IL-6 differentiates CD8⁺ T cells into IL-21–producing effector cells

The aforementioned data show that high levels of IL-21, but low levels of IFN-γ, are produced by CD8⁺ T cells activated with IL-6. The levels of IL-2 produced by CD8⁺ T cells activated with IL-6 were also lower (Fig. 2 A). IL-6 also reduced the levels of Ifng (Fig. 2 B) and Il2 mRNA (Fig. 2 C), indicating that this effect is not caused by increased consumption. Thus, although IL-6 induces IL-21, it also has a negative effect on the expression of IFN-γ and IL-2 in CD8⁺ T cells.

Open in a separate window

Figure 2.

IL-6 reprograms effector CD8⁺ T cells into IL-21–producing effector cells. (A) IL-2 production from CD8⁺ T cells activated using anti-CD3/CD28 Abs in the absence or presence of IL-6 for 72 h (n = 3). (B and C) Relative mRNA levels for IFN-γ (B) and IL-2 (C) in CD8⁺ T cells activated with or without IL-6 or IL-2. (D) Intracellular staining for IFN-γ and IL-2, and surface staining for hCD4 (surrogate for IL-21) on activated CD8⁺ IL-21-hCD4 reporter T cells. Numbers indicate the percentage of cells in each quadrant. (E and F) Percentage of different cytokine-producing populations as determined in D (n = 3). (G–I) CD8⁺ T cells were activated in the absence (primary, Med) or presence (primary, IL-6) of IL-6 for 2 d, harvested, and restimulated using anti-CD3 Ab in the absence (Restim, Med) or presence (Restim, IL-6) of IL-6. After 24 h, IL-21 production (G), the relative IL-21 mRNA levels (H), and IL-2 production (I) were determined (n = 3). (J and K) CD8⁺ T cells were activated with or without IL-6 that was added at the indicated time after activation. After 72 h of activation (J) and 72 h after the addition of IL-6 (K), IL-21 production was determined (n = 3). Error bars represent the mean ± SD. *, P < 0.05, as determined by Student’s t test (A), one-way ANOVA (B, C, J, and K), or two-way ANOVA (E–I). Results are representative of two to three experiments.

These results suggested that IL-6 could promote the differentiation of CD8⁺ T cells into effector CD8⁺ T cells characterized by high IL-21, low IFN-γ, and low IL-2 production. We therefore performed IL-2 and IFN-γ intracellular staining in CD8⁺ T cells from IL-21-reporter mice expressing a truncated human CD4 (hCD4) in the Il21 locus (unpublished data). No hCD4 was detected in CD8⁺ T cells activated in the absence of IL-6 (Fig. 2 D). However, a subset of hCD4 (IL-21)-expressing CD8⁺ T cells was present when cells were activated with IL-6 (Fig. 2 D). The percentage of IFN-γ– and IL-2–producing cells were markedly reduced in CD8⁺ T cells activated with IL-6 (Fig. 2 D). Importantly, most IL-21–producing cells did not make IL-2 (Fig. 2 E) or IFN-γ (Fig. 2 F). Thus, IL-6 promotes the generation of an effector CD8⁺ T cell subpopulation that produces high levels of IL-21, but minimal levels of IFN-γ and IL-2.

We then investigated whether this IL-21–producing subset of effector CD8⁺ T cells represented a relatively committed subset. CD8⁺ T cells were activated for 2 d with anti-CD3/CD28 antibodies (Abs) in the presence or absence of IL-6 (primary activation), extensively washed, and an equal number of cells was then restimulated with anti-CD3 Ab only or anti-CD3 Ab plus IL-6. IL-21 levels were determined in the supernatant 24 h later. CD8⁺ T cells where primary activation was done in the absence of IL-6 did not produce IL-21 after restimulation even when IL-6 was added during the restimulation phase (Fig. 2 G). In contrast, high levels of IL-21 were induced after restimulation with anti-CD3 Abs alone if effector CD8⁺ T cells were initially activated in the presence of IL-6 (Fig. 2 G). Similar results were obtained when Il21 mRNA levels were assayed (Fig. 2 H), indicating that the absence of IL-21 was not caused by increased consumption. Moreover, CD8⁺ T cells that received IL-6 during the primary activation still produced low levels of IL-2 after restimulation (Fig. 2 I). These data show that the presence of IL-6 during the primary activation promotes the differentiation of CD8⁺ T cells into a unique effector CD8⁺ T cell subset that remains capable of producing IL-21 upon activation.

To better dissect the period of time during activation when CD8⁺ T cells remain responsive to IL-6 and produce IL-21, CD8⁺ T cells were activated with anti-CD3/CD28 Abs, and IL-6 was administered at different periods of time after stimulation. IL-21 levels were measured 3 d after the stimulation. The maximum production of IL-21 was achieved when IL-6 was added at the initiation of stimulation (Fig. 2 J). Delayed addition of IL-6 resulted in progressively decreased production of IL-21 by CD8⁺ T cells (Fig. 2 J). Addition of IL-6 48 h after activation did not induce IL-21 production when examined a day later (Fig. 2 J) or even when cells were incubated for 3 additional days after addition of IL-6 (Fig. 2 K). Thus, the differentiation of CD8⁺ T cells into IL-21–producing effector cells requires IL-6 signaling early during differentiation.

IL-6R defines a CD8⁺ T cell subset capable of producing IL-21

IL-6 induces signals by binding to its membrane IL-6R, which then associates with gp130, a common signal transducer that activates the Jak/Stat3 pathway (Heinrich et al., 2003). To determine the fraction of CD8⁺ T cells that could produce IL-21 by responding to IL-6, we analyzed the cell surface levels of IL-6R by flow cytometry. Only a subset of CD8⁺ T cells had detectable levels of IL-6R (IL-6R^high; Fig. 3 A). The levels of surface expression of IL-6R in CD8⁺ T cells were lower than in CD4⁺ T cells, both in intensity (Fig. 3 B) and frequency of expressing cells (Fig. 3 C). Analysis of IL-6R expression during thymocyte development showed that single-positive CD4⁺ thymocytes already expressed higher levels of IL-6R than single-positive CD8⁺ thymocytes (Fig. 3 D), whereas no IL-6R expression could be found in double-positive thymocytes (Fig. 3 D). Expression of IL-6R on the cell surface has been shown to be down-regulated in CD4⁺ T cells during activation (Rincón et al., 1997). Analysis of IL-6R during activation of CD8⁺ T cells also showed a down-regulation (Fig. 3 E). Down-regulation of IL-6R with activation correlates with the failure of IL-6 to induce Stat3 phosphorylation (unpublished data) and IL-21 production (Fig. 2, J and K) in CD8⁺ T cells when administrated later during activation.

Open in a separate window

Figure 3.

IL-6R defines a CD8⁺ T cell subset capable of producing IL-21. (A) IL-6R expression in fresh CD4⁺ and CD8⁺ T cells. IL-6R KO T cells were used as negative controls (shaded). The gate defines the IL-6R^high (IL-6R^hi) population. (B and C) IL-6R MFI (B) and percentage of IL-6R^high (C) as in A (n = 3). (D) IL-6R expression on CD4/CD8 double positive (DP) and single CD4- or CD8-positive (CD4 SP or CD8 SP). Unstained DP thymocytes were used as negative control (shaded). (E) IL-6R expression on fresh CD8⁺ T cells (0 h) or after activation for the indicated periods of time. The unstained cells were used as negative controls (shaded). (F) Expression of CD25, KLRG1, CD127, and CD44 on IL-6R^high (IL-6R^hi) and IL-6R^low (IL-6R^lo) CD8⁺ T cells. The gate based on an unstained negative control is shown. (G) Percentage of CD44^high cells as gated in F (n = 3). (H) IL-6R MFI on CD44^high and CD44^low CD8⁺ T cells (n = 3). (I) FACS sorted CD44^low and CD44^high CD8⁺ T cells were activated in the absence or presence of IL-6. IL-21 production (72 h) was determined (n = 3). (J) Expression of IL-6R on OT-I CD8⁺ T cells (shaded histogram shows unstained cells). (K) IL-21 production from OT-I CD8⁺ T cells activated in the absence or presence of IL-6 for 72 h (n = 3). (L) FACS sorted IL-6R^hi and IL-6R^lo CD8⁺ T cells were activated in the absence or presence of IL-6. IL-21 production after 72 h was determined (n = 3). Error bars represent mean ± SD. *, P < 0.05, as determined by Student’s t test (B, C, G, H, and K) and two-way ANOVA (I and L). Results are representative of two to three experiments.

We performed a phenotypic analysis of the CD8⁺ T cell subset expressing detectable IL-6R (IL-6R^hi) before activation. No CD25 (IL-2Rα) expression could be detected in either the IL-6R^hi or IL-6R^lo subsets (Fig. 3 F). Only minimal KLRG1 (short-lived effector cell marker) and low levels of CD127 (IL-7R) were detected, but there was no difference between the two subsets (Fig. 3 F). However, the levels of CD44 were different between IL-6R^hi and IL-6R^lo CD8⁺ T cell subsets (Fig. 3 F). The IL-6R^hi CD8⁺ T cell subset contained a lower frequency of cells expressing high levels of CD44 (CD44^high; Fig. 3 G). Similarly, analysis of IL-6R expression in gated CD44^low (naive) and CD44^high (memory/activated) showed higher expression of IL-6R in naive CD8⁺ T cells (Fig. 3 H). Accordingly, analysis of IL-21 production in FACS-sorted CD44^high and CD44^low CD8⁺ T cells showed higher levels of IL-21 in CD44^low CD8⁺ T cells (Fig. 3 I). To further confirm the production of IL-21 in bona fide naive CD8⁺ T cells, we used CD8⁺ T cells from OT-I mice that express a TCR recognizing OVA. Similar to WT CD8⁺ T cells, IL-6R was expressed on most OT-I CD8⁺ T cells (Fig. 3 J). Whereas no IL-21 was detected in the absence of IL-6, OT-I CD8⁺ T cells activated with IL-6 produced high levels of IL-21 (Fig. 3 K). Thus, naive CD8⁺ T cells express IL-6R and can produce high levels of IL-21 during activation when IL-6 is present.

To determine whether the cell surface expression of IL-6R could be used as a marker to identify the subset of CD8⁺ T cells that can differentiate into IL-21–producing cells, IL-6R^hi and IL-6R^lo CD8⁺ T cells were FACS sorted and activated with or without IL-6. IL-6R^hi CD8⁺ T cells produced high levels of IL-21 in response to IL-6 (Fig. 3 L), whereas IL-21 was almost undetectable in IL-6R^lo CD8⁺ T cells (Fig. 3 L). Thus, the expression of IL-6R defines the subset of CD8⁺ T cells capable of producing IL-21.

IL-21–producing CD8⁺ T cells promote IgG isotype switching in B cells in vitro

IL-21 has an essential role in the B cell antibody responses (Ozaki et al., 2002), as it induces isotype switching to IgG isotypes, germinal center (GC) formation, and plasma cell differentiation (Ettinger et al., 2005; Linterman et al., 2010). IL-21 is produced primarily by CD4⁺ Tfh cells (Linterman et al., 2010). We investigated whether CD8⁺ T cells could also acquire a B cell helper function through IL-21. WT and IL-21R KO B cells were activated with LPS and an anti-CD40 Ab in the absence or presence of supernatant from WT or IL-21 KO CD8⁺ T cells previously activated with or without IL-6. Culture supernatants from WT CD8⁺ T cells activated with IL-6 significantly increased the production of IgG1 in WT B cells (Fig. 4 A). In contrast, supernatants from IL-21 KO CD8⁺ T cells had no effect on IgG1 production (Fig. 4 A). Supernatants from WT CD8⁺ T cells activated with IL-6 had no effect on IgG1 production by IL-21R KO B cells (Fig. 4 A), although these B cells still could produce IgG1 in response to IL-4 (Fig. 4 B). Similar results were found when B cells were activated with LPS alone (Fig. 4 C), an anti-IgM Ab alone (Fig. 4 D), or an anti-CD40 Ab alone (Fig. 4 E). Thus, IL-21 produced by CD8⁺ T cells in vitro is sufficient to promote IgG1 production by B cells when activated through different pathways.

Open in a separate window

Figure 4.

IL-6 reprograms CD8⁺ T cells to become B cell helpers through their production of IL-21 in vitro. (A) Culture supernatants (SN) were collected from WT and IL-21 KO CD8⁺ T cells activated in absence (Med SN, CD8) or presence (IL-6 SN, CD8) of IL-6 for 3 d. IgG1 production by WT and IL-21R KO B cells activated for 6 d with LPS and anti-CD40 Abs using these SN was determined (n = 3). (B) WT or IL-21R KO B cells were activated using LPS+IL-4 for 6 d in SN from WT CD8⁺ T cells activated for 3 d in the presence of IL-6. IgG1 production was determined (n = 3). (C–E) WT or IL-21R KO B cells were activated using LPS (C), anti-IgM Ab (D), or anti-CD40 Ab (E) in the presence of SN obtained from WT CD8⁺ T cells activated as in A. After 6 d, IgG1 production was determined (n = 3). (F and G) In co-culture system, WT or IL-21R KO B cells were activated with LPS while WT CD8⁺ T cells were activated using anti-CD3/CD28 Abs with or without IL-6. After 6 d, the levels of IgG1 (F) and IgM (G) in the co-cultures were determined (n = 3). (H) IgG1 levels in 6 d co-cultures of WT or IL-6 KO B cells with WT CD8⁺ T cells activated as in F (n = 3). (I) IgG1 levels in 6-d co-cultures of IL-6 KO B cells with WT or IL-6R KO CD8⁺ T cells activated as in F (n = 3). (J) Co-cultures of B cells and CD8⁺ T cells were established as described in F, but a combination of an anti-IgM Ab and CpG was used to activate B cells. IgG1 levels in the co-cultures were determined (n = 3). (K) IgG1 levels in 6 d co-cultures of WT B cells with WT or IL-21 KO CD8⁺ T cells activated as in J (n = 3). Error bars represent the mean ± SD. *, P < 0.05, as determined by Student’s t test (B), one-way ANOVA (H), or two-way ANOVA test (A, C, D, E–G, and I–K). Results are representative of two to three experiments.

We then established co-cultures of CD8⁺ T cells with B cells where CD8⁺ T cells were activated with anti-CD3/CD28 Abs with or without IL-6, and B cells with LPS. WT B cells, but not IL-21R KO B cells, co-cultured with CD8⁺ T cells in the presence of IL-6 produced higher levels of IgG1 than those co-cultured in the absence of IL-6 (Fig. 4 F), indicating that the effect of IL-6 on IgG1 production was dependent on IL-21 acting on B cells. IL-21 has no effect on the production of IgM (Ozaki et al., 2002). Accordingly, IgM levels were comparable between WT and IL-21R KO B cells co-cultured with CD8⁺ T cells, either with or without IL-6 (Fig. 4 G), showing that IL-21R KO and WT B cells are equally activated. In the absence of IL-6, the baseline IgG1 produced in co-cultures of IL-21R KO B cells was lower than in co-cultures with WT B cells (Fig. 4 F). Because TLR ligands can trigger IL-6 production in B cells (Barr et al., 2007), this could be a result of endogenous B cell–derived IL-6 inducing IL-21 in CD8⁺ T cells. To address this question, we established co-cultures using WT and IL-6 KO B cells with WT CD8⁺ T cells. The levels of IgG1 were practically undetectable in co-cultures containing IL-6 KO B cells (Fig. 4 H). Addition of exogenous IL-6 to those co-cultures was able to restore the effect of CD8⁺ T cells in promoting antibody production by B cells (Fig. 4 H). Moreover, addition of IL-6 to co-cultures with IL-6R KO CD8⁺ T cells failed to promote the production of IgG1 (Fig. 4 I). In addition to TLR4 (LPS receptor), CD8⁺ T cells also enhanced IgG isotype switching in B cells activated with an anti-IgM Ab and the TLR9 ligand unmethylated CpG when IL-6 is present (Fig. 4 J), and this help was dependent on their ability to produce IL-21 (Fig. 4 K). Together, these results demonstrate that CD8⁺ T cells activated in the presence of IL-6 promote IgG isotype switching of B cells in vitro through their ability to produce IL-21.

Cell purification and activation in vitro

In most experiments, CD8⁺ T cells were isolated from spleen and lymph nodes by negative selection, as previously described (Diehl et al., 2002). CD8⁺ T cells isolated by negative selection are typically >85% in purity and have <2% CD4⁺ T cell contaminants. CD8⁺ T cells from Stat3 conditional KO mice and IL-21 reporter mice and their corresponding WT controls, as well as CD8⁺ T cells from WT and IL-21 KO mice for adoptive transfer experiments, were purified by positive selection using anti-CD8 magnetic microbeads (Miltenyi Biotec). Positive selection resulted in >92% CD8⁺ T cell purity, with <0.5% CD4⁺ T cell contamination. CD8⁺ T cells from the lung were isolated using the lung dissociation kit (Miltenyi Biotec) as recommended by the manufacturer, followed by positive selection (Miltenyi Biotec). B cells from WT or IL-21R KO mice were purified by positive selection (Miltenyi Biotec; typically >92% in purity). IL-6R^hi and IL-6R^lo CD8⁺ T cells, and CD44^high and CD44^low CD8⁺ T cells, were isolated by immunostaining and cell sorting (FACSAria; BD). IL-6R KO cells stained with anti–IL-6R Abs were used as the negative control for IL-6R^hi and IL-6R^lo CD8⁺ T cell gating.

CD8⁺ T cells were activated with plate-bound anti-CD3 (145-2C11; 5 µg/ml) and soluble anti-CD28 (1 µg/ml; BioXCell) Abs in the presence or absence of IL-6 (50 ng/ml; Miltenyi Biotec), except where indicated. Human recombinant IL-2 (50 U/ml; BioLegend), mouse recombinant IL-12 (5 ng/ml; Amgen), and IL-15 (20 ng/ml; Amgen) were used. B cells were activated as listed below: first, LPS (1 µg/ml; Sigma-Aldrich); second, anti-IgM F(ab’)₂ (10 µg/ml; Jackson ImmunoResearch Laboratories) alone; third, anti-CD40 Ab (1 µg/ml; BD) alone; fourth, LPS (1 µg/ml) and IL-4 (100 U/ml); fifth anti-CD40 Ab (1 µg/ml) with LPS (1 µg/ml); and sixth, anti-IgM F(ab’)₂ (10 µg/ml) with CpG (1 µg/ml; Sigma-Aldrich).

Flow cytometry analysis

For IL-6R staining experiments, splenocytes were stained with anti-CD4 Ab (BioLegend), anti-CD8 Ab (Miltenyi), anti-CD44 Ab (BioLegend), anti–IL-6Rα Ab (BD), anti-KLRG1 Ab (eBioscience), anti-CD25 Ab (BioLegend), and anti-CD127 Ab (BD). For GC B cell staining, lung cell homogenates were generated using the lung dissociation kit (Miltenyi Biotec), as recommended by the manufacturer. Lung cells or spleen cells were stained with anti-CD8 Ab (Miltenyi Biotec), anti-Fas Ab (BD), anti-B220 Ab (BioLegend), anti-CD24 Ab (BioLegend), anti-IgD Ab (BD), and biotin-peanut agglutinin (Vector Laboratories), followed by secondary staining using Streptavidin staining Abs (BD). The immunostaining was always performed in the presence of Fc block (BD). For intracellular cytokine staining experiments, CD8⁺ T cells from IL-21 hCD4 reporter mice were activated for 3 d with or without IL-6. Cells were harvested, stained with anti-CD8 Ab (Miltenyi Biotec), anti-hCD4 Ab (BioLegend), and live-dead staining (Invitrogen). Cells were then fixed, permeabilized using saponin buffer (Sigma-Aldrich) as previously described (Yang et al., 1998), and stained with anti–IFN-γ Ab (BioLegend) or anti–IL-2 Ab (BioLegend). Expression of CD69 on activated cells was determined by immunostaining with an anti-CD69 Ab (BioLegend) and flow cytometry analysis. CD8⁺ T cell survival was determined by LIVE/Dead Viability kit, as recommended by the manufacturer (Molecular Probes) and flow cytometry analysis. CD8⁺ T cell proliferation was determined by CFSE staining (Molecular Probes) and flow cytometry analysis. All flow samples were run on FACS-LSR II (BD) and FlowJo Version 10.1 (FlowJo).

RNA isolation and real-time RT-PCR

Total RNA was isolated using the micro RNeasy kit (QIAGEN), as recommended by manufacture. cDNA synthesis was performed as previously described (Yang et al., 2015). cDNA was used for real-time RT-PCR assay using specific probe/primer sets for mouse Il21, Il2, Ifng, and β-microglobulin (Assays-on-Demand; Applied Biosystems). We followed the comparative C_T method (2^−ΔΔCT) as described by the manufacturer (Applied Biosystems; Livak and Schmittgen, 2001). In brief, Il21, Il2, and Ifng mRNA levels relative to the expression of β-microglobulin as housekeeping gene were obtained (ΔC_T). ΔΔC_T values were then obtained by subtracting the ΔCt value from a given reference sample (Figs. 1, I and K; and 2, B, C, and H) as calibrator to the rest of the samples. For Fig. 5 A, the mean of the ΔC_T value for the five individual mice in the WT/Lung CD8⁺ T cell group was used as calibrator. The final relative expression data are provided as 2^−ΔΔCT, defined as RQ value (relative quantitation).

ELISPOT assay

The detection of IL-21–producing CD8⁺ T cells by ELISPOT was performed as previously described (Dienz et al., 2009). In brief, purified CD8⁺ T cells (5 × 10⁴ cells/well) were plated in an anti–IL-21 Ab–coated ELISPOT plate and incubated (9 h) in the presence of PMA (20 ng/ml) and ionomycin (1 μg/ml). Plates were developed as previously described (Dienz et al., 2009). The number of spots (number of IL-21–producing cells) per well, as well as the total area of spots (overall IL-21 production) per well, were measured.

Western blot analysis

Western blot analyses were performed as previously described (Yang et al., 2015), using the following antibodies: anti-actin (Santa Cruz Biotechnology, Inc.), anti-pSTAT3 Tyr705 (Cell Signaling Technology), and anti-STAT3 (Cell Signaling Technology), anti–rabbit HRP (Jackson ImmunoResearch Laboratories), and anti–goat HRP (Santa Cruz Biotechnology, Inc.).

Lung imaging

Lungs frozen in Tissue-Tek O.C.T. compound (Sakura Finetek) were sectioned using a CM1850 cryostat (Leica). 30-µm frozen sections were blocked for 30 min with Background Buster (Innovex Biosciences) at room temperature, stained 6 h at 4°C in a humidified chamber with anti-CD8α (BD), anti-GL-7 (BioLegend), and anti-B220 (BioLegend) in PBS with 2% goat serum. Alternatively (for lymphoid structures in the lung), whole mount imaging was performed on 250 µm scalpel cut lung sections that were then fixed with 2% PFA (1 h at 4°C), blocked, and stained with the antibody cocktail mentioned above and anti-CD4 Ab (BD). Immunofluorescence confocal microscopy was performed with the Zeiss 780 laser scanning microscope (Carl Zeiss; air objective 20× Plan-Apochromat with NA [numerical aperture] 0.5 or water objective 10x Plan-Apochromat with NA 0.30) using multichannel frame scans. Image processing was performed using Imaris 8.1 software. Images were analyzed using LSM5 image browser and cells in three to four images from each mouse were counted using ImageJ (National Institutes of Health) cell counter plugin.

ELISA

IL-21, IL-2, IL-4, and IFN-γ levels in cell culture supernatants were determined by ELISA, as previously described (Diehl et al., 2000; Yang et al., 2015). IgG1 levels in cell culture supernatants were determined by ELISA, as previously described (Dienz et al., 2009). Levels of Influenza A PR8-specific IgG, IgM, IgG1, IgG2c, and IgG3 were determined using UV-inactivated influenza PR8 as antigen, as previously described (Dienz et al., 2009). Antibody levels in serially diluted serum samples were analyzed using HRP-conjugated Abs specific for mouse IgM, IgG, IgG1, IgG2c, and IgG3 (SouthernBiotech).

We thank T. Hunter and J. Hoffman for help with real-time PCR analysis (DNA Analysis Facility, University of Vermont, Burlington, VT), R. del Rio-Guerra for help with flow cytometry analysis (Flow Cytometry Facility, University of Vermont, Burlington, VT), Brian Silverstrim for technical help, and Roberta Pelanda and Raul Torres for helpful discussion (Department of Immunology, University of Colorado, Denver, CO).

This work was supported by National Institutes of Health grants R56AI094027 (to M. Rincon) and P20GM103496 (to M. Rincon and R. Yang). R. Yang was supported through the American Association of Immunologists Careers in Immunology Fellowship Program.

The authors declare no competing financial interests.

Author contributions: R. Yang, A.R. Masters, K.A. Fortner, D.P. Champagne, N. Yanguas-Casás, and D.J. Silberger performed experiments; R. Yang, and M. Rincon interpreted data; R. Yang, L. Haynes, C.T. Weaver, and M. Rincon designed experiments; and R. Yang, L. Haynes, C.T. Weaver, and M. Rincon wrote the paper.