Edited by Marc Feldmann, Imperial College London, London, United Kingdom, and approved June 27, 2011 (received for review May 10, 2011)
Tissue plasminogen activator is the only treatment option for stroke victims; however, it has to be administered within 4.5 h after symptom onset, making its use very limited. This report describes a unique target for effective treatment of stroke, even 12 h after onset, by the administration of αB-crystallin (Cryab), an endogenous immunomodulatory neuroprotectant. In Cryab−/− mice, there was increased lesion size and diminished neurologic function after stroke compared with wild-type mice. Increased plasma Cryab was detected after experimental stroke in mice and after stroke in human patients. Administration of Cryab even 12 h after experimental stroke reduced both stroke volume and inflammatory cytokines associated with stroke pathology. Cryab is an endogenous anti-inflammatory and neuroprotectant molecule produced after stroke, whose beneficial properties can be augmented when administered therapeutically after stroke.
Stroke is the most common cause of disability and the third most common cause of death in adults worldwide (1). The only US Food and Drug Administration (FDA)-approved treatment is tissue plasminogen activator (tPA), where therapy is time dependent and must be given within 4.5 h of stroke onset (2, 3). Consequently, there is an unmet need for therapy that could be administered later than tPA, and that is directed toward protection rather than clot dissolution.
There are several molecular mechanisms involved in the progression and evolution of the damage caused by stroke, including oxidative/nitrosative stress, excitotoxicity, apoptosis, calcium dysregulation, and inflammation (4⇓⇓⇓–8). Although several therapeutic approaches have been shown to be effective in various stroke animal models, the only FDA-approved drug is tPA, which acts through conversion of plasminogen into active plasmin to cleave the blood clot. tPA exemplifies strategies that aim to alter the obstructive blood clot, rather than protect the damaged brain.
Cryab, a member of the family of small heat shock proteins, designated sHSP B5, is constitutively expressed in the lens of the eye and muscle and is induced in many types of brain injury. Cryab is the most abundant induced transcript in multiple sclerosis (MS) lesions, and is highly expressed in areas of inflammation (9). Cryab has both antiapoptotic (10) and immunomodulatory properties (11). When first described in MS lesions, it was described as an autoantigen (12); however, subsequent studies demonstrated that Cryab acted as a negative regulator on inflammation and that it was imbued with protective properties for the nervous system (11, 13).
To investigate the effects of Cryab deficiency on cerebral ischemia, wild-type and Cryab−/− mice were exposed to 30 min of middle cerebral artery occlusion. The Cryab−/− mice had significantly larger lesion sizes at 2 d as assessed by triphenyltetrazolium chloride (TTC) staining, which stains the viable tissue in red due to the activity of the mitochondrial dehydrogenases (Fig. 1A). This difference remained at 7 d after stroke as assessed by silver stain (used to detect neuronal and axonal degeneration; Fig. 1B and SI Appendix, Fig. S1), indicating that the deficiency of Cryab affected both the early and delayed phases of ischemic damage. Functional outcome was assessed by a 28-point neurobehavioral scoring test (14, 15). The Cryab−/− mice had significantly worse scores at both 2-d and 7-d time points compared with wild-type controls (Fig. 1C). No differences were seen in cerebral blood flow measured by laser Doppler flow meter immediately after the occlusion and at 15 and 30 min of reperfusion between the groups (SI Appendix, Fig. S2).
Cryab−/− mice have larger lesions and worse neurological scores. (A) Representative images of TTC-stained brain sections of wild-type and Cryab−/− mice (Left) and quantification of lesion sizes 2 d after stroke (Right, n = 10 per group). (B) Representative images of silver-stained brain sections (Left) and quantification of lesion sizes 7 d after stroke (Right). (C) Quantification of 28-point neurological scoring in wild-type and Cryab−/− mice at 12 h, 2 d, and 7 d after stroke. For B and C, n = 10 and 7 for wild-type and Cryab−/− groups, respectively. *P < 0.05, Student t test. Data represent means ± SEM.
To investigate the role of the immune system in ischemic neuropathology associated with stroke, the profile of mononuclear cells in the ischemic wild-type and Cryab−/− brains was characterized. We identified four distinct cell populations according to their CD45 and CD11b expression patterns: CD11b+CD45low (microglia), CD11bhighCD45high (granulocytes and macrophages), CD11blowCD45high (subpopulation of monocytes), and CD11b−CD45high (lymphoid cells; SI Appendix, Fig. S3). The brain invasion of granulocytes and macrophages was higher in both groups at the earlier (2-d) time point, whereas the lymphoid cells increased more at the later (7-d) time points (Fig. 2A), consistent with other reports (16). The total number of microglial cells was equivalent in wild-type and Cryab−/− mice at both the 2-d and 7-d time points (Fig. 2B). However, the numbers of granulocyte and macrophage populations and the subpopulation of monocytes were significantly higher in the Cryab−/− mice brains at 2 d but not at 7 d compared with wild-type mice (Fig. 2B). Moreover, the numbers of lymphoid cells were significantly higher in the Cryab−/− group at 7 d (Fig. 2B).
There is more immune cell infiltration in the brains of Cryab−/− mice, and Cryab deficiency in the immune system causes larger lesion sizes after stroke. (A) Representative CD45 vs. CD11b flow cytometry plots of brain immune cells in wild-type and Cryab−/− mice 2 d and 7 d after stroke. (B) Quantification of total cell numbers of microglia (CD11b+CD45low), granulocytes and macrophages (CD11bhighCD45high), subpopulation of monocytes (CD11blowCD45high), and lymphoid cells (CD11b−CD45high) in wild-type and Cryab−/− mice brains at 2 d and 7 d after stroke. (C) Schematic plots showing the analysis method with F4/80 and Gr1 markers of CD11bhighCD45high population (F4/80−Gr1+: granulocytes, F4/80+Gr1+: activated macrophages, F4/80+Gr1−: macrophages). (D) Quantification of granulocytes, activated (act.) macrophages, and macrophages in wild-type and Cryab−/− mice brains 2 d and 7 d after stroke. (E) Representative CD45 vs. CD3 flow cytometry plots of brain immune cells in wild-type and Cryab−/− mice 2 d and 7 d after stroke. (F) Quantification of number of CD3+, CD3+CD4+, CD3+CD8+, and CD3+γδTCR+ T cells. n = 6 per group for each time point. *P < 0.05, Mann–Whitney analysis. (G) Representative γδ-TCR vs. IL-17a flow cytometry plots of brain immune cells in wild-type and Cryab−/− mice 7 d after stroke. (H) Scheme of timetable for the bone marrow-chimeric mice experiments. (I) Quantification of lesion sizes 7 d after stroke in the bone marrow-chimeric mice, wild-type cells into wild-type host: WT→WT; Cryab−/− cells into wild-type host: KO→WT; wild-type cells into Cryab−/− host: WT→KO; and Cryab−/− cells into Cryab−/− host: KO→KO (n = 10 for all groups except KO→KO, n = 6). *P < 0.05, **P < 0.01, #P = 0.063, Student t test. Data represent means ± SEM.
The granulocyte and macrophage population was further analyzed with Gr1 and F4/80 markers to identify granulocytes and macrophages separately. In the CD11bhighCD45high population, three subpopulations were identified (Fig. 2C): F4/80−Gr1+ (granulocytes), F4/80+Gr1+ (activated macrophages), and F4/80+Gr1− (macrophages). There were more granulocytes and significantly higher numbers of activated macrophages in Cryab−/− mice at 2 d (Fig. 2D). The number of macrophages was equivalent at both time points between the groups (Fig. 2D).
The analysis of T cells (CD3+) showed that the total number of T cells increased at 7 d compared with 2 d in both groups, and there were significantly more T cells in the brains of Cryab−/− mice than wild-type mice at 7 d (Fig. 2 E and F). When we analyzed the T-cell subpopulations, no difference in CD4+ and CD8+ T cells was observed between wild-type and Cryab−/− groups (Fig. 2F); however, there were significantly more γδ-TCR+ (γδ-T) cells in the brains of Cryab−/− mice at 2 d and 7 d after ischemia (Fig. 2F). Moreover, these γδ-T cells were producing IL-17 (Fig. 2G), causing damage rather than tolerance as suggested previously (7, 17–20).
The number of cells in various populations in spleens of Cryab−/− mice was significantly decreased at 2 d after stroke, suggesting a stronger inflammatory response associated with larger lesion sizes (21) (SI Appendix, Tables S1 and S2). There was no difference in the numbers in various splenocyte populations before and 7 d after stroke between wild-type and Cryab−/− mice (SI Appendix, Tables S1 and S2). There were no differences in the number of CD11b+ cells and CD3+, CD4+, and CD8+ T cells in blood of wild-type and Cryab−/− mice before, or 2 d and 7 d after, stroke (SI Appendix, Tables S3 and S4). However, there were more γδ-T cells in the blood of Cryab−/− mice before and 7 d after, but not 2 d after, stroke (SI Appendix, Table S4). The difference of blood γδ-T-cell numbers between Cryab−/− and wild-type mice was 2.7-fold before stroke and sixfold at 7 d, implying a stronger inflammatory response in Cryab−/− mice (SI Appendix, Table S4). There were no differences in the number of any mononuclear cell populations in the brains of wild-type and Cryab−/− mice before stroke (SI Appendix, Table S5). Overall, these data suggest that a deficiency of Cryab might lead to a more vigorous inflammatory response to stroke. The extent of inflammation can be proportional to the lesion size (22), and various inflammatory markers correlate with clinical outcome after stroke (23).
Cryab has been shown to have anti-inflammatory properties by inhibition of NF-κB and p38 MAP kinase (11) as well as antiapoptotic properties by inhibition of activation of caspase-3 (10, 24–26). To further dissect whether the more vigorous inflammatory response seen in Cryab−/− mice might be due to the larger lesion sizes in Cryab−/− mice, or vice versa, we performed bone marrow chimera experiments. After a lethal dose of total-body γ-irradiation, wild-type or Cryab−/− bone-marrow cells were injected into either wild-type or Cryab−/− mice to regenerate the immune system. We had four groups: wild-type cells transferred to wild-type hosts (WT→WT), wild-type cells transferred to Cryab−/− hosts (WT→KO), Cryab−/− cells transferred to wild-type hosts (KO→WT), and Cryab−/− cells transferred to Cryab−/− hosts (KO→KO). The irradiation was established to be lethal, by the death within 2 wk of irradiated wild-type and Cryab−/− mice that were not given the cells. Six weeks after the irradiation, we confirmed the chimerism by PCR of blood leukocytes for Cryab so that the mice that were given wild-type and Cryab−/− cells had only wild-type and Cryab−/− genotypes in their immune system, respectively (Fig. 2H). The KO→KO group had significantly larger lesion sizes than WT→WT group (Fig. 2I). This is concordant with the previous finding of Cryab−/− mice having larger lesions than wild-type mice. The KO→WT group had significantly larger lesions than the WT→WT group (Fig. 2I). There was also a trend of larger lesions in the KO→KO group than the WT→KO group (Fig. 2I). Taken together, these studies indicate that a deficiency of Cryab in the immune system is associated with larger lesions in both wild-type and Cryab−/− hosts. Therefore, the presence or absence of Cryab in the immune system is critical in the evolution of infarct. Furthermore, the WT→KO group had significantly larger lesions than the WT→WT group (Fig. 2I). There was also a trend of larger lesions in the KO→KO group than in the KO→WT group (Fig. 2I), which suggests that the difference in lesion sizes between wild-type and Cryab−/− mice is due not only to the Cryab deficiency in the immune system but also to the deficiency of Cryab outside the immune system and within the central nervous system. These results indicate that Cryab deficiency in both the immune system as well as Cryab deficiency in the brain independently contribute to the larger lesions in Cryab−/− mice compared with wild-type mice.
Several reports have shown that Cryab expression is up-regulated in neurons (27) and astrocytes (28) after cerebral ischemia, although some reports indicate that it is not up-regulated in brain after ischemia (29). We analyzed levels of Cryab in plasma in wild-type mice before and at 12 h, 2 d, and 7 d after stroke onset by ELISA. The Cryab levels were significantly increased at the 12-h time point, with a gradual decrease over the ensuing 7 d (Fig. 3A).
Plasma Cryab increases after stroke, and restoration of plasma Cryab in Cryab−/− mice confers neuroprotection and modulates the peripheral immune response. (A) Quantification of plasma Cryab levels in naïve mice and in mice at 12 h, 2 d, and 7 d after stroke (n = 8, 11, 10, and 8 for naïve, 12-h, 2-d, and 7-d groups, respectively). *P < 0.05, Kruskal–Wallis analysis. (B) Quantification of the levels of Cryab in human plasma and (C) correlation of lesion volume with the levels of Cryab in human plasma at presentation (<4 h), 24 h, and 72 h after the symptom onset (n = 5, 7, and 6 for healthy controls, younger and older patients, respectively). (D) Quantification of lesion sizes in PBS-treated wild-type, PBS-treated Cryab−/−, and Cryab-treated Cryab−/− mice 7 d after stroke as assessed by silver stain (Upper) and timetable of Cryab injections as indicated by vertical black arrows (Lower; n = 15, 12, and 13 for PBS-treated wild-type, PBS-treated Cryab−/−, and Cryab-treated Cryab−/− groups, respectively). *P < 0.05, one-way ANOVA. (E) Quantification of cytokines after stimulation with anti-CD3/anti-CD28, Concanavalin-A (ConA), or LPS; representative graphs from one of the two experiments with similar results. *P < 0.05, **P < 0.01, ***P < 0.005, #P = 0.053, ##P = 0.056, §P = 0.06, Student t test.
As an exploratory approach, we also examined the plasma concentrations of Cryab in patients with ischemic stroke who were admitted to the Stanford Stroke Center and healthy young individuals. All patients were treated according to Stanford Stroke Center's current protocols. After the analysis, the patients were grouped according to their age, as younger (ages 39–66) and older (ages 82–93; SI Appendix, Tables S6 and S7). Cryab levels at presentation to the hospital (<4 h after symptom onset for all patients) were higher in the younger patient population compared with the control group (Fig. 3B). However, the older patients did not show increased levels of Cryab (Fig. 3B). Because the lesion size might affect the response, the lesion volumes were determined by diffusion-weighted magnetic resonance imaging. Although the mean lesion volume was higher in the older population, there was no significant difference between older and younger patient groups (SI Appendix, Fig. S4). Interestingly, the lesion volume and the plasma Cryab levels at presentation highly correlated only in the younger patient group and not in the older group (Fig. 3C), perhaps indicating that the body's endogenous response to stroke is age dependent. These findings have not been observed before for any small heat shock protein after stroke. There are several potential biomarkers in plasma (e.g., C-reactive protein, S100B, and matrix metalloproteinase-9) that can be used for diagnosis or outcome prediction of stroke (30). Given that Cryab is secreted from human retinal pigment epithelial cells (31), whether it can be used as a biomarker after stroke requires more extensive studies. Variations with age of the patient must be considered in such studies.
To test the hypothesis that increased plasma Cryab after stroke was beneficial, Cryab−/− mice were given i.p. injections of recombinant Cryab protein starting 1 h before stroke onset and continuing at 12 h, 24 h, and daily afterward for 7 d in total. The lesion sizes in the Cryab-treated Cryab−/− mice were significantly decreased, to the levels of PBS-treated wild-type mice, compared with PBS-treated Cryab−/− mice (Fig. 3D).
The effects of Cryab treatment on splenocytes isolated from PBS-treated wild-type and Cryab−/− mice and Cryab-treated Cryab−/− mice were assessed at 7 d after stroke. Splenocytes from PBS-treated Cryab−/− mice produced more proinflammatory IL-2, IL-17, and TNF when stimulated with anti-CD3/anti-CD28 than both PBS-treated wild-type mice and Cryab-treated Cryab−/− mice splenocytes (Fig. 3E). When stimulated by concanavalin-A, splenocytes from PBS-treated Cryab−/− mice produced more proinflammatory IFN-γ, TNF, and IL-12p40, and less anti-inflammatory IL-10 compared with the splenocytes from both PBS-treated wild-type and Cryab-treated Cryab−/− mice (Fig. 3E). When stimulated with LPS, splenocytes from PBS-treated Cryab−/− mice produced more TNF and less IL-10 compared with the splenocytes from both PBS-treated wild-type mice and Cryab-treated Cryab−/− mice (Fig. 3E). These data indicate that restoration of plasma Cryab by systemic treatment modulates the peripheral inflammatory response and is sufficient to decrease the lesion sizes in Cryab−/− mice to the levels of wild-type mice after stroke.
Because the restoration of plasma Cryab in Cryab−/− mice decreased the lesion size, we next investigated whether administration of Cryab into wild-type mice would have a similar effect. When Cryab was administered 1 h before and 12 h after the stroke onset, the lesion size at 2 d was not different between PBS- and Cryab-treated wild-type mice groups (Fig. 4A). However, when it was administered 1 h before, 12 h and 24 h after, and daily afterward for 7 d in total, the lesion sizes were significantly reduced in the Cryab-treated group compared with the PBS-treated group (Fig. 4B). Moreover, starting the initial treatment even 12 h after the stroke onset—making the treatment highly relevant if translated into the clinic—conferred neuroprotection in the Cryab-treated group (Fig. 4C). There were no differences between PBS- and Cryab-treated groups in blood pressure, pulse rate, plasma glucose and lactate levels, and arterial blood gas analysis before, during, or after cerebral ischemia (SI Appendix, Tables S8 and S9).
Cryab administration to wild-type mice reduces the lesion size after stroke and modulates the peripheral immune response. (A) Quantification of lesion sizes in PBS- and Cryab-treated wild-type mice 2 d after stroke as assessed by TTC staining (Upper) and timetable of injections (Lower; n = 7 per group). (B) Quantification of lesion sizes in PBS- and Cryab-treated (starting the treatment 1 h before stroke) wild-type mice 7 d after stroke as assessed by silver stain (Upper) and timetable of injections (Lower; n = 7 per group). (C) Quantification of lesion sizes in PBS- and Cryab-treated (starting the treatment 12 h after stroke) wild-type mice 7 d after stroke as assessed by silver stain (Upper) and timetable of injections (Lower; n = 14 and 15 for PBS- and Cryab-treated groups, respectively). *P < 0.05, **P < 0.005, Student t test. (D–K) Quantification of IL-2, IL-17, IFN-γ, IL-12p40, IL-6, IL-10, and TNF levels after ConA and LPS stimulation of splenocytes from PBS- and Cryab-treated mice 7 d after stroke; representative graphs from one of the four experiments with similar results. *P < 0.05, **P < 0.01, ***P < 0.005, #P = 0.053, Student t test.
The effects of Cryab treatment on the splenocyte cytokine response were assessed 7 d after stroke in wild-type mice. The total number of splenocytes in PBS- and Cryab-treated mice was equivalent 7 d after stroke. When stimulated by Con A, splenocytes from Cryab-treated mice produced fewer proinflammatory IL-2, IL-17, IFN-γ, IL-12p40, and IL-6, and more anti-inflammatory IL-10 than the splenocytes from PBS-treated mice (Fig. 4 D–I). When stimulated by LPS, Cryab-treated mice splenocytes produced fewer proinflammatory IL-6 and TNF than the splenocytes from PBS-treated mice (Fig. 4 J and K).
The T-cell populations in brain were analyzed 7 d after stroke. Cryab treatment did not change the total number of the brain CD3+, CD4+, CD8+, and γδ-T cells (Fig. 5 A–E). Moreover, there were no differences in the percentages of IL-17+ cells among CD3+, CD4+, or γδ-T cells and of IFN-γ+ cells among CD3+ or CD4+ cells between Cryab-treated and PBS-treated mice brains 7 d after stroke (SI Appendix, Fig. S5). However, when we quantitatively analyzed how many cytokine molecules were present per cell, we found fewer IL-17-PE molecules per CD3+, CD4+, and γδ-T cells in Cryab-treated mice brains compared with PBS-treated ones (Fig. 5 F and G). There were no differences in the number of IFN-γ-PE molecules per CD3+ or CD4+ cells between the groups (Fig. 5H).
The effect of Cryab treatment on brain T cells after stroke. (A) Representative flow cytometry plots of CD3+ cells from PBS- and Cryab-treated mice brains at 7 d after stroke. (B–E) Quantification of CD3+, CD4+, CD8+, and γδ-T cells in brains of PBS- and Cryab-treated mice 7 d after stroke (n = 6 and 7, respectively); pooled analysis of two experiments with similar results. (F) Representative histogram analysis of IL-17+ γδ-T cells in brains of PBS- and Cryab-treated mice 7 d after stroke. (G) Quantitative analysis of number of IL-17-PE molecules per cell in brains of PBS- and Cryab-treated mice 7 d after stroke (n = 10 and 11, respectively); pooled analysis of three experiments with similar results. *P < 0.05, #P = 0.053, Student t test. (H) Quantitative analysis of number of IFN-γ-PE molecules per cell in brains of PBS- and Cryab-treated mice 7 d after stroke (n = 4 and 6, respectively); pooled analysis of two experiments with similar results.
Our findings describe a therapeutic role for Cryab in stroke, and emphasize how it functions as an endogenous neuroprotectant by modulating the immune system. tPA is also an endogenous protein that increases in plasma after stroke (32) and is augmented when given therapeutically. In a similar logic, administration of Cryab is also augmenting a naturally occurring anti-inflammatory pathway.
Cryab is a key protective response element of the body after stroke. Variation in plasma Cryab levels may be one of the several differences between younger and older patients to explain the worse outcome in older patients (33, 34). Earlier studies have described Cryab as a guardian molecule in brain inflammation in multiple sclerosis, and this descriptive name clearly translates to its role in stroke (10, 35). Its presence as an endogenous protectant can be exploited by administering it in larger quantities as a therapeutic agent. Its benefit seen with starting the treatment 12 h after stroke would represent a significant improvement over tPA if translated to the clinic.
Several immunomodulatory agents for ischemic stroke treatment have the potential to provide a more expanded time window than provided by tPA to start intervention (7, 36). However, some potential targets in the immune system have not been fully characterized (37). Many defined targets have been described that show promise (7, 36–42). Cryab is notable because it is endogenously produced after stroke, and has known anti-inflammatory effects on inflammatory immune responses (11). Its further augmentation when administered exogenously reduces stroke volume and inflammatory cytokine production.
Recently, another exciting approach for stroke was published showing that modulation of GABA could influence stroke in a model system in rodents (43). The GABA pathway was shown to have inhibitory effects on the immune system, and GABA agonists could suppress CNS inflammation in vivo (44). In stroke, GABA modulation beginning 3 d after stroke promoted functional recovery but did not influence the size of the stroke (43), in contrast to the current results with Cryab, where stroke size was reduced when the treatment was started even 12 h after the stroke. New approaches to protect the nervous system with administration of recombinant protectants such as Cryab and with inhibitors of synaptic modulators such as inverse agonists of GABAA receptors may offer real advantages over approaches aimed at merely shrinking the arterial blood clot or inhibiting coagulation.
Some proposed targets for therapy of stroke have dual roles, causing deleterious effects initially and becoming beneficial later on (45). Moreover, there is an endogenous response for compensation and remodeling (46). These physiological responses after stroke can be influenced by genetic, environmental, physiological, or pharmacological factors, as well as cross-talks between neuronal, glial, vascular, and immune compartments (46). Cryab exemplifies a molecule with intricate protective properties, elicited following damage to the CNS, in wide-ranging conditions, including inflammation in MS (11), retinal ischemia (13), and, as described here, stroke. Cryab suppresses inflammation, acting at the interface of the central nervous system and the immune system.
All animal procedures were approved by the Stanford University Administrative Panel on Laboratory Animal Care. Filament occlusion stroke model was used as described in SI Appendix. The lesion sizes were assessed by TTC and silver stains. Flow cytometry analysis was performed on a Becton Dickinson LSR-II (Stanford Shared FACS Facility) as described in SI Appendix. Human samples were obtained from Stanford Stroke Center. Statistical methods are described in detail in SI Appendix.
Detailed materials and methods are described in SI Appendix.
We thank Andrew Nepomuceno for performing the neuroscore test; May Han, Robert C. Axtell, and Michael Kurnellas for critical suggestions; David Kunis for laboratory management; and Elizabeth Hoyte for figure preparation. This study was supported in part by National Institutes of Health Grant 1PO1 NS37520-08; Russell and Elizabeth Siegelman; Bernard and Ronni Lacroute; the William Randolph Hearst Foundation (G.K.S.); the National Multiple Sclerosis Society (L.S.); a Stanford University Graduate Fellowship (S.E.B.); and a Stanford University School of Medicine Dean's Fellowship from the Evelyn L. Neizer Fellowship Fund (A.A.).
↵1A.A. and S.E.B. contributed equally to this work.
↵2L.S. and G.K.S. contributed equally to this work.
Author contributions: A.A., S.E.B., J.B.R., L.S., and G.K.S. designed research; A.A., S.E.B., J.B.R., R.M.K., and M.P.P. performed research; J.B.R., C.C., and G.W.A. contributed new reagents/analytic tools; A.A., S.E.B., C.C., R.M.K., M.P.P., and G.W.A. analyzed data; and A.A., S.E.B., J.B.R., L.S., and G.K.S. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1107368108/-/DCSupplemental.
