the NanoString nCounter platform ( Figure S6). We examined the normalized, log2-transformed expression of HAS2 and binned cases by its expression level being greater or less than 1 standard deviation away from the mean ( Figure S8). This resulted in 15 HAS2-high and 15 HAS2-low cases respectively, with the remaining 64 cases classified as intermediate. Differential gene expression testing comparing HAS2-high vs. HAS2-low tumor samples was performed, and we identified 123 genes with significantly altered levels (Supplementary Materials File S1). We noted that clinicopathologic variables such as ER status, Nottingham grade, and PAM50 molecular subtype were heterogeneously distributed across the cohort, ordered by HAS2 expression ( Figure 7A). Importantly, we observed significant enrichment of a focused inflammatory signaling gene signature in the cases, with all enriched genes demonstrating upregulation in the HAS2-high cases ( Figure  7A,B). Similarly, the differential expression of a larger signature encompassing genes involved in stromal remodeling and other stromal biologic processes in addition to inflammatory signaling was also significantly enriched in the HAS2-high cases ( Figures 7B and S3). Gene ontology enrichment testing showed significant overlap for inflammatory and stromal biologic processes among the genes upregulated in the HAS2-high group (Supplementary Materials File S1). Specifically, we noted higher levels of stromal related genes involved in collagen production (COL1A1 and COL6A1-3) and matrix remodeling (CMA1), among others ( Figures 7C and S9). Furthermore, the analysis of genes with wellestablished roles in inflammation demonstrated significantly higher expression levels in the HAS2high samples. These include IL6, IL1B, CXCL2, CXCL8/IL-8 and CD68 (a macrophage marker). Because CCL2 was not included on the focused Nanostring gene codeset, we were not able to directly assess its differential expression in these cases. However, these results clearly demonstrate an association between HAS2 and inflammatory genes, consistent with the in vitro findings (Figure 3). Additionally, HAS2-high samples had an increased expression of genes known to promote cancer, including KIT, STAT5A, and MYC. Overall, these findings confirm the association between elevated Images were acquired on a Leica DM400B microscope at 400× magnification. Statistical analysis was performed using Student's unpaired, two-tailed t-test. Error bars represent standard error of the mean. P values * p < 0.05, ns = not significant.

Increased HAS2 Gene Expression is Associated with Inflammatory and Stromal Biology Gene Signatures in Human Cases of Breast Cancer
Our findings thus far suggest that the HA-CD44 axis directly impacts the formation of an inflammatory microenvironment. To determine whether HAS2 expression is related to inflammation within human breast cancer samples, we examined the expression levels of inflammatory and other stromal-related genes in a cohort (n = 94) of human breast cancer samples ( Figure S7), using a custom 356 gene codeset on the NanoString nCounter platform ( Figure S6). We examined the normalized, log2-transformed expression of HAS2 and binned cases by its expression level being greater or less than 1 standard deviation away from the mean ( Figure S8). This resulted in 15 HAS2-high and 15 HAS2-low cases respectively, with the remaining 64 cases classified as intermediate. Differential gene expression testing comparing HAS2-high vs. HAS2-low tumor samples was performed, and we identified 123 genes with significantly altered levels (Supplementary Materials File S1). We noted that clinicopathologic variables such as ER status, Nottingham grade, and PAM50 molecular subtype were heterogeneously distributed across the cohort, ordered by HAS2 expression ( Figure 7A). Importantly, we observed significant enrichment of a focused inflammatory signaling gene signature in the cases, with all enriched genes demonstrating upregulation in the HAS2-high cases ( Figure 7A,B). Similarly, the differential expression of a larger signature encompassing genes involved in stromal remodeling and other stromal biologic processes in addition to inflammatory signaling was also significantly enriched in the HAS2-high cases ( Figure 7B and Figure S3). Gene ontology enrichment testing showed significant overlap for inflammatory and stromal biologic processes among the genes upregulated in the HAS2-high group (Supplementary Materials File S1). Specifically, we noted higher levels of stromal related genes involved in collagen production (COL1A1 and COL6A1-3) and matrix remodeling (CMA1), among others ( Figure 7C and Figure S9). Furthermore, the analysis of genes with well-established roles in inflammation demonstrated significantly higher expression levels in the HAS2-high samples. These include IL6, IL1B, CXCL2, CXCL8/IL-8 and CD68 (a macrophage marker). Because CCL2 was not included on the focused Nanostring gene codeset, we were not able to directly assess its differential expression in these cases. However, these results clearly demonstrate an association between HAS2 and inflammatory genes, consistent with the in vitro findings (Figure 3). Additionally, HAS2-high samples had an increased expression of genes known to promote cancer, including KIT, STAT5A, and MYC.
Overall, these findings confirm the association between elevated HAS2 expression and an inflammatory microenvironment in human breast cancer samples, representative of all molecular subtypes.

Discussion
Clinical studies have shown that chronic inflammation is a catalyst for cancer progression and thought to contribute to approximately 25% of all human malignancies [45]. As the important relationship between cancer cells and their microenvironment becomes increasingly evident, we sought to investigate CD44 as a regulator of inflammatory cytokines within TNBC, through interactions with its ligand, hyaluronan (HA), and subsequent effects on the TME.
Our results demonstrate that breast cancer cells synthesize and fragment HA and express CD44 on the cell surface. Additionally, our findings suggest that HA-CD44 interactions promote cancer-associated inflammation and contribute to early tumor formation. Importantly, our in vitro and in vivo results in a murine model were supported by data from human breast cancer cases, where increased HAS2 expression is significantly correlated with an inflammatory gene signature. These findings support previous reports, suggesting that high levels of HA within the tumor serve as a poor prognostic indicator, not only in breast cancer [6], but also in ovarian [46], prostate [14], colon [47], and gastric cancers [48], and lung adenocarcinomas [17,49]. Therefore, our results emphasize that HA-CD44 interactions potentially have broad applications across multiple cancers.
Consistent with the literature, HA was deposited throughout primary tumors derived from Hs578T and MDA-MB-231 cells ( Figure 2D). Therefore, we aimed to characterize HA machinery within these cells, to determine whether tumor cells contribute to stromal HA accumulation. Based on previous findings [25,26], we confirmed that both Hs578T and MDA-MB-231 cells synthesized HA (Figure 2A), suggesting that these tumor cells contribute to HA deposition within the TME. Additionally, HA accumulation was present within tumor nests in murine mammary tumors. Interestingly, we also noticed that in vivo HA deposition was heterogenous. These findings raise unique questions regarding the importance of HA-rich vs. HA-absent regions and how they affect surrounding malignant and infiltrating immune cells for further investigation.
While HA staining via immunofluorescence allows for the visualization of HA within cells, this technique does not allow for the identification of HA fragmentation. Therefore, we adapted a protocol by Cowman and colleagues [50], to isolate HA from conditioned medium and separate HA fragments via gel electrophoresis. To our knowledge, this is the first time anyone visualized the unique banding patterns of hyaluronan produced by two TNBC cell lines. Importantly, the Hs578T and MDA-MB-231 cells produced both fragments of both high and low molecular mass ( Figure 2B). Understanding the biological function of HA fragmentation is critical, as there is current interest in utilizing pegylated hyaluronidase (PEGPH20) to abolish the HA capsule surrounding a tumor and improve drug delivery in cancer patients. Recently, a Phase III clinical trial (NCT02715804) was performed using PEGPH20, in combination with first-line chemotherapeutics to treat patients with HA-high stage IV pancreatic adenocarcinoma [51]. Unfortunately, this study was terminated, as PEGPH20 did not improve response rate or overall survival. Further studies are needed to better understand mechanisms of toxicity surrounding PEGPH20 treatment, and to identify those patients who might respond to hyaluronidase therapy.
Additionally, it is crucial to understand how malignant cells within the tumor might sense the presence of HA within the TME. Along with being a primary HA receptor, CD44 is also highly expressed on breast carcinoma cells [21], suggesting that this transmembrane signaling protein is involved in the crosstalk between tumor cells and their surrounding environment. Upon the deletion of CD44 in breast cancer cells, we found decreased expression of genes involved in adhesion and cytokine activity (Figure 3). While the role of CD44 in adhesion is well documented [21,34,52,53], its role in cytokine synthesis and regulation is not as well-characterized. Therefore, we further validated the changes in inflammatory gene expression at the protein level using an inflammatory cytokine array (Figure 4). Furthermore, as was observed by limiting CD44 expression, 4MU-mediated knockdown of HA synthesis caused a significant decrease in CCL2 production in both TNBC cell lines, as well as IL-8 within the Hs578T cell line. Because CCL2 and IL-8 chemokines have known roles in angiogenesis and cancer-associated inflammation [36,[41][42][43]54], HA-CD44 signaling has important implications within human cases of TNBC. In addition, our findings agree with Karalis et al. [55], where HA depletion (using 4MU) within breast cancer cells resulted in a decrease in cytokine synthesis (MMPs, IL-6, IL-8). Interestingly, this group also discovered that CD44 expression decreased upon HA depletion in vitro, suggesting a unique feedback loop between HA-CD44 signaling and inflammatory cytokine production.
Because others have demonstrated that CD44 can modulate the signaling pathways associated with the regulation of inflammatory factors (such as STAT3, PI3K-Akt, MAPK/ERK pathways) [22], we aimed to assess changes in canonical inflammatory signaling pathways within our CD44 KO cells via immunoblot analysis. However, we did not observe significant changes in the activation of known inflammatory signaling pathways examined between CD44 WT and KO cells, including Src, ERK, NF-κB (p65, p105, IκBα), p38, and STATs 1,3, and 5. These findings could be due to a number of reasons, including alternative and non-canonical mechanisms regulating the pathways that may be impacted. For example, both CD44 and RHAMM have been shown to alter the duration of downstream ERK signaling [5,56], which may not be readily observable, by assessing steady state levels of activation via immunoblot analysis. In addition, CD44 can be post-translationally modified via proteolytic cleavage, releasing extracellular and intracellular domain fragments [57]. The intracellular domain (ICD) fragment migrates to the nucleus and acts as a transcription factor for various genes, including MMP9 [21,58] (known to cleave CD44) and CD44 itself [58]. Thus, this mechanism could also explain how a transmembrane receptor lacking kinase activity is able to alter gene expression. Further studies are required to fully understand the mechanism through which the HA-CD44 axis modulates inflammatory cytokine expression.
In breast cancer, the "reactive" TME is maintained by pro-inflammatory cytokines secreted both by tumor cells and their surrounding stroma. Our results suggest that tumor cells responding to HA through CD44 help maintain a chronic inflammatory state by regulating cytokine synthesis. This finding is important, since cytokine signaling leads to the recruitment and activation of a myriad of immune cells within the TME. Notably, production of the chemokine CCL2 within breast cancer cells was significantly decreased in conditions where either CD44 was deleted or HA was depleted ( Figure 4). CCL2 is relevant as it has been shown to recruit monocytes into the primary tumor, where they can differentiate into tumor-associated macrophages (TAMs) [54,59]. These infiltrating immune cells have been linked with decreased overall survival in breast cancer patients, promoting primary tumor growth and metastatic dissemination [60]. Thus, the inhibition of HA-CD44 signaling associated with inflammation may disrupt macrophage infiltration and improve survival in breast cancer patients.
More specifically, we noticed a dramatic reduction in early tumor formation using the Hs578T orthotopic transplant model when CD44 was deleted from cancer cells. Furthermore, this finding was accompanied by a decrease in CD206+ macrophages within the TME ( Figure 6). CD206+ positive macrophages are typically associated with tumor promoting functions including the promotion of angiogenesis, stromal remodeling and other tumorigenic functions [61][62][63]. These results support a model in which HA-CD44 associated inflammation and macrophage infiltration may, in part, be responsible for early tumor formation. However, future studies are required to determine whether CD206+ macrophages impact stromal modulation and/or vessel formation in early tumor formation.
Additionally, in vitro experiments demonstrated that Hs578T cells lacking CD44 exhibited impaired wound closure and increased HA production ( Figure 5C,D), factors that may also contribute to the delay in tumor formation. Recent reports [64,65] suggest that CD44 is able to bind and internalize HA, which would explain why HA production appears to be increased when CD44 is deleted within Hs578T cells. However, since HA can be tumor preventing or promoting, further studies are needed to determine whether HA being produced by the CD44 KO cells may have tumor suppressive properties.
It is also important to note that we did not see an effect on early tumor formation within the MDA-MB-231 orthotopic transplant model when CD44 was deleted from the cancer cells. This finding may be due to intrinsic differences between the two breast cancer cell lines. While Hs578T cells are derived from a primary tumor, the MDA-MB-231 cell line is derived from a distant, metastatic site [44]. Because the MDA-MB-231 cells may be too aggressive to model early tumor formation in breast cancer, we investigated whether CD44 expression impacted metastatic lesion formation in vivo using a tail vein injection model [66,67]. Through this work, we discovered that mice receiving CD44 KO MDA-MB-231 cells had a significant reduction in lung colonization when compared to CD44 WT controls. This could be due to a number of factors, including decreased CCL2 signaling (limiting metastatic nice formation) [68][69][70], along with changes in tumor cell adhesion [34].
Due to the growing body of literature utilizing CD44 as a marker of cancer cell stemness [71][72][73][74], we were somewhat surprised to find that time to tumor endpoint (2 cm 3 ) in vivo did not differ between CD44 WT and KO conditions in either breast cancer cell line. These differences could be due to a number of factors. Because CD44 is often used as a marker of cancer stem cells, findings are often correlative rather than functional. Interestingly, HAS2 knockdown within cancer stem-like (CD44 + /CD24 − ) MDA-MB-231 BoM cells has been shown to reduce both the number and growth of metastatic lesions in vivo [75]. This finding might suggest a role for additional HA receptors in maintaining cancer stem cell populations. Overall, our data support others [34], in which CD44 contributes to TNBC aggressiveness through adhesion and cytokine synthesis.
Finally, we found a correlation between increased HAS2 gene expression and a pro-tumorigenic inflammatory gene signature in primary human breast cancer samples (Figure 7). Our findings agree with two independent reports [41,76], in which inflammatory gene signatures were discovered within TNBC patients. Together, these findings suggest that the HA-CD44 axis may directly drive the expression of this inflammatory gene signature identified within triple negative cases. Gene ontology also revealed that HAS2-high samples demonstrated differential expression of stromal-associated genes, including genes involved in collagen production (COL1A1 and COL6A1-3) and matrix remodeling (CMA1), among others. This correlation is intriguing, as it might imply that the deposition of various matrices is heavily dependent on one another in human cancers.
Ultimately, our findings suggest that HAS2 expression supports a chronic inflammatory state within human breast cancer, which has been shown to foster tumor progression. These results also explain, in part, why HAS2 expression is linked to poor patient survival in breast cancer patients [8].

Mice
Athymic nude mice (Foxn1) were purchased from the Jackson Laboratory (Manassas, VA, USA). All experiments were performed using 6-8-weeks-old female mice, housed in specific pathogen-free facilities. All animal care and procedures were approved by the Institutional Animal Care and Use Committee of the University of Minnesota (Approval #1909-37381A) and were in accordance with the procedures detailed in the Guide for the Care and Use of Laboratory Animals [77].

Nanostring Analysis of Human Breast Cancer Samples
Tissue specimens were obtained following the approval of the University of Minnesota Institutional Review Board (Approval #1409E53504). A retrospective review of surgically-resected female breast cancer cases in a 30-month window between 2011-2013 was performed to select potentially eligible samples. Samples were selected from sequential cases in the pathology record up to pre-determined quotas of hormone receptor subtypes as follows: 50% ER+, 30% HER2+, and 20% triple negative breast cancer. Histologic slides were reviewed by a pathologist to ensure a minimum of 20% tumor cellularity on tissue sections submitted for analysis. De-identified clinicopathologic disease features were recorded from the pathology record, in accordance with the IRB-approved protocol. Formalin-fixed paraffin embedded tissue was extracted with Qiagen FFPE RNeasy reagents (Qiagen, Germantown, MD, USA), following the manufacturer's protocols. Nanostring gene expression analysis was performed by the University of Minnesota Genomics Core using a custom designed 356 gene codeset (Supplementary Materials File S1). Overall, 120 primary breast cancer samples had sufficient cellularity and RNA yield for subsequent Nanostring gene expression analysis; the average age of this cohort was 58 (range 30-97). In addition, 35 breast cancer cell lines (obtained through ATCC) were also analyzed. Data quality control analysis was completed in R (v 3.4.4), using the NanoStringQCPro (version 1.10.0) package, which excluded 26 patient samples due to low gene detection (i.e., missing >20% of genes surveyed by the codeset). The final patient cohort was therefore comprised of 94 samples, and patient characteristics are described in Figure S7 HAS2-low samples was tested and significant DE genes were identified (absolute value log2-fold change ≥0.5 and Benjamini and Hochberg adjusted p-value < 0.05). Gene ontology over-representation analysis was performed in R using the enrichGO function from the clusterProfiler (version 3.14.0) package, where significant (q < 0.05) terms from the CC, BF, and MF groups were identified. General over-representation analysis (one-sided Fisher's Exact test) was performed using the enricher function in clusterProfiler, where significant DE gene lists (up, down, or all) from the HAS2-high vs. HAS2-low comparison were tested for enrichment of our custom genesets (Stromal Biology or Inflammatory Signaling, listed in Supplementary Materials File S1). Significant enrichment was considered for tests with q < 0.05.

CD44 Knockout vis CRISPR/Cas9
Using methods provided by Church and colleagues [78], gRNA was designed to target exons 1 and 5 of the human CD44 sequence adjacent to an NGG PAM sequence.

CD44 Knockout
Cells were each plated into 6 cm dishes at 1E6 cells per plate, using 1 µg/mL puromycin to maintain selective pressure. The next day, experimental plates were transfected with 2 µg Exon 1 gRNA and 2 µg Exon 5 gRNA per plate in X-tremeGENE ™ HP DNA Transfection Reagent (Roche, Mannheim, Germany), while control plates were treated with transfection reagent only. Twenty-four hours later, Cas9 was induced via 1 µg/mL doxycycline treatment. Following a 24-h incubation, gRNA transfection was repeated, and cells were incubated overnight. CD44 knockout and control cells were

RNA Extraction and Sequencing
Cells were plated into a 6-well plate at 500,000 cells/well in complete medium. The next day, cells were starved in 1% serum for 24 h. RNA was extracted using the RNeasy Mini Kit (Qiagen #74104, Germantown, MD, USA) per manufacturer's instructions, and RNA was quantified using UV spectroscopy. Triplicate RNA samples were submitted to the University of Minnesota's Genomics Core for Illumina Sequencing. Twelve dual-indexed Illumina TruSeq stranded mRNA libraries were pooled for sequencing on 1 lane of a HiSeq 2500, using v4 chemistry to generate 2 × 50 bp paired-end reads.

RNA-Seq Differential Gene Expression and Pathway Analysis
Fastq files were mapped to the GRCh38 human reference genome using HiSat2 (version 2.1.0) [79]. Counts were estimated using the Subread featureCounts (version 1.6.2) tool [80], using the gencode.v28.GRCh38.annotation.gtf annotation [81]. Count data were filtered to only keep genes that were less than 300 nt in length and had a cpm (counts per million) value greater than 1 cpm in at least 2 sample replicates, across all experimental conditions. The quasi-likelihood test was used to evaluate differential expression (DE) with edgeR (version 3.22.5) [82,83]. The Benjamini-Hochberg method was used to adjust p-values for multiple hypothesis testing and an adjusted p-value ≤ 0.01, with a log2 fold change > 0 was used as a DE significance threshold. For gene ontology (GO) pathway analysis, we used the R package clusterProfiler (version 3.8.1) [84,85].

Inflammatory Cytokine Array
Cells were plated in a 6-well plate (500,000 cells/well) overnight in complete medium. Cells were starved in 1% serum for 24 h and conditioned media were collected for analysis by R&D Systems™ Proteome Profiler Human Cytokine Array Kit (#ARY005B), according to the manufacturer's instructions.

ELISAs
Cells were plated in a 6-well plate (200,000 cells/well) overnight in complete medium. Cells were starved of serum for 24 h and conditioned media were collected for analysis by ELISA per manufacturer's instructions. HA production was quantified using an HA ELISA (Echelon Biosciences #K-1200, Salt Lake City, UT, USA), CCL2 production was quantified using the human CCL2/MCP-1 Quantikine ELISA Kit (R&D Systems #DCP00, Minneapolis, MN, USA), and IL-8 production was quantified using the human CXCL8/IL-8 Quantikine ELISA Kit (R&D Systems #D8000C, Minneapolis, MN, USA).

Inhibition of HA Synthesis
HA was depleted from cultured cells using 4-Methylumbelliferone sodium salt (Sigma-Aldrich #M1508), dissolved in methanol. Cells were treated with 600 µM 4MU for 24 h and HA depletion was confirmed using an HA ELISA, as described above.

Scratch Assay
Hs578T and MDA-MB-231 CD44 WT and KO cells were plated in a 6-well plate using complete medium. Once cells reached confluency, a scratch was made using a p200 pipette tip, as previously described [86]. Images were taken at 0, 6, 12, and 24 h, and percent wound closure was calculated relative to the initial time point. Difference in wound closure between CD44 WT and KO cells was determined.

In Vivo Experiments
In the in vivo experiments, 2.5E6 cancer cells were resuspended in 50% Matrigel/PBS solution and orthotopically injected into the fourth mammary fat pad of athymic nude mice.

Effects of CD44 on Animal Survival
Once palpable, tumors were calipered every other day, to determine growth rate and total tumor volume. Animals were humanely euthanized using CO 2 once tumors reached endpoint (2 cm 3 ). Tumors were excised for histologic analysis.

Effects of CD44 on Early Tumor Formation
Four weeks following injection, animals were humanely euthanized, and mammary glands were harvested for histologic analysis.

Effects of CD44 on Lung Colonization
Using methods previously described [66,87], 750,000 cells were resuspended in 200µL PBS (per animal) and injected into the tail vein of athymic nude mice. Mice were sacrificed at seven weeks and lungs were collected for histologic analysis.

Microscope Imaging
All images were acquired on a Leica DM400B microscope (Leica, Wetzlar, Germany), at either 20× or 40× objectives. Images were acquired using a Leica DFC310 FX camera (Leica, Wetzlar, Germany) and LAS V3.8 software. Five images of at least 3 representative tumors were analyzed.

Statistical Analysis
Statistical analysis was performed using Student's unpaired, two-tailed t-test. Comparisons between multiple groups were performed using one-way ANOVA with Dunnett's multiple comparisons test. Error bars represent standard error of the mean (SEM).

Conclusions
HA-CD44 interactions promote cancer-associated inflammation in triple negative breast cancer. Although further investigation is needed, the disruption of HA-binding and subsequent signaling through CD44 may be an exciting new approach for the treatment of triple negative breast cancer.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/5/1325/s1, Figure S1: Large format copies of the 400× images in Figure 1C, Figure S2: Expression of HAS2 and CD44 in breast cancer cell lines, Figure Figure S6: Flow chart highlighting the case selection process used within the Nanostring analysis of human breast cancer samples, Figure S7: Patient characteristics of the final 94 patient samples included within the Nanostring analysis, Figure S8: HAS2 expression across 94 human breast cancer cases, Figure S9: Stromal biology gene expression in human breast cancers, Figure S10: Full length western blot of CD44 in Hs578T CD44 WT and CD44 KO cells, Figure S11: Full length western blot of β-tubulin loading control in Hs578T CD44 WT and CD44 KO cells, Figure S12