Interleukin-17 enhances immunosuppression by mesenchymal stem cells

IL-17 enhances the immunosuppressive effect of MSCs

Our previous studies have shown that a combination of IFNγ with TNFα is required to elicit the immunosuppressive effect of MSCs. We examined whether IL-17 could affect IFNγ and TNFα-induced immunosuppression of MSCs. We found that T-cell proliferation was reduced in the presence of IFNγ and TNFα, and surprisingly, this suppression was markedly enhanced by IL-17 (Figure 1a). As IL-17 signals through IL-17RA and IL-17RC, the expression of these receptors in MSCs was determined by quantitative RT-PCR (Figure 1b), in which a mouse macrophage cell line Raw 264.7 was used as a positive control. Cell surface expression of the receptor was also confirmed by flow cytometric analysis (Figure 1c).

As the concentrations of inflammatory cytokines vary at different stages of an inflammatory response, we examined the concentration dependence of IFNγ and TNFα on IL-17–enhanced immunosuppression. IL-17 was able to enhance the immunosuppressive effect of MSCs on T cells at IFNγ and TNFα concentrations as low as 1–2 ng/ml each (Figures 1d and e). Even at higher concentrations of IFNγ and TNFα (10–20 ng/ml), IL-17 still improved immunosuppression by MSCs though the effect was less pronounced. Nonetheless, in the presence of MSCs, as little as 0.5 ng/ml IL-17 was sufficient to elicit a dramatic decrease in T-cell proliferation (Figure 1f). We also examined the role of T-cell death in this effect of IL-17. However, our results showed that the strongly inhibited T-cell proliferation by MSCs pretreated with IFNγ+TNFα+IL-17 was not due to T-cell death (Supplementary Figure S1).

To verify that IL-17 contributes to the immunosuppressive effect of MSCs during a T-cell immune response, antibody against IL-17 was used to neutralize it in a co-culture system with MSC-activated splenocytes. While proliferation of activated splenocytes was markedly inhibited by MSCs, this inhibition was partially reversed upon addition of antibody against IL-17 (Figure 1g). The optimal effect of anti-IL-17 occurred with a MSC/splenocyte ratio of 1 : 40. At a MSC/splenocyte ratio of 1 : 20, the reversal effect was less pronounced, but still statistically significant. Furthermore, this enhancing effect of IL-17 on immunosuppression of MSCs was completely reversed by iNOS inhibitor, L-NMMA (Figure 1h), suggesting that iNOS has a crucial role in this action of IL-17.

To expand our studies, we compared the function of IL-17 between bone marrow–derived MSCs and adipose-derived MSCs. IL-17 could also dramatically enhance IFNγ and TNFα-induced immunosuppression in adipose-derived MSCs (Supplementary Figure S2A), indicating that this effect of IL-17 may be independent of the source of MSCs. As the self-renewal capacity and plasticity may be altered after several passages, we examined whether the responsiveness of MSCs to IL-17 treatment was compromised as passage number increased. IL-17 still enhanced the immunosuppressive effect of MSCs at higher passage number, even though IFNγ and TNFα can induce modestly more immunosupression in this situation (Supplementary Figure S2B).

IL-17 synergizes with inflammatory cytokines to induce the expression of immune modulatory genes in MSCs

We have shown that IL-17 can enhance the immunosuppressive effect of MSCs. To reveal the mechanism, we initially hypothesized that IL-17 affects the proliferation of MSCs, especially considering that IL-17 has been reported to promote the proliferation of human MSCs.²⁶ However, we found that in the presence of IFNγ and TNFα, the addition of IL-17 modestly inhibited the proliferation of MSCs (Supplementary Figure S2C). Thus, the promoting effect of IL-17 on immunosuppression by MSCs is not through influencing the proliferation of MSCs.

We previously showed that NO and chemokines, acting in concert, are the key molecules mediating the immunosuppressive effects of MSCs. The induction of iNOS and some chemokine genes in MSCs is initiated by IFNγ and TNFα.²⁷ As IL-17 enhanced immunosuppression by MSCs cultured with IFNγ and TNFα (Figures 1d and e), and iNOS was essential for this effect of IL-17 (Figure 1h), we hypothesized that IL-17 synergizes with IFNγ and TNFα to induce the expression of iNOS or chemokines in MSCs. To test this hypothesis, MSCs were cultured with various combinations of IFNγ, TNFα, and/or IL-17, and the expression levels of some key immune modulatory genes were assessed. Compared with MSCs cultured with single or double combinations of cytokines, addition of IFNγ, TNFα, and IL-17 dramatically increased the expression of iNOS mRNA (Figure 2a). Western blotting analysis and assay of nitrate in supernatant also confirmed an increase of iNOS protein (Figure 2b, Supplementary Figure S2D). However, the expression of other chemokines such as CCL2, CCL5, CXCL9, and CXCL10, which have pivotal roles in the immunosuppressive effects of MSCs, were all unaffected by addition of IL-17 (Figure 2c), indicating a specific effect of IL-17 on iNOS expression. The addition of IL-17 can induce MSCs to produce large amounts of NO, which can further inhibit the proliferation of T cells. Excessive NO can also exhibit a cytotoxic effect on MSCs, resulting in suppressed proliferation of IFNγ+TNFα+IL-17 treated MSCs (Supplementary Figure S2C).

To verify that the effects on gene expression were due to IL-17, MSCs were cultured with supernatant from anti-CD3 and anti-CD28–activated splenocytes in the presence or absence of neutralizing antibody against IL-17. Addition of anti-IL-17 to supernatant diminished its ability to induce iNOS gene expression without affecting CCL2, CCL5, CXCL9, or CXCL10 (Figures 2d and e). Furthermore, we also verified our findings by blocking signal transduction of IL-17 in MSCs by knocking down Act1, a key component in the signal transduction pathway from the IL-17 receptor. As has been shown in other cells,¹⁰ the phosphorylation of IκBα, ERK, p65, and JNK, key molecules in the IL-17 signaling pathway, were impaired in these Act1 knockdown MSCs (Supplementary Figure S3A). As expected, IL-17 cannot upregulate IFNγ and TNFα-induced iNOS expression in Act1 knockdown MSCs (Supplementary Figures S3B, C). At the cellular level, the immunosuppressive effect of IFNγ and TNFα could not be further enhanced by IL-17 in the absence of Act1 (Supplementary Figure S3D).

IL-17 together with IFNγ and TNFα-pretreated MSCs alleviate ConA-induced liver injury in an iNOS-dependent manner

To investigate the promoting effect of IL-17 on IFNγ and TNFα-induced immunosuppression of MSCs in vivo, we employed the ConA-induced liver injury mouse model. ConA-induced liver injury is a well-established in vivo model of hepatocyte apoptosis induced by acute immune responses, in which T lymphocytes are identified as the major effector cells.²⁸ Wild-type MSCs were pretreated or not with IFNγ and TNFα in the presence or absence of IL-17 for 12 h. After ConA administration for 30 min, these differently pretreated MSCs were intravenously injected into the mice. The immunosuppression mediated by MSCs is exerted locally. Thus, GFP-transgenic MSCs were transfused into mice that received ConA injection 30 min earlier. Localization of administrated MSCs in the damaged liver was indeed observed (Supplementary Figure S4). Interestingly, liver injury was not inhibited in mice administrated with wild-type MSCs or wild-type MSCs pretreated with IFNγ and TNFα. However, MSCs pretreated with IFNγ, TNFα, and IL-17 almost completely inhibited liver damage, exhibited as a dramatic reduction of serum alanine aminotransferase (ALT) level, centrilobular necrosis, and inflammatory infiltration (Figures 3a, c and d). Detailed analyses of immune cells in the liver showed that the infiltration of mononuclear cells (MNCs), CD3⁺CD4⁺, and CD3⁺CD8⁺ T cells were dramatically decreased in mice administered with wild-type MSCs pretreated with IFNγ and TNFα, together with IL-17 (Figure 3b). However, the ameliorated effects of MSCs pretreated with IFNγ, TNFα, and IL-17 in liver injury was not observed in iNOS^−/− MSCs (Figures 3a, c and d), indicating that iNOS is a key mediator in the promoting effect of IL-17 on immunosuppression by MSCs in vivo.

IL-17 reverses the suppression of gene expression imposed by RNA-binding protein AUF1

Messenger RNAs encoding iNOS and many cytokines/chemokines are rapidly degraded after interaction with mRNA decay factors, which provides a mechanism to limit the abundance of these molecules during an immune response. Activation of signaling pathways, particularly during immune responses, stabilizes many of these mRNAs to increase their expression. Indeed, a major mechanism by which IL-17 influences the expression of many inflammatory mediator genes is by stabilizing their mRNAs.^{29, 30} AUF1 has been shown to promote the degradation of iNOS mRNA.²⁰ We thus hypothesized that AUF1 may act to limit the expression of iNOS mRNA and that IL-17 may block this activity of AUF1 (thereby increasing gene expression).

To test this hypothesis, cytokine-induced gene expression was compared between MSCs derived from bone marrow of wild-type and auf1^−/− mice.³¹ Cells were cultured as before with IFNγ and TNFα, with or without IL-17. Addition of IL-17 (together with IFNγ and TNFα) dramatically further increased iNOS mRNA in wild-type MSCs (Figure 4a). In contrast, iNOS mRNA was much higher in auf1^−/− MSCs cultured with IFNγ and TNFα; interestingly, addition of IL-17 had little effect (i.e., IL-17 increased iNOS mRNA abundance less than twofold). Likewise, IFNγ and TNFα were sufficient to maximally induce iNOS protein in auf1^−/− MSCs without the need for IL-17, in contrast to wild-type MSCs (Figure 4b, compare lanes 7 and 8 with lanes 2 and 5). To verify these findings, we knocked down AUF1 in wild-type MSCs (Figure 4c), and treated these cells with IFNγ and TNFα, with or without IL-17. Consistent with auf1^−/− MSCs, IFNγ and TNFα induced maximal iNOS mRNA and protein expression in AUF1-knockdown MSCs (Figures 4d and e), and the enhancing effect of IL-17 was abated. Moreover, IL-17 further increased the suppressive activity of IFNγ and TNFα-stimulated wild-type MSCs on T-cell proliferation; however, IFNγ and TNFα were sufficient to induce maximal immunosuppression in auf1^−/− MSCs and IL-17 did not further enhance the inhibitory effect on T-cell proliferation (Figure 4f).

IL-17 enhances gene expression through modulating AUF1-mediated mRNA decay in MSCs

Given the effects of AUF1 and IL-17 on gene expression and immunosupression by MSCs, we considered the possibility that knockout of AUF1 alone would be sufficient to provide the degree of mRNA stabilization that would normally require IL-17. mRNA decay assays were performed with wild-type and auf1^−/− MSCs, cultured with IFNγ and TNFα, with or without IL-17. In wild-type MSCs, iNOS mRNA was relatively unstable with a half-life of 4.3±1.4 h; IL-17 led to a twofold stabilization of iNOS mRNA (Figure 5a); P<0.05. The CCL2 and CXCL10 mRNAs, which did not respond to IL-17 (see Figure 2c), were not stabilized by IL-17, as would be expected (Figure 5a; t_1/2=∼2 h for both mRNAs, with or without IL-17). In contrast to wild-type MSCs, knockout of AUF1 strongly stabilized iNOS mRNA (Figure 5b; t_1/2>10 h). In these auf1^−/− MSCs, IL-17 had no effect on the half-life of iNOS mRNA (Figure 5b). We also verified these findings in AUF1-knockdown MSCs (Figure 5c).

To assess the interaction of IL-17 and AUF1 on the enhancing effect of IL-17 on MSC immunosuppression, the protein level of AUF1 was analyzed in MSCs treated with IFNγ and TNFα, in the presence or absence of IL-17, as a function of time. We found that the addition of IL-17 could reduce AUF1 protein levels in IFNγ+TNFα-treated MSCs (Figure 5d). Therefore, IL-17 enhances iNOS mRNA stability by reducing AUF1 protein levels. Thus, AUF1 is one key factor mediating the enhancing effect of IL-17 on IFNγ and TNFα-induced iNOS expression by MSCs. Further experiments were employed to assess potential interactions between AUF1 and IL-17 downstream molecules. As either knockdown of Act1 or knockout of AUF1 in MSCs abolished the ability of IL-17 to enhance their immunosuppressive effects (Supplementary Figure S3 and Figure 4), we assessed interaction of Act1 and AUF1 in MSCs upon IL-17 stimulation using co-immunoprecipitation. IL-17 stimulation increased the interaction of endogenous Act1 with AUF1 (Figure 5e). These data support the ideas that AUF1 downregulation and Act1–AUF1 interactions are key steps in the enhancing effect of IL-17 on both gene expression and immunosupression in MSCs.

Critical roles of AUF1 in IL-17–enhanced therapeutic effect of MSCs on ConA-induced acute liver injury

Given the crucial role of AUF1 in the enhancing effect of IL-17 on immunosuppression by MSCs in vitro, we next examined the effects of IL-17 on immunosuppression in vivo by wild-type and auf1^−/− MSCs. Wild-type and auf1^−/− MSCs were first pretreated with or without IFNγ and TNFα, in the presence or absence of IL-17 for 12 h, and then these cells were transfused into mice that received ConA injection 30 min earlier. Consistent with the results in Figure 3, wild-type MSCs pretreated with IFNγ, TNFα, and IL-17 showed the best therapeutic effect on liver injury, in which liver injury was almost completely alleviated with sharply reduced serum ALT activity, liver necrosis, and lymphocyte infiltration (Figures 6a, c and d). However, maximal therapeutic effects on ConA-induced liver injury were also achieved with auf1^−/− MSCs pretreated with only IFNγ and TNFα, without the need for IL-17 (Figures 6a, c and d). Infiltration of MNCs and CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells in liver was also dramatically decreased in mice administered with wild-type MSCs pretreated with IFNγ, TNFα, and IL-17, or in mice administered auf1^−/− MSCs pretreated with IFNγ and TNFα, either with or without IL-17 (Figure 6b). The percentages of CD3⁺CD4⁺ T cells, CD3⁺CD8⁺ T cells, Th1, Th17, and Treg subpopulations were analyzed in liver-infiltrated cells among the different groups. However, no significant changes were observed among them (Supplementary Figures S5A, B and C). Thus, the therapeutic effect of MSCs pretreated with IFNγ, TNFα, and IL-17 on ConA-induced liver injury was mainly dependent on their robust suppression of T-cell proliferation, but not by affecting differentiation of T-cell subpopulations.

Reagents and mice

Recombinant mouse IFNγ, TNFα, IL-17A, and antibodies against IL-17A were from eBiosciences (La Jolla, CA, USA). Recombinant mouse IL-2 was from R&D Systems (Minneapolis, MN, USA). Antibodies against β-actin, GAPDH, iNOS, p65, p-IκBα, p-p65, p-JNK, and p-ERK1/2 were from Cell Signaling Technology (Danvers, MA, USA). Antibody against Act1 was from Santa Cruz Biotechnology (Dallas, TX, USA). L-NMMA, propidium iodide, and actinomycin D were purchased from Sigma-Aldrich (St Louis, MO, USA). Concanavaline A (ConA) was from Vector Labs (Burlingame, CA, USA).

C57BL/6 mice were purchased from the Shanghai Laboratory Animal Center of the Chinese Academy of Sciences (Shanghai, China) and maintained under specific pathogen-free conditions in the vivarium of Shanghai Jiao Tong University School of Medicine with water and food provided ad libitum. All animal protocols are approved by our Institutional Animal Care and Use Committee.

Cells

MSCs were generated using our previously described protocol.¹³ Briefly, tibia and femur bone marrow of wild-type, iNOS^−/−, or auf1^−/− mice was harvested. Cells were cultured in DMEM medium supplemented with 10% FBS, 2 mM glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (complete medium, all from Invitrogen, Carlsbad, CA, USA). All nonadherent cells were removed after 24 h (hr), and adherent cells were maintained. Medium was changed every 2–3 days. To obtain MSC clones, cells at confluence were harvested and seeded into 96-well plates by limited dilution. Individual clones were then picked and expanded. These MSCs were capable of differentiating into adipocytes and osteocytes under the respective differentiation conditions. Cells were used before the 20th passage.

The generation of T-cell blasts was done as previously described.^{13, 43}

Proliferation assay

To assay cell proliferation, 0.5 μCi of ³H-thymidine (Tdr, Shanghai Institute of Applied Physics, Chinese Academy of Sciences, China) was added to each well of 96-well plates 6 h before termination of the cultures by freezing. Plates were then thawed, cells were harvested, and incorporated ³H-Tdr was assessed with a Wallac Microbeta scintillation counter (Perkin-Elmer, Waltham, MA, USA).

DNA content analysis

DNA content analysis was done as previously described.⁴⁴ Briefly, A1.1 cells were co-cultured with MSCs for 12 h, harvested, and washed once with PBS. The cells were then resuspended in staining buffer (PBS supplemented with 50 μg/ml propidium iodide, 0.2% saponin, 10 μg/ml RNase A). After 30 min incubation in dark, samples were analyzed by flow cytometry on a FACS Calibur flow cytometer (Becton Dickinson, San Jose, CA, USA).

RNA isolation and gene expression assay

Total RNA was isolated with the RNAprep Pure Cell/Bacteria Kit (TIANGEN Biotech, Beijing, China). First-strand cDNA synthesis was performed with a cDNA synthesis kit (TIANGEN Biotech). The levels of mRNAs were measured by quantitative RT-PCR (7900 HT; Applied Biosystems, Foster City, CA, USA) with SYBR Green Master Mix (TaKaRa Biotech, Dalian, China) and normalized to the level of β-actin mRNA. Sequences of forward and reverse primer pairs are as follows:

AUF1: forward 5′-CAGCAGAGTGGTTATGGGAAAGTATCC-3′ and reverse

5′-GACAGGAGCCACCTTCAAATGAATC-3′

β-actin: forward 5′-CCACGAGCGGTTCCGATG-3′ and reverse

5′-GCCACAGGATTCCATACCCA-3′

CCL2: forward 5′-TCTCTCTTCCTCCACCACCATG-3′ and reverse

5′-GCGTTAACTGCATCTGGCTGA-3′

CCL5: forward 5′-TTTCTACACCAGCAGCAAGTGC-3′ and reverse

5′-CCTTCGTGTGACAAACACGAC-3′

CXCL9: forward 5′-AGTGTGGAGTTCGAGGAACCCT-3′ and reverse

5′-TGCAGGAGCATCGTGCATT-3′

CXCL10: forward 5′-TAGCTCAGGCTCGTCAGTTCT-3′ and reverse

5′-GATGGTGGTTAAGTTCGTGCT-3′.

IL-6: forward 5′-GAGGATACCACTCCCAACAGACC-3′ and reverse

5′-AAGTGCATCATCGTTGTTCATACA-3′

IL-17RA: forward 5′-AGTGTTTCCTCTACCCAGCAC-3′ and reverse

5′-GAAAACCGCCACCGCTTAC-3′

IL-17RC: forward 5′-GCTGCCTGATGGTGACAATGT-3′ and reverse

5′-TGGACGCAGGTACAGTAAGAAG-3′

iNOS: forward 5′-CAGCTGGGCTGTACAAACCTT-3′ and reverse

5′-CATTGGAAGTGAAGCGTTTCG-3′.

Messenger RNA decay assays

Messenger RNA decay assays were performed essentially as described.^{45, 46, 47} Briefly, MSCs were incubated with cytokine combinations for 6 h. Actinomycin D (Act.D) was added into the medium at a final concentration of 5 μg/ml to stop transcription. At various time points after addition of Act.D, cells were harvested for the extraction of total RNA. Levels of iNOS, IL-6, CCL2, and CXCL10 mRNAs were measured at each time point by quantitative RT-PCR. Percentages of mRNA remaining at each time point were plotted versus time after Act.D addition. First-order decay constants, k, were determined by nonlinear regression. The associated mRNA half-lives, t_1/2, were calculated with the equation t_1/2=ln2/k.

RNA interference

Plasmids bearing shRNA targeting AUF1 mRNA (5′-GTTGTAGACTGCACTCTGA-3′) or Act1 mRNA (5′-GACAGTGATGATAATCGTA-3′) were constructed by the insertion of annealed oligonucleotides into the Hind III and BamH I sites of pSILENCER 2.0-U6 (Ambion, Austin, TX, USA). Plasmids bearing scrambled sequences (shCTRL), shRNA targeting AUF1 mRNA (shAUF1) or Act1 mRNA (shAct1) were introduced into MSCs (1.5 × 10⁶ cells) using Amaxa Nucleofector, with program U-023, and the Nucleofector kit V (Amaxa, Cologne, Germany). Cells were seeded in DMEM complete medium in six-well plates. After 2–3 days, the medium was changed, and stably transfected clones were picked and expanded using DMEM complete medium containing 250 μg/ml hygromycin B (Invitrogen) for 3–4 weeks.

Western blotting analysis

Cells were washed twice with ice-cold PBS, harvested and lysed in the RIPA buffer (Millipore, Temecular, CA, USA) containing a cocktail of protease inhibitors (Roche, Natley, NJ, USA) and PMSF (Sigma) for 30 min on ice. Lysates were clarified by centrifugation at 16 000 × g for 15 min. Protein concentration of the supernatant was determined by the Bradford assay (Bio-Rad, Hercules, CA, USA). Protein samples were diluted in 5 × SDS loading buffer and fractionated in a 10% SDS–polyacrylamide gel. Proteins were electroblotted onto a nitrocellulose membrane (Whatman Inc., Clifton, NJ, USA) and incubated for 1 h in 5% nonfat dry milk dissolved in TBST (150 mM NaCl, 50 mM Tris-HCl, pH 7.5, 0.05% Tween 20) at room temperature. The blotting membranes were incubated with primary antibodies overnight at 4 °C, extensively washed in TBST, incubated with hP-conjugated secondary antibody (Cell Signaling) for 1.5 h at room temperature, and washed again with TBST. The blotting membranes were developed with chemiluminescent reagents (Millipore, Billerica, MA, USA) according to the instructions provide by the manufacturer.

Immunoprecipitation

Cells were lysed in RIPA buffer (Millipore, Temecular, CA, USA) containing a cocktail of protease inhibitors (Roche) and PMSF (Sigma) for 30 min on ice. Lysates were clarified by centrifugation at 16 000 × g for 15 min. Protein concentration of the supernatant was determined by the Bradford assay (Bio-Rad). After precleared with protein G sepharose beads, cell extracts were incubated with 20 μl protein G sepharose beads in the presence of 1 μg Act1 antibody or nonspecific IgG overnight at 4 °C. After washing four times with RIPA buffer, beads were suspended in 20 μl 2 × SDS loading buffer, boiled at 100 °C for 10 min, and analyzed by western blot.

ConA-induced liver injury in mice

C57BL/6 mice (8-10 week old) were intravenously injected with ConA in PBS at 15 mg/kg to induce acute inflammatory liver injury. MSCs (5 × 10⁵) derived from wild-type mice, iNOS^−/− mice or auf1^−/− mice were treated with or without IFNγ and TNFα in the presence or absence of IL-17 (10 ng/ml for each cytokine) for 12 h, and then intravenously administrated into mice that have been treated with ConA for 30 min. Mice were killed and serum and liver tissues were sampled after another 7.5 h. Serum ALT activity was determined by an ALT detection kit (Yihua Medical Scientific, Shanghai, China). Formalin-fixed liver histological sections were stained with hematoxylin & eosin.

Localization of MSCs in mouse liver

Mice were transfused with GFP-transgenic mouse MSCs 30 min after ConA injection. After another 7.5 h, livers were harvested. Frozen sections were stained with anti-GFP antibody (Abcam, New Territories, Hong Kong) and labeled with Alexa Fluor 594 conjugated donkey anti-rabbit secondary antibody (Invitrogen) and DAPI (Sigma-Aldrich) to reveal localization of MSCs.

Isolation of liver MNCs and flow cytometric analysis

Liver mononuclear cells (MNCs) were purified by a 40%/70% percoll gradient⁴⁸ and stained with anti-CD3-PE, anti-CD4-PerCP/Cy5.5, anti-CD8a-APC, or anti-CD25-FITC (eBiosciences) for 30 min at 4 °C in staining buffer (PBS, 3% FCS). For detection of intracellular proteins, cells were first stimulated with PMA (Sigma) and inomycin (Sigma) for 4 h and stopped by addition of Brefeldin A (Sigma). After cell surface staining, cells were fixed and permeabilized (eBiosciences), and then stained respectively with anti-IFNγ-FITC, anti-IL-17 A-PE, or anti-Foxp3-PE. For detection of surface expression of IL-17RA in cloned MSCs, cells were stained with anti-IL-17RA-PE (eBiosciences). All samples were analyzed by flow cytometry on a FACS Calibur flow cytometer (Becton Dickinson).

Statistical analysis

Nonlinear regression and statistical analyses were performed with PRISM v5 software (GraphPad Software, Inc., La Jolla, CA, USA). Comparisons between samples were performed with the unpaired t-test. Differences with P<0.05 were considered significant (*, P<0.05; **, P<0.01; ***, P<0.001); N.S., not significant.