TGF-β Increases MFGE8 Production in Myeloid-Derived Suppressor Cells to Promote B16F10 Melanoma Metastasis

There is growing evidence that myeloid-derived suppressor cells (MDSCs) are directly involved in all stages leading to metastasis. Many mechanisms for this effect have been proposed, but mechanisms of coregulation between tumor cells and MDSCs remain poorly understood. In this study, we demonstrate that MDSCs are a source of milk fat globule-epidermal growth factor (EGF) factor 8 (MFGE8), which is known to be involved in tumor metastasis. Interestingly, TGF-β, an abundant cytokine in the tumor microenvironment (TME), increased MFGE8 production by MDSCs. In addition, co-culturing MDSCs with B16F10 melanoma cells increased B16F10 cell migration, while MFGE8 neutralization decreased their migration. Taken together, these findings suggest that MFGE8 is an important effector molecule through which MDSCs promote tumor metastasis, and the TME positively regulates MFGE8 production by MDSCs through TGF-β.

Metastasis is a complex process through which tumor cells spread to other organs. It progresses through several steps, including epithelial-mesenchymal transition (EMT), local invasion, neovascularization, and pre-metastatic niche formation [17][18][19][20]. MDSCs have been studied as an important regulator of tumor metastasis. For example, MDSCs promote tumor growth by remodeling the TME to facilitate angiogenesis [21,22]. MDSCs are also known to establish a pre-metastatic niche and induce EMT [19,20]. These data highlight that elucidating the mechanisms through which MDSCs communicate with tumor cells is important for understanding tumor progression and for the development of anti-tumor therapies.
Previously, we demonstrated the importance of TGF-β in the immunosuppressive function of MDSCs during inflammation [7]. However, TGF-β is also an important component of the TME and a mediator of tumor progression and immune modulation in cancer [23][24][25]. Thus, we considered the possibility that TGF-β-induced MDSCs activation could also affect the tumor-related response. Microarray analysis identified several potential targets, including milk fat globule-epidermal growth factor (EGF) factor 8 (MFGE8).

Retroviral Overexpression
To overexpress Mfge8, retrovirus was generated from Phoenix Eco cells by transient transfection with a retroviral pMigR1_GFP vector. Bone marrow cells were seeded for 24 h, then proliferating bone marrow cells were treated with viral supernatant containing polybrene (hexadimethrine bromide; Sigma-Aldrich) and centrifuged. GFP-positive cells were sorted by flow cytometry.

RNA Extraction and Quantitative RT-PCR
Total RNA from Mo-and Gr-MDSCs was isolated using the RNeasy Micro Kit (Qiagen, Hilden, Germany) and total RNA from B16F10 cells was isolated using TRI Reagent (Molecular Research Center, Cincinnati, OH, USA). Using the extracted RNA, cDNA was synthesized with TOPscript RT DryMIX (enzynomics, Daejeon, Korea), and relative mRNA Biomedicines 2021, 9, 896 3 of 14 levels were determined by quantitative PCR using TOPreal qPCR 2X PreMix (enzynomics) and the CFX Connect Real-Time System (Bio-Rad, Hercules, CA, USA). The following primer pairs were used: GAPDH (Forward, 5 -GCC TTC TCC ATG GTG GTG AA; Reverse,  5 -GCA CAG TCA AGG CCG AGA AT) The relative mRNA expression was calculated by the ∆∆C T method, with normalization to the expression of GAPDH.

Enzyme-Linked Immunosorbent Assay (ELISA)
After sorting of MDSCs, 1.25 × 10 5 cells were seeded in 96-well plates for 24 h. Supernatant was harvested and used for measurement of MFGE8 by ELISA. The Mouse Milk Fat Globule 1 ELISA kit (Abcam, Cambridge, UK) was used according to the manufacturer's protocol. The TGF-β ELISA was conducted using the Human/Mouse TGF β1 Uncoated ELISA kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol with supernatant from B16F10 cells after 24 h of culture.

Migration and Invasion Assays
For migration assays, 5 × 10 5 B16F10 cells suspended in serum-free DMEM were loaded in the upper chamber of a Falcon Permeable Support for 24-well Plate with 8.0 µm Transparent PET Membrane. The lower part of the chamber was filled with DMEM containing 10% FBS and MFGE8 (10 ng/mL). After 24 h, non-migrating cells were removed with a cotton swab and migrating cells were fixed with 70% ethanol and stained with 0.5% crystal violet solution (BD Bioscience).
For invasion assays, transwell inserts were coated with 50 µL Matrigel Matrix (Corning, NY, USA) mixed with DMEM (1:4 ratio). 5 × 10 5 B16F10 cells were loaded in the upper chamber and MFGE8 (10 ng/mL) was added to the lower well. After 48 h, non-invading cells were removed with a cotton swab and invading cells were fixed with 70% ethanol and stained with 0.5% crystal violet solution.

Microarray and Data Analysis
Mo-MDSCs and Gr-MDSCs were treated with TGF-β during in vitro differentiation. After sorting, Mo-and Gr-MDSCs were lysed with TRI Reagent and extracted mRNA was analyzed on a GeneChip Mouse Gene 2.0 ST array (Affymetrix, Santa Clara, CA, USA). Three independent microarray experiments were performed. The resulting data was analyzed by Gene Set Enrichment Analysis (GSEA; https://www.gsea-msigdb.org/gsea/ index.jsp, accessed on 9 March 2020) [34] and heat maps were generated using Multiple Experiment Viewer (MeV).
For the gene expression analysis of Mfge8 in tumor-infiltrated MDSCs, the Gene Expression Omnibus (GEO) datasets GSE110772 and GSE21927 were retrieved and analyzed.

Statistical Analysis
The bar graphs show the means of 2-3 independent experiments and the error bars represent the standard deviation (SD). p-values were calculated using two-tailed Student's t-tests, and p-values of <0.05 were considered statistically significant.

TGF-β Is an Important Regulator of MDSCs Immunosuppressive Function
Previously, we have shown that TGF-β is an important factor for the regulation of MDSCs' immune suppressive function [7]. Therefore, in this study, to identify the differentially expressed genes related to the functional enhancement of MDSCs by TGF-β, MDSCs were treated with TGF-β and gene expression changes were analyzed by microarray. In this analysis, we confirmed that treatment with TGF-β increased the Mo-MDSC population and decreased the Gr-MDSC population, as shown in our previous report ( Figure 1A). In addition, the MDSC effector molecules arginase1, iNOS, and IL-10 were up-regulated at the mRNA level by TGF-β treatment of MDSCs ( Figure 1B,C). We also analyzed differential gene expression by GSEA and confirmed that TGF-β signaling pathway-related molecules were significantly enriched in both Mo-MDSCs and Gr-MDSCs treated with TGFβ ( Figure 1D,E), indicating that the immunosuppressive function of MDSCs is increased by TGF-β.

TGF-β Induces Gene Expression Changes including Pro-Metastatic Genes Such as Mfge8
To analyze the TGF-β-induced differential gene expression in MDSCs, we created a two-dimensional plot of the microarray data ( Figure 2A). The X-axis represents the Fold Change ('Fold'), a sum of the gene expression changes in TGF-β-treated Mo-MDSCs and Gr-MDSCs on a log2 scale. The Y-axis represents the Influence Factor ('IF'), the product of the 'Fold' and the difference between the fold changes in Mo-MDSCs and Gr-MDSCs on a log2 scale. Thus, the 'IF' represents whether TGF-β affected the expression of a given gene more strongly in Mo-MDSCs or Gr-MDSCs. 'Distance' is the distance from the origin of the plot; thus the 'Distance' is dependent on both the 'Fold' and the 'IF'. From the microarray data analysis, 227 genes with a 'Fold' value greater than 2.5 were selected and displayed on a heat map ( Figure 2B). The two-dimensional plot of the microarray data showed the differential gene expression profiles of Mo-MDSCs and Gr-MDSCs after TGF-β treatment ( Figure 2C). The 227 genes were divided into six groups according to the 'IF' value, representing the differential impact of TGF-β on each subset of MDSCs. Genes with an 'IF' value greater than 1 (the impact of TGF-β on the expression of the gene was greater in Mo-MDSCs) were assigned to the 'Mo-Up' or 'Mo-Down' group. Genes with an 'IF' value lower than −1 (the impact of TGF-β on the expression of the gene was greater in Gr-MDSCs) were assigned to the 'Gr-Up' or 'Gr-Down' group. Genes with an 'IF' value greater than −1 and lower than 1 were assigned to the 'Up' or 'Down' group. Finally, the three genes with the largest 'Distance' in six groups were selected, and the TGF-β-induced differential mRNA expression was confirmed by qPCR ( Figure 2D).

TGF-β Induces Gene Expression Changes including Pro-Metastatic Genes Such as Mfge8
To analyze the TGF-β-induced differential gene expression in MDSCs, we created a two-dimensional plot of the microarray data ( Figure 2A). The X-axis represents the Fold Change ('Fold'), a sum of the gene expression changes in TGF-β-treated Mo-MDSCs and Gr-MDSCs on a log2 scale. The Y-axis represents the Influence Factor ('IF'), the product of the 'Fold' and the difference between the fold changes in Mo-MDSCs and Gr-MDSCs on a log2 scale. Thus, the 'IF' represents whether TGF-β affected the expression of a given gene more strongly in Mo-MDSCs or Gr-MDSCs. 'Distance' is the distance from the origin of the plot; thus the 'Distance' is dependent on both the 'Fold' and the 'IF'. From the microarray data analysis, 227 genes with a 'Fold' value greater than 2.5 were selected and displayed on a heat map ( Figure 2B). The two-dimensional plot of the microarray data showed the differential gene expression profiles of Mo-MDSCs and Gr-MDSCs after TGFβ treatment ( Figure 2C). The 227 genes were divided into six groups according to the 'IF' value, representing the differential impact of TGF-β on each subset of MDSCs. Genes with an 'IF' value greater than 1 (the impact of TGF-β on the expression of the gene was greater in Mo-MDSCs) were assigned to the 'Mo-Up' or 'Mo-Down' group. Genes with an 'IF' value lower than −1 (the impact of TGF-β on the expression of the gene was greater in Gr- TGF-β is a cytokine with various functions, including immune suppression [6,7] and tumor progression [23,24,35]. Therefore, we hypothesized that genes with differential expression by TGF-β treatment may have similar function with TGF-β and the 18 genes that were selected from the differential expression analysis described above were analyzed for their roles in inflammation or tumor progression (Table 1). Among the 18 genes, six genes (Nt5e, Elovl3, Mfge8, Tshr, Gbp5, and Havcr2) have been previously described to have roles in cancer progression or immune regulation. Mfge8 was selected as a target gene of interest because Mfge8 had the largest 'Distance' value when recalculated with the qPCR results (Table A1), and because Mfge8 has an anti-inflammatory function [32,36] as well as a role in promoting tumor metastasis [28,30]. Table 1. Known functions of the 18 genes selected from the microarray data. The top three genes in each of the six groups were selected for further analysis. The function of some genes has not been well-studied, and these genes are denoted as 'unknown'.

Group Gene Name Function
Up

Elovl3
Fatty acid chain elongation [39] Mfge8 Anti-inflammatory [40,41], Tumor metastasis [27,30] Down Gbp2 * Better prognosis in breast cancer [42], Glioblastoma invasion [ MDSCs) were assigned to the 'Gr-Up' or 'Gr-Down' group. Genes with an 'IF' value greater than −1 and lower than 1 were assigned to the 'Up' or 'Down' group. Finally, the three genes with the largest 'Distance' in six groups were selected, and the TGF-β-induced differential mRNA expression was confirmed by qPCR ( Figure 2D). TGF-β is a cytokine with various functions, including immune suppression [6,7] and tumor progression [23,24,35]. Therefore, we hypothesized that genes with differential expression by TGF-β treatment may have similar function with TGF-β and the 18 genes that were selected from the differential expression analysis described above were analyzed for their roles in inflammation or tumor progression (Table 1). Among the 18 genes, six genes (Nt5e, Elovl3, Mfge8, Tshr, Gbp5, and Havcr2) have been previously described to have roles in cancer progression or immune regulation. Mfge8 was selected as a target gene of interest because Mfge8 had the largest 'Distance' value when recalculated with the qPCR results (Table A1), and because Mfge8 has an anti-inflammatory function [32,36] as well as a role in promoting tumor metastasis [28,30].

MFGE8 Does Not Promote the Immunosuppressive Function of MDSCs
As there are two isoforms of MFGE8, the short isoform (S-MFGE8) and the long isoform (L-MFGE8), we confirmed which isoform is expressed in MDSCs by qPCR before investigating the role of MFGE8 in MDSCs ( Figure 3A). S-MFGE8 mRNA was preferentially expressed over L-MFGE8 in both Mo-MDSCs and Gr-MDSCs. Therefore, the following experiments were performed with S-MFGE8.

MFGE8 Induces Metastasis of B16F10 Melanoma
Although MFGE8 did not increase the immunosuppressive function of MDSCs, it has been previously reported that MFGE8 induces tumor metastasis in various models [27,30]. Therefore, we hypothesized that MFGE8, as an effector molecule of MDSCs, may promote cancer metastasis in TGF-β-rich environments such as the TME. To confirm MFGE8 secretion by MDSCs, we measured the MFGE8 concentration by ELISA ( Figure 4A). Both Mo-MDSCs and Gr-MDSCs secreted MFGE8, and TGF-β significantly increased its secretion. Open-source microarray data showed that tumor-infiltrated MDSCs also express higher * p < 0.05, ** p < 0.01, and *** p < 0.001 by two-tailed t-test.
Next, we overexpressed S-MFGE8 in Mo-MDSCs and Gr-MDSCs using retroviral infection to find out whether MFGE8 promotes the immunosuppressive function of MDSCs. MFGE8 mRNA expression was increased in Mo-MDSCs and Gr-MDSCs upon infection with the MFGE8-expressing retrovirus compared with cells infected with the vector control retrovirus ( Figure 3B). Next, the expression of MDSC immunosuppressive effector molecules was measured. Surprisingly, the arginase1 mRNA level decreased ( Figure 3C) upon MFGE8 overexpression, and the levels of other effector molecules, including iNOS, IL-10, and TGF-β, were unaffected ( Figure 3D). Although MFGE8 is known to have antiinflammatory properties, it did not promote the expression of immunosuppressive effector molecules in MDSCs. MFGE8 treatment showed similar results to MFGE8 overexpression ( Figure 3E,F). These results indicate that, in contrast to our expectations, MFGE8 did not promote the immunosuppressive function of MDSCs.

MFGE8 Induces Metastasis of B16F10 Melanoma
Although MFGE8 did not increase the immunosuppressive function of MDSCs, it has been previously reported that MFGE8 induces tumor metastasis in various models [27,30]. Therefore, we hypothesized that MFGE8, as an effector molecule of MDSCs, may promote cancer metastasis in TGF-β-rich environments such as the TME. To confirm MFGE8 secretion by MDSCs, we measured the MFGE8 concentration by ELISA ( Figure 4A). Both Mo-MDSCs and Gr-MDSCs secreted MFGE8, and TGF-β significantly increased its secretion. Open-source microarray data showed that tumor-infiltrated MDSCs also express higher levels of MFGE8 mRNA than splenic MDSCs ( Figure A1). Furthermore, to confirm B16F10 cells as a source of TGF-β, the concentration of TGF-β was measured by ELISA ( Figure 4B). Since B16F10 cells can also secrete MFGE8, the concentration of MFGE8 protein was also measured in B16F10 supernatants by ELISA and isoform expression levels of MFGE8 in B16F10 were examined by qPCR ( Figures 4C and A2). B16F10 melanoma secreted MFGE8 (1.65 ± 0.04 ng/mL), which is similar amount of MFGE8 secreted by Mo-MDSCs and Gr-MDSCs without TGF-β (1.06 ± 0.44 ng/mL and 1.91 ± 1.23 ng/mL, respectively) where S-MFGE8 is preferentially expressed over L-MFGE8. However, both Mo-MDSCs and Gr-MDSCs secreted substantially higher amounts of MFGE8 in response to TGF-β treatment (3.86 ± 0.95 ng/mL and 7.10 ± 1.70 ng/mL, respectively). Therefore, these ELISA results indicate that MDSCs are the main source of MFGE8 in the TME.
Biomedicines 2021, 9, x FOR PEER REVIEW 9 of 15 levels of MFGE8 mRNA than splenic MDSCs ( Figure A1). Furthermore, to confirm B16F10 cells as a source of TGF-β, the concentration of TGF-β was measured by ELISA ( Figure  4B). Since B16F10 cells can also secrete MFGE8, the concentration of MFGE8 protein was also measured in B16F10 supernatants by ELISA and isoform expression levels of MFGE8 in B16F10 were examined by qPCR ( Figure 4C and Figure A2). B16F10 melanoma secreted MFGE8 (1.65 ± 0.04 ng/mL), which is similar amount of MFGE8 secreted by Mo-MDSCs and Gr-MDSCs without TGF-β (1.06 ± 0.44 ng/mL and 1.91 ± 1.23 ng/mL, respectively) where S-MFGE8 is preferentially expressed over L-MFGE8. However, both Mo-MDSCs and Gr-MDSCs secreted substantially higher amounts of MFGE8 in response to TGF-β treatment (3.86 ± 0.95 ng/mL and 7.10 ± 1.70 ng/mL, respectively). Therefore, these ELISA results indicate that MDSCs are the main source of MFGE8 in the TME.
After confirming MDSCs and B16F10 as a source of MFGE8 and TGF-β, respectively, we evaluated B16F10 proliferation, EMT marker expression, and matrix metalloproteinase (MMP) expression after MFGE8 treatment. MMPs are enzymes that play an important role in cancer cell invasion. For tumor cells to become invasive, MMPs are required to degrade the basal membrane [53]. Although the proliferation of B16F10 cells was not affected by MFGE8 treatment (Figure 4D), MFGE8 induced EMT of B16F10 melanoma cells, as the expression of E-cadherin mRNA decreased while the expression of vimentin and twist1 mRNA increased. Moreover, the mRNA expression of MMP13 was significantly increased by MFGE8 treatment ( Figure 4E). Taken together, these data show that MFGE8 induces metastasis of B16F10 melanoma.  After confirming MDSCs and B16F10 as a source of MFGE8 and TGF-β, respectively, we evaluated B16F10 proliferation, EMT marker expression, and matrix metalloproteinase (MMP) expression after MFGE8 treatment. MMPs are enzymes that play an important role in cancer cell invasion. For tumor cells to become invasive, MMPs are required to degrade the basal membrane [53]. Although the proliferation of B16F10 cells was not affected by MFGE8 treatment (Figure 4D), MFGE8 induced EMT of B16F10 melanoma cells, as the expression of E-cadherin mRNA decreased while the expression of vimentin and twist1 mRNA increased. Moreover, the mRNA expression of MMP13 was significantly increased by MFGE8 treatment ( Figure 4E). Taken together, these data show that MFGE8 induces metastasis of B16F10 melanoma.

MDSCs Increase B16F10 Migration in an MFGE8-Dependent Manner
To further evaluate the relationship between MDSCs and B16F10 melanoma, we conducted a co-culture experiment of MDSCs and B16F10 cells ( Figure 5A). Co-culture of B16F10 and MDSCs significantly increased B16F10 transmigration ( Figure 5B). Furthermore, MFGE8 neutralization significantly reduced B16F10 transmigration ( Figure 5C), while MFGE8 treatment increased B16F10 transmigration and invasion ( Figure 5D,E). Since MFGE8 treatment did not affect B16F10 proliferation ( Figure 4D), the increased cell numbers are a result of augmented B16F10 migration. These results suggest that MDSCs can induce B16F10 metastasis by secreting MFGE8 in response to TGF-β. These results highlight an interaction between MDSCs and B16F10 melanoma cells and demonstrate that MFGE8 is a key molecule that promotes B16F10 melanoma migration.

MDSCs Increase B16F10 Migration in an MFGE8-Dependent Manner
To further evaluate the relationship between MDSCs and B16F10 melanoma, we conducted a co-culture experiment of MDSCs and B16F10 cells ( Figure 5A). Co-culture of B16F10 and MDSCs significantly increased B16F10 transmigration ( Figure 5B). Furthermore, MFGE8 neutralization significantly reduced B16F10 transmigration ( Figure 5C), while MFGE8 treatment increased B16F10 transmigration and invasion ( Figure 5D,E). Since MFGE8 treatment did not affect B16F10 proliferation ( Figure 4D), the increased cell numbers are a result of augmented B16F10 migration. These results suggest that MDSCs can induce B16F10 metastasis by secreting MFGE8 in response to TGF-β. These results highlight an interaction between MDSCs and B16F10 melanoma cells and demonstrate that MFGE8 is a key molecule that promotes B16F10 melanoma migration.

Discussion
MDSCs accumulate and interact with tumor cells in the TME. It is widely accepted that MDSCs accumulation results in tumor progression and metastasis [54], however the detailed mechanisms are still not fully understood. In this study, we showed that MDSCs interact with B16F10 melanoma, increasing cancer cell migration. TGF-β treatment of MDSCs resulted in increased MFGE8 secretion and addition of MFGE8 neutralizing antibody demonstrated that MDSC-derived MFGE8 promoted the migration of B16F10 melanoma cells. Previous studies have shown that MFGE8 promotes tumor progression or metastasis in various cancer models [27,29,30], yet the reported source of MFGE8 was mainly TAMs or stromal cells. Therefore, our findings suggest a new mechanism through which MDSCs promote tumor progression, as a source of MFGE8. In addition, although MDSCs preferentially express S-MFGE8, comparing S-MFGE8 function with L-MFGE8 in the tumor microenvironment could be an important issue in future studies.
In a previous study, we showed that TGF-β increased the immunosuppressive function of MDSCs in a murine colitis model [7]. However, the function of TGF-β is not restricted to prolonged inflammatory response alleviation. Many studies have shown the roles of TGF-β in tumor metastasis [24,25] and immune modulation. Here, we demonstrate an additional role of TGF-β as an immune modulator in the TME. TGF-β increased the Mo-MDSC population, which directly interact with tumor cells and induce EMT in the TME [19]. The immunosuppressive function of MDSCs was assessed based on their mRNA expression of effector molecules. Arginase1, iNOS, and IL-10 mRNA levels increased with TGF-β treatment. Most importantly, TGF-β altered the gene expression profile of MDSCs, upregulating pro-metastatic genes such as Mfge8. Although MFGE8 did not promote MDSCs immunosuppressive function itself, it affected the phenotype of B16F10 melanoma cells, suggesting a role of TGF-β and TGF-β-induced genes in tumor metastasis. Furthermore, MFGE8 produced by MDSCs may also be involved in the inhibition of T cell proliferation in the tumor microenvironment, as a previous report showed inhibition of T cell proliferation by MFGE8 [55].
Generally, cancer metastasis is related to the prognosis and survival rate. Often, melanoma can be detected easily, but it is highly metastatic, thus increasing the mortality rate [56,57]. In this study, we showed that MDSCs induce the migration of B16F10 mouse melanoma cells by secreting MFGE8, and MFGE8 neutralization suppressed their migration. A previous study also suggested MFGE8 antibody as a novel therapeutic for ovarian carcinoma [27]. Together, these results suggest the therapeutic potential of targeting MDSCs or MFGE8 for metastatic melanoma or other cancers.

Conclusions
We demonstrate that TGF-β alters the gene expression profile of MDSCs. Mfge8 and a number of other metastatic genes were upregulated in TGF-β-treated MDSCs. MFGE8 did not directly increase MDSC effector molecule expression, but MDSC-derived MFGE8 promoted the metastasis of B16F10 melanoma cells, suggesting that MDSC-derived MFGE8 could be a promising target for metastatic cancer therapies.  Institutional Review Board Statement: The study was approved by the Institutional Animal Care and Use Committee of Gwangju Institute of Science and Technology (GIST-2020-086, approved on 26 August 2020).

Informed Consent Statement: Not applicable.
Data Availability Statement: The datasets are available from the corresponding authors upon reasonable request.

Conflicts of Interest:
The authors declare no potential conflict of interest.  Figure A1. Mfge8 expression is increased in tumor-infiltrated MDSCs. Mfge8 express infiltrated MDSCs in the GEO datasets GSE110772 and GSE21927. Data are expressed ± SD. * p < 0.05 and ** p < 0.01 by two-tailed t-test. Figure A1. Mfge8 expression is increased in tumor-infiltrated MDSCs. Mfge8 expression in tumorinfiltrated MDSCs in the GEO datasets GSE110772 and GSE21927. Data are expressed as the mean ± SD. * p < 0.05 and ** p < 0.01 by two-tailed t-test. Figure A1. Mfge8 expression is increased in tumor-infiltrated MDSCs. Mfge8 expressi infiltrated MDSCs in the GEO datasets GSE110772 and GSE21927. Data are expressed ± SD. * p < 0.05 and ** p < 0.01 by two-tailed t-test.