Evaluation of the hyaluronan biosynthetic and degradative axis using human protein atlas data

The human protein atlas V21.0 (Uhlén et al. 2015) was used to initially characterize the protein expression of the enzymes studied in this project comparing normal ovarian tissue with the expression in tumour tissue (Supplementary Figure S1). HAS1 was lowly expressed in normal tissue in ovarian stromal cells and was undetectable in follicular cells. In comparison, HAS1 expression in tumour cells was low to medium measured in 5 of 12 samples. HAS1 was not detected in the remaining samples. HAS2 could not be evaluated with the human protein atlas because protein data based on antibody staining were not available. HAS3 expression could not be detected in normal tissue in ovarian stromal cells and only to a low level in follicular cells. Tumour tissue showed a similar tendency. Low HAS3 expression was measured in one sample, while it was not detected in any of the other samples. HYAL1 and HYAL2 expression could not be detected in the ovarian stromal cells or follicular cells of normal tissue. In the tumour tissue, 3 of 12 and 11 samples, respectively, showed low to medium HYAL1 and HYAL2 expression, being undetectable in the remaining samples. For HYAL3 low-intensity expression was detected in follicular cells in normal tissue. No expression was shown for ovarian stromal cells. 3 of 11 samples showed low to medium expression in ovarian tumour tissue. HYAL4 could be detected at low intensity in ovarian stromal cells and not in follicular cells. In tumour tissue, all samples showed HYAL4 expression, with low expression in 8 samples and medium expression in 3 probes. PH20 expression could not be detected in 8 normal or tumour tissue using the antibodies and staining conditions utilized in the human protein atlas V21.0 project. In summary, normal tissue showed no or only a low expression of the investigated enzymes, whereas in tumour tissue a slightly increased expression of some of the enzymes was detectable (see Supplemental Figure S1 for details). High protein expression was not measured in any of the tissue samples for all enzymes examined.

The gene expression levels of HAS1, HAS2 and HYAL1-4 are significantly altered in ovarian tumour tissue compared to control tissue

We next aimed at investigating a potentially altered expression of the HA biosynthetic and degradative axis in a larger collective evaluating mRNA expression as a more robust means of quantification. The TNMplot database (Bartha and Győrffy 2021) was used to compare the gene expression of the enzymes studied in this project in normal tissue (n = 46) and tumour tissue (n = 744) of the ovary. The database did not include data for HAS3, PH20 and HYALP1, so these genes could not be examined. Significant changes in gene expression were seen for all genes examined, as shown in Fig. 1. HAS1 showed reduced gene expression in ovarian cancer tissue with a fold change of 0.11. The expression of HYAL1 (fold change = 0.69) and HYAL4 (fold change = 0.86) also was slightly reduced. In contrast, the expression of HAS2 (fold change = 1.42), HYAL2 (fold change = 1.84) and HYAL3 (fold change = 1.94) was significantly increased in the tumour tissue (Fig. 1).

Fig. 1

The gene expression levels of HAS1, HAS2 and HYAL1-HYAL4 of normal ovarian tissue (n = 46) compared with tumour tissue of the ovary (n = 744). While HAS1, HYAL1 and HYAL4 expression levels were significantly lower in tumour tissue compared to control ovary, the expression of HAS2, HYAL2, and HYAL3 were significantly upregulated in tumours compared to controls. The data are derived from the TNMplot database (Bartha and Györffy 2021) (https://tnmplot.com/analysis/, accessed on 24 Feb 2022)

HAS2, HYAL2 and HYAL3 have a differential impact on the survival of ovarian cancer patients

In this project, the Kaplan–Meier-Plotter database was used to present the influence of various enzymes of the HA system on the OS and PFS of patients. The number of specific patient cases per classification is shown in Table 1. The original patient collective was described in reference (Győrffy et al. 2012). HAS2 had a significant negative impact on survival in terms of both, OS and PFS (Table 1, Fig. 2a, b). The HR (hazard ratio) of OS was 1.23 (p-value = 0.0019) and the one of PFS was 1.14 (p-value = 0.042). A positive correlation was given to the high expression of HYAL2 and HYAL3 referred to PFS. The HR was 0.86 for both with a p-value of 0.023 for HYAL2 (Table 1, Fig. 2c) and 0.024 for HYAL3 (Table 1, Fig. 2d). Furthermore, subgroup analysis was done to find out whether enzymes had a particular stronger influence on certain patient groups. A distinction was made between histology, staging, grading and different chemotherapy approaches. Histology was subdivided into serous and endometrioid ovarian cancer, staging into stage I + II compared to III + IV and grading into grade I + II compared to III. The chemotherapy approaches were taxol compared to cisplatin and the combination of taxol and cisplatin. The results of these subgroup analyses are shown in Table 2 (HAS1-3), Table 3 (HYAL1-3) and Table 4 (HYAL4, PH20) and in Table S3 (HYALP1) in the supplement information.

Table 1 Correlation between the expression of HAS1-HAS3, HYAL1-HYAL4, PH20 and HYALP1 and the overall survival (OS) or progression free survival (PFS) of ovarian cancer patientsFig. 2

Prognostic value of HAS2, HYAL2 and HYAL3 for the survival of patients with ovarian cancer. The analysis was done by the Kaplan–Meier-Plotter. For each enzyme the Kaplan–Meier-curve, the hazard ratio (95% confidence interval) and the p-value were given. a OS in correlation with HAS2 expression (n = 1656), b PFS in correlation with HAS2 expression (n = 1435), c PFS in correlation with HYAL2 (LuCA-2) (n = 1435), d PFS in correlation with HYAL3 (n = 1435)

Table 2 Correlation between the expression of HAS1-HAS3 and the OS or PFS of ovarian cancer patientsTable 3 Correlation between the expression of HYAL1-HYAL3 and the OS or PFS of ovarian cancer patientsTable 4 Correlation between the expression of HYAL4, PH20 and HYALP1 and the OS or PFS of ovarian cancer patients

Table 2 shows that high expression of HAS1 (HR = 1.3, p-value = 0.00033) had a negative impact on PFS for serous ovarian cancer patients. Furthermore, HAS2 had a negative correlation with the OS (HR = 1.26, p-value = 0.0027) and PFS (HR = 1.31, p-value = 0.00021). High expression of HYAL3 was correlated with a better OS (HR = 0.83, p-value = 0.016), which is shown in Table 3. It is not possible to compute a Hazard Rate in case there is no event in one of the cohorts defined by the gene expression, as the HR will be either 0 or infinite in these cases. In such cases, we adjusted the HR to < 0.1. For patients with endometrioid ovarian cancer expression of HYAL2 (HR = 0.17, p-value = 5.00 × 10–4) and PH20 (HR = 0.3, p-value = 0.015) was associated with better PFS (Tables 3, 4).

Referred to staging, no correlation of the HA-associated genes was found for patients in staging I + II (Table 2–4). High expression of HAS1 (HR = 1.24, p-value = 0.0033) was associated with worse PFS of patients in staging III + IV, as shown in Table 2. For staging III + IV HAS1 also correlated with worse OS (HR = 1.18, p-value = 0.03) and PFS (HR = 1.2, p-value = 0.012) (Table 2). A positive influence on the OS had the expression of HYAL3 (HR = 0.83, p-value = 0.016) and HYAL4 (HR = 0.85, p-value = 0.035) as shown in Tables 3 and 4.

For patients in grade I + II high expression of HAS2 showed a negative impact on the OS of ovarian cancer patients (HR = 1.44, p-value = 0.013). For grade III there was a worse PFS for patients with high expression of HAS1 (HR = 1.29, p-value = 0.0029) and HAS2 (HR = 1.29, p-value = 0.0027) (Table 2). High expression of HYAL3 (HR = 0.69, p-value = 0.011) had a positive association with the OS of patients (Table 3).

Finally, we analysed the prognostic impact of HA pathway constituents in patients related to chemotherapy treatment with the combination of taxol and cisplatin. The expression of HAS1 had a negative impact on the PFS of patients (HR = 1.25, p-value = 0.011) (Table 2). The expression of HAS2 was associated with worse OS (HR = 1.21, p-value = 0.045) and PFS (HR = 1.23, p-value = 0.019) (Table 2). Correlations in the context of treatment with taxol or cisplatin can be found in the supplementary information (Supplementary Table S2).

In conclusion, HAS2 appeared to be the enzyme of the HA system with the biggest impact on the survival of ovarian cancer patients. Therefore, we decided to study the functional impact of HAS2 depletion using an in vitro siRNA approach in human ovarian cancer cell lines.

HAS2 depletion results in a moderate cell-type specific dysregulation of HYAL3

As a first step, a panel of ovarian cancer cell lines (i.e., SKOV3, Caov-3, SW 626 and PA-1) was analyzed for the expression of diverse genes involved in HA metabolism. These cell lines correspond to the ATCC ovarian cancer panel with varying degree of genetic complexity. HAS1, HAS2 and HYAL3 expression levels showed to be comparable in all adenocarcinoma cell lines SKOV-3, Caov-3 and SW 626 cells, whereas the teratocarcinoma cell line PA-1 cells expressed higher levels of all the three genes (Figs. 3a, b, e). As for HAS3, the expression was slightly higher in Caov-3 cells, with respect to all the other cell lines (Fig. 3c). HYAL2 expression was comparable in SKOV3 and SW 626 and substantially higher with respect to Caov-3 and PA-1 cells (Fig. 3d). Since SKOV3 and SW 626 displayed very similar gene expression profiles related to HA metabolism and showed the highest HAS2 expression in this panel, we decided to focus on these lines for our subsequent experiments.

Fig. 3

HAS1, HAS2, HAS3, HYAL2 and HYAL3 gene expression levels in SKOV3, Caov-3, SW 626 and PA-1ovarian cancer cells, as measured by qPCR. The mean value was given with the standard error. Data represent the results of 3 independent experiments with n = 2 or 3 independent replicates under the same conditions. The gene expression levels are shown relative to the expression in PA-1 cells, which were set to a mean value of 1

At this point, we first asked whether HAS2 knockdown affects the expression of other enzymes of the HA system. For this purpose, we used siRNA transfection and qPCR to detect the HAS2 knockdown and to compare the expression of HAS1, HAS3, HYAL2 and HYAL3 in HAS2 knockdown cells with the expression in control cells. These enzymes were chosen because they showed the greatest impact on patient survival in the Kaplan–Meier-Plotter. It was shown that the knockdown of HAS2 resulted in a significant (p < 0.001) and substantial downregulation (> 75%) of its expression in both SKOV3 and SW 626 cell lines (Fig. 4a, b, Supplementary Figure S2). No HAS2 expression was detectable in 6 samples of SKOV3 cells in the HAS2 knockdown group, suggesting that the expression rate was below the detection limit. While these data speak for a successful knockdown, these values were not included in the calculation. Evaluation of the kinetics of HAS2 knockdown revealed that HAS2 mRNA levels were substantially and significantly downregulated by > 75% for 24 h and 48 h after knockdown, returning to basal levels 4d–7d after transient transfection (Supplementary Figure S2). The measurement of pericellular HA confirmed that both SW 626 and SKOV3 cells had significantly lower amounts of pericellular HA upon HAS2 silencing with respect to control cells (Fig. 4c, d). The HAS2 knockdown did not have a significant influence on the expression of HAS1, HAS3 and HYAL2. HYAL3 was marginally downregulated by about 20% in SW 626 cells (p < 0.05), whereas it was upregulated to a similar extent in SKOV3 cells (p < 0.05) (Fig. 4a, b). mRNA levels of the HA receptors CD44 and RHAMM, and the HA binding proteoglycan versican (VCAN) were not changed upon HAS2 silencing with respect to control in both cell lines (Supplemental information, Figure S3).

Fig. 4

Impact of HAS2 knockdown and its influence on HA production and the expression of HAS1, HAS3, HYAL2 and HYAL3 in SW 626 and SKOV3 ovarian cancer cells. a, b qPCR confirmation of HAS2 knockdown and impact on the expression of HA-related genes. The mean value is given with the standard error. Data represent the results of 4 independent experiments with n = 2 or 3 independent replicates under the same conditions. a The mean value was calculated out of 7 values, in 6 HAS2 knockdown samples the HAS2 expression levels were under the limit of detection. c Representative images of particle exclusion assay of SW 626 and SKOV3 pericellular space as readout of HA production. Cells were transfected for 24 h with 20 nM siRNA against HAS2 or scrambled control siRNA or treated with 2 U/ml Streptomyces hyalurolyticus hyaluronidase. d, e Quantification of HA pericellular matrix for SW 626 and SKOV3 cell lines. Data are shown as mean ± SEM of three independent experiments. Results are expressed as the ratio between the area of ECM delimited by red blood cells and the area of the cell by using ImageJ software. *p ≤ 0.05, ***p ≤ 0.0001

Furthermore, the influence of HAS2 knockdown on the expression of HAS1, HAS3, HYAL2 and HYAL3 in cells treated with chemotherapy was investigated. A distinction was made between therapy with taxol, cisplatin and the combination of taxol and cisplatin. With all therapies, no significant difference was found with regard to the expression of HAS1, HAS3, HYAL2 and HYAL3. More detailed data are provided in the Supplementary Figure S4. In conclusion, we could prove a successful HAS2 knockdown in SKOV3 and SW 626 cells. As a result, HYAL3 was moderately, yet significantly dysregulated in a cell-type specific manner.

HAS2 knockdown affects the formation of tumour cell spheroids

HA is an important factor that influences cell cohesion and stability. Moreover, a role for HA in cancer stem cell function has been described (D. Vitale et al. 2019a, b). To test the possible influence of HAS2 knockdown on the capability of ovarian cancer cells to form tumour spheroids, a hanging drop assay was performed. In the hanging drop method, these factors were represented by the size of the area and the perimeter of cell spheres in each drop. Differences were between control and HAS2 siRNA treated cells were analyzed regarding the area and perimeter of the spheres and area and perimeter of the spheres plus a diffuse edge/margin that was visible under some of the treatment conditions. Spheroids of control SKOV3 cells and HAS2 knockdown cells on day 4 and day 7 are shown in Fig. 5c We found that the area of spheres of HAS2 knockdown cells was significantly smaller on day 4 (p-value = 0.017) (Fig. 5a). Furthermore, the perimeter of these was smaller for HAS2 knockdown cells compared to control cells on both days (day 4 p-value = 2.72 × 10–4, day 7 p-value = 0.035) (Fig. 5b). For values including the diffuse edge, the area of HAS2 knockdown cells was significantly higher on day 4 (p-value = 0.047) and day 7 (p-value = 4.97 × 10–10) (Supplementary Figure S5a). Referred to the perimeter, measured values were significantly higher for HAS2 knockdown cells on day 7 (p-value = 4.25 × 10–5) (Supplementary Figure S5b). Besides this, we observed that a diffuse edge was formed in 25% of the drops with control cells on day 4 and in 54% on day7 (Supplementary Figure S5c). Compared to this HAS2 knockdown cells formed a diffuse edge in 76% of the drops on day 4 and 100% on day 7 (Supplementary Figure S5c). In the evaluation of the diffuse edge, only the drops that formed an edge were included. Two drops of control cells on day 4 and 4 drops on day 7 did not form spheroids. HAS2 knockdown cells did not form a spheroid in 2 drops for both days. These samples were not included in the results. SW 626 cells failed to form proper spheroids in the hanging drop assay, as only loose cell aggregates were seen, precluding an analysis of HAS2 depletion in this assay (Supplementary Figure S6). To conclude, we found out that HAS2 knockdown SKOV3 cells formed significant smaller spheres with bigger edges, especially on day 7. Furthermore, knockdown SKOV3 cells formed this edge more often.

Fig. 5

Impact of HAS2-depletion on the sphere formation capacity of SKOV3 cells. Hanging drop method was used to show differences in cell cohesion and sphere formation capability of HAS2 knockdown SKOV3 cells compared to control SKOV3 cells. a, b Area and perimeter of the spheres excluding the diffuse edge. The area or perimeter of spheres of HAS2 knockdown and control cells was measured at day 4 and day 7. *p ≤ 0.05, c Representative pictures of spheres in drops of HAS2 knockdown cells and control cells. Note presence of a solid dark core and a light, diffuse edge. The values were built out of data of 4 experiments á 12 drops for HAS2 knockdown and control cells (n = 48). AU arbitrary unit

Impact of HAS2 siRNA depletion on cell viability and the response to chemotherapy

As our Kaplan–Meier-Plotter analysed had indicated an impact of HAS2 on OS and PFS of patients with prior chemotherapy treatment, we analysed the impact of HAS2-depletion in SKOV3 cells subjected to different concentrations of chemotherapy in vitro. The MTT Assay was used as a well-established and robust assay (Sargent 2003) to assess whether the ovarian cancer cell viability is influenced by HAS2 knockdown. In all chemotherapy treatment conditions, 8–11 serial 1:2 dilutions of the combinatorial treatment with taxol and cisplatin were applied to control cells and HAS2 knockdown cells (Fig. 6).

Fig. 6

Viability of HAS2 knockdown and control ovarian cancer cells at different concentrations of taxol and cisplatin, measured by MTT assay. a SKOV3 cells, b SW 626 cells. All values are given in % based on the concentration of control cells at 0.00 nM chemotherapy treatment. Results represent mean value ± SD for 3 experiments under same conditions. *p ≤ 0.05, **p ≤ 0.001

Starting with the lowest concentration of chemotherapy the therapeutics had a concentration of 0.00 nM. The viability of HAS2 knockdown SKOV3 cells under these basal conditions was 46.68% lower compared to control cells (p-value = 0.0002) (Fig. 6a). A significant value was measured at 0.6703 nM taxol and 0.0744 nM cisplatin. The viability of HAS2 knockdown SKOV3 cells was 38.7% smaller (p-value = 0.022) (Fig. 6a). Similar effects were seen in SW 626 cells, where HAS2 knockdown resulted in a significant decrease in cell viability, albeit to a lesser extent (16%, p < 0.01) compared to SKOV3 cells (Fig. 6b). The significant difference in viability of HAS2 knockdown and control SW 626 cells persisted upon treatment with 0.16758 nM taxol and 0.0186 nM cisplatin. All in all, in both cell lines tested, the viability of HAS2 knockdown cells was lower than control cells, indicating that the impact of HAS2 depletion alone on cell viability was higher than a possible effect of HAS2 on the chemotherapy response under our assay conditions. The effects on SKOV3 cell viability of taxol or cisplatin treatment alone are shown in the supplementary information (Supplementary Figure S7).

Finally, we evaluated if HAS2 depletion may affect the migratory capacity of ovarian cancer cells, employing a scratch wound assay. No significant impact on SKOV3 migration was observed upon HAS2 knockdown (Supplementary Figure S8). SW 626 cells detached as cell sheets at the scratch wound margins, precluding meaningful quantitative analysis in this assay (data not shown).

String analysis reveals the interconnection of the HA system and pathogenetic factors in ovarian cancer

Our last step of the analysis was the use of the STRING tool to show interactions of HAS1-3 and HYAL1-5 (Fig. 7) between each other and the 10 closest interactions with other proteins. For each protein, the interactions were analyzed related to gene neighbourhood, gene fusions, gene co-occurrence, experimentally determination, curated databases, co-expression, protein homology and text mining. HYALP1 was not analyzed by the STRING tool. Referred to the HA system it was shown that there is high interaction between HAS2 and HAS3. HAS1-3 interacted with PH20 due to co-expression and text mining. Text mining indicated an interaction between HAS1-3 and HYAL2 and HYAL3 and between HAS2 and HYAL4 (Fig. 7).

Fig. 7

STRING analysis for protein–protein interactions of HA pathway constituents. With the use of STRING database (https://string-db.org) the interactions of the proteins, analyzed in this study, are shown. Medium confidence threshold of 0.004

Related to other proteins, especially HAS2 and HAS3 showed high interaction with UGDH, CD44 and HMMR. Besides this, HAS1 and HAS2 cooperated with VCAN. In addition, a strong interaction between PH20 and ADAM2 was evident (Fig. 7).

For HYALs, a high interaction was found between HYAL2 and CD44, HMMR, Macrophage stimulating 1-receptor (MST1R) and WWOX. PH20 showed high interaction with ADAM2 (Fig. 7).

The other proteins shown in Fig. 7 are interacting in less relevant size. Therefore, they were not named in detail.