PKR inhibits the DNA damage response, and is associated with poor survival in AML and accelerated leukemia in NHD13 mice

The double-stranded RNA-activated protein kinase (PKR) can be activated by a variety of cellular stresses to play a pivotal role in proapoptotic and inflammatory signaling pathways.^1-11  Due to proapoptotic functions, PKR has been considered to have tumor suppressor properties. However, PKR knockout mice (PKRKO) do not display any increased tumor incidence.¹²  In addition, we recently discovered that mice expressing a PKR transgene (transgenic PKR [TgPKR]) specifically in hematopoietic cells develop a preleukemic, myelodysplastic syndrome (MDS)-like phenotype that includes bone marrow (BM) dysplasia and increased BM blasts.³  Furthermore, increased PKR has been reported in patients with acute leukemias, as well as breast, melanoma, and colon cancers.^13-17  Thus, PKR may have a previously unrecognized role that contributes to oncogenesis.

Although the role of PKR in the cytoplasm to inhibit protein synthesis has been well studied, at least 20% of PKR resides in the nucleus but the function of nuclear PKR is unclear.¹⁸  Increased nuclear PKR activity is found in CD34⁺ blasts from high-risk MDS patients but not from low-risk MDS patients or healthy donors, and PKR is mainly nuclear in phosphatase and tensin homologue-deficient acute leukemia cell lines, suggesting that nuclear PKR signaling may play a role in tumorigenesis.^17,19,20  Since we recently observed that Lineage-negative (Lin⁻) cells isolated from BM of TgPKR mice are more sensitive, whereas cells from PKRKO mice are highly resistant to genotoxic stresses including ionizing irradiation (IR),³  we investigated whether nuclear PKR may regulate DNA damage response (DDR) signaling and whether decreased PKR expression/activity may safeguard genomic fidelity.

Peripheral blood (PB) and BM were collected from 414 patients with newly diagnosed AML at The University of Texas M. D. Anderson Cancer Center between September 1999 and March 2007 during routine diagnostic assessments under Institutional Review Board-approved protocol 05-0654. Informed consent was obtained in accordance with the Declaration of Helsinki. Proteomic profiling of PKR in CD34⁺ cells was performed by reverse phase protein array (RPPA) using the PKR M02 monoclonal antibody, clone 1D11 (Abnova, Taipei, Taiwan). Isolation of CD34⁺ cells from patient samples, RPPA processing, and statistical analysis was performed as described.^21,22  In addition, CD34⁺ cells were isolated from an independent set of 6 randomly selected AML patients by positive selection using a CD34 MicroBead Kit (Miltenyi Biotec, Auburn, CA), and primary human CD34⁺ isolated from the BM of healthy donors were purchased from Lonza (Walkersville, MD). Measurement of PKR gene expression in CD34⁺ AML and normal cells is described in supplemental Methods on the Blood Web site.

C57BL/6-Tg (Vav-NUP98/HOXD13) (NHD13) mice were purchased from The Jackson Laboratory (#010505; Bar Harbor, ME). PKR (TgPKR) and PKRKO mice were described previously.³  Breeding to generate NHD13-TgPKR and NHD13-PKRKO, and genotyping and analysis of PB and BM for evidence of MDS/leukemia is described in supplemental Methods. In addition, treatment of mice with PKR inhibitor (PKRI) is described in supplemental Methods. The University of Florida Institutional Animal Care and Use Committee approved these experiments (#201102224 and #201208669).

REH, K562, and HL60 cells were from American Type Culture Collection (Manassas, VA) and propagated under standard growth conditions.¹¹  Generation of cell lines expressing short hairpin RNA (shRNA) is described in supplemental Methods. PKRI compound, okadaic acid (OA), and anti-phospho(serine 1981)-ataxia-telangiectasia mutated (p-ATM) antibody, clone 10H11.E12, were purchased from EMD Millipore (Darmstadt, Germany). FTY720 was obtained from Cayman Chemical Company (Ann Arbor, MI). Antibodies for actin, HSP90 protein phosphatase 2A (PP2A)/C, PP2A-B55α, and PP2A-B56γ were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). p-PKR (threonine 451) antibody was purchased from Invitrogen/BioSource (Grand Island, NY). Antibodies for H2AX, γ-H2AX, ATM, LSD1, NBS1, and p-NBS1 (serine 343) were purchased from Cell Signaling Technology (Beverly, MA).

Staining of surface antigens was performed as described.³  For staining of intracellular antigens, cells were fixed with 1% methanol-free formaldehyde, washed with phosphate-buffered saline (PBS), and permeabilized with 70% ethanol. After washing with PBS containing 1% bovine serum albumin and 0.02% Tween 20, cells were stained with primary antibody overnight at 4°C, and washed and stained with fluorochrome-conjugated secondary antibody. Flow cytometry was performed on an Accuri C6 flow cytometer at the University of Florida core facility (BD Biosciences, San Jose, CA).

Cell cytospins were prepared, fixed with 4% paraformaldehyde for 30 minutes, permeabilized with 1% Triton in PBS for 10 minutes and stained with a 1:30 dilution of the indicated primary antibody, and 1:100 dilution of fluorochrome-conjugated secondary antibody. Slides were mounted with ProLong Gold antifade reagent containing 4,6 diamidino-2-phenylindole (DAPI, Life Technologies, Grand Island, NY) and viewed with a Zeiss Axioplan 2 microscope equipped with epifluorescence optics at ×63. Images were captured using Openlab 5.

Nuclear and cytoplasmic subcellular fractions were prepared using an NE-PER Kit according to the manufacturer’s protocol (Pierce, Rockford, IL). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis and immunoblotting was performed as previously described.¹¹  For IP, 500 μg of protein lysate was precleared with nonspecific mouse immunoglobulin (Ig)G and protein G, then incubated overnight with the indicated antibody (1:100 dilution) and protein G agarose (Life Technologies, Grand Island, NY) at 4°C. Agarose-conjugated complexes were washed with PBS and detected by immunoblotting as described.¹¹  PP2A was immunoprecipitated from nuclear extracts and activity assayed using a PP2A IP phosphatase kit (17-313; Millipore, Billerica, MA) according to the manufacturer’s protocol and as previously described.¹¹ 

DNA repair was quantified using a Comet Assay Kit and neutral conditions according to the manufacturer’s protocol (Trevigen, Gaithersburg, MD). SYBR green-stained slides were observed at ×40 by fluorescence microscopy using the equipment and software described above. Comet Olive Tail Moment calculations were performed on 50 randomly chosen cells for each sample by Wimasis Image Analysis (Munich, Germany).

Data are presented as the mean ± standard error of the mean of at least 3 independent experiments. Significance (P < .05) was determined by Student t test using Prism 6 (GraphPad Software, La Jolla, CA). Kaplan–Meier analysis and Cox proportional hazard modeling were performed on RPPA results using Statistica Version 10 (StatSoft). Kaplan–Meier and Log-rank analysis of mice were performed using Prism 6.

Increased PKR expression and activity have been reported in leukemia patient samples and leukemia-derived cell lines.^16,17,20  To elucidate the role of PKR in leukemia, we tested whether PKR expression may be associated with clinical outcome by RPPA analysis of primary CD34⁺ blast cells from 414 newly diagnosed AML patients.^22,23  When stratified for PKR, AML patients with the highest one-third of PKR expression in CD34⁺ cells had a significantly worse overall survival (OS) than patients with the lowest two-third of PKR expression (Figure 1A) (median survival 41.6 vs 54.6 weeks; P < .05). Furthermore, patients with low PKR expression were approximately twice as likely to remain in remission after 2 years compared with patients with high PKR expression (Figure 1B). These correlations were most significant for AML patients with unfavorable cytogenetics (Figure 1C-D and supplemental Figure 1), and are similar to other published reports and those in curated databases such as www.oncomine.org and www.kmplot.com, which indicate that high PKR expression is found in tumor vs normal tissues and correlates with reduced survival in large cohort studies of breast, lung, and ovarian cancer patients (supplemental Figure 2).^13-16  Because PKR’s reported proapoptotic function cannot readily account for these findings, we examined whether PKR may have a previously unrecognized oncogenic function.

Because we previously demonstrated that cells from PKRKO mice or treated with a pharmacologic PKRI are resistant to genotoxic stress,^3,11  we tested whether PKR may regulate DDR signaling in primary CD34⁺ cells isolated from BM of AML patients or healthy donors, murine Lin⁻ BM cells, and leukemia cell lines. As a measure of DDR signaling, ATM autophosphorylation on serine 1981 (p-ATM), phosphorylation of histone H2AX (γ-H2AX), and phosphorylation of the ATM target NBS1 (p-NBS1) on serine 343 were examined by western blotting, IF microscopy, and flow cytometry (supplemental Figure 3A-C).^24-26 

Phospho(threonine 451)-PKR (p-PKR), indicative of activated PKR,²⁷  was observed in the nucleus of CD34⁺ cells isolated from BM or PB of 6 AML patients, and treatment with a PKRI effectively decreased activated but not total PKR (supplemental Figure 3D and data not shown). Furthermore, IF staining and flow cytometry demonstrated that PKRI treatment increased p-ATM, γ-H2AX, and p-NBS1 in primary CD34⁺ AML cells after IR (Figure 2A-C). In addition, the rate and abundance of γ-H2AX foci formation following IR, as well as resolution of foci, was increased by PKRI treatment of CD34⁺ AML cells compared with control cells (Figure 2D). Significantly, when the percentage of cells positive for γ-H2AX or p-ATM after IR was plotted against relative PKR expression for each AML sample, patients with the highest PKR expression in CD34⁺ BM cells displayed the lowest IR-induced γ-H2AX and p-ATM (Figure 2E-F). In support of this, the inhibition of PKR activity by either PKRI treatment or knockdown of PKR expression by shRNA in primary CD34⁺ cells also demonstrated increased IR-induced γ-H2AX and p-ATM formation (Figure 2G-H and supplemental Figure 3F-K). Thus, an inverse relationship exists between PKR and DDR signaling in primary human CD34⁺ hematopoietic stem and progenitor cells from AML patients or healthy donors.

To determine whether increased PKR expression inhibited DDR signaling in primary hematopoietic cells, we compared DDR signaling following IR in Lin⁻ cells collected from the BM of TgPKR, PKRKO, and wild-type (WT) mice. Lin⁻ cells isolated from the BM of TgPKR mice show a significantly impaired rate and amplitude of γ-H2AX formation compared with cells from WT or PKRKO mice (Figure 2I). Furthermore, Lin⁻ cells from TgPKR mice have a decreased percentage of cells positive for p-ATM and p-NBS1 (Figure 2J-K), and treatment of TgPKR Lin⁻ BM cells with PKRI restored DDR signaling to the level of WT Lin⁻ cells (supplemental Figure 3E and data not shown).

Next, we tested whether PKR expression or activity was required to regulate DDR signaling in leukemic cell lines. Knockdown of PKR in REH, K562, or HL60 cells (supplemental Figure 3L) promoted more rapid γ-H2AX formation and significantly increased p-ATM and p-NBS1 following IR as compared with control cells (Figure 2L and supplemental Figure 3L-R). Furthermore, inhibition of PKR activity by PKRI treatment also promoted increased p-ATM and p-NBS1 after IR (Figure 2M, lane 4 vs lane 2). In contrast, treatment of cells with either poly(I:C), a double-stranded RNA activator of PKR, or interferon (IFN)-γ, to increase PKR expression, significantly reduced the level of p-ATM and p-NBS1 following IR and delayed γ-H2AX formation (Figure 2M, lanes 6 and 8 vs lane 2, and supplemental Figure 3S-T). These results indicate that PKR kinase activity has a previously unrecognized role to antagonize DDR signaling in primary human and mouse hematopoietic cells as well as in leukemic cell lines.

Next, we tested whether PKR was a component of any ATM-containing complex by reciprocal co-IP. PKR co-IPs with ATM under normal growth conditions and PKRI treatment decreased, whereas poly(I:C) increased PKR-ATM co-precipitation (Figure 3A). Importantly, a significant reduction of PKR-ATM co-precipitation was observed after IR (Figure 3B). Furthermore, inhibition of PKR expression or activity promoted co-IP of ATM with NBS1 or γ-H2AX following IR (Figure 3C, lanes 4 and 5 vs lane 3). In contrast, treatment with poly(I:C) or IFN-γ greatly reduced association of these proteins (Figure 3C, lanes 6 and 7 vs lane 3), strongly suggesting that activated PKR interacts with ATM to antagonize ATM activation and association with downstream targets.

Our laboratory and others have demonstrated that PKR can activate PP2A, which has been reported to inhibit autophosphorylation and activation of ATM.^8,9,11,28  Thus, we tested whether nuclear PKR may inhibit ATM by a mechanism dependent on PP2A.

PKR efficiently co-precipitated with PP2A from the nuclear fraction of REH and K562 cell lysates, and this interaction was significantly decreased following IR (Figure 4A). Furthermore, cells with knocked-down PKR expression (sh-PKR) had an approximately twofold decrease in nuclear PP2A activity, suggesting that PKR is necessary to maintain steady-state nuclear PP2A activity (Figure 4B). Significantly, the association of ATM and PP2A is decreased in sh-PKR knockdown cells under normal growth conditions and to an even greater extent following IR (Figure 4C, lane 2 vs lane 1 and lane 5 vs lane 4). In all cases, decreased PP2A-ATM association was concomitant with increased p-ATM in nuclear extracts (Figure 4C). Furthermore, treatment of sh-PKR cells with the PP2A activator, FTY720, restored PP2A co-precipitation with ATM and reduced p-ATM to a level similar to that of control cells (Figure 4C, lane 3 vs lane 2 and lane 6 vs lane 5). Furthermore, although activation of PP2A activity by FTY720 promoted, inhibition of PP2A activity by OA decreased PKR-ATM association in nuclear extracts, suggesting that PKR-ATM association requires PP2A activity (Figure 4D). These results indicate that nuclear PKR promotes PP2A-ATM interaction, which antagonizes ATM autophosphorylation.

Because PP2A regulatory B subunits may be a downstream target of PKR, and IR has been reported to induce dissociation of the PP2A B55α subunit from the nuclear PP2A heterotrimer, we tested whether PKR may regulate nuclear B55α.^8,9,11,28,29  Following IR, B55α decreased in the nucleus and increased in the cytoplasm (Figure 4E, lane 1 vs 3 and lane 5 vs 7). Furthermore, sh-PKR cells had markedly decreased nuclear B55α that was even more evident following IR compared with control nuclear extracts (Figure 4E, lane 2 vs 1 and lane 4 vs 3). Significantly, when B55α expression was knocked down (Figure 4F, sh-B55α), cells with reduced B55α displayed increased γ-H2AX and p-ATM following IR compared with control (sh-ctrl) cells (Figure 4G, lane 5 vs 2 and Figure 4H). Furthermore, inhibition of PKR activity by PKRI treatment failed to promote γ-H2AX or p-ATM following IR in sh-B55α cells as it did in control cells (Figure 4G, lane 6 vs 3 and Figure 4H). Thus, nuclear PKR suppresses ATM autophosphorylation by promoting nuclear localization of B55α and activation of PP2A (Figure 4I).

Next, we measured whether PKR may also delay the kinetics of DNA DSB repair using a neutral Comet assay to specifically detect DSBs.³⁰  At various times following IR, DNA damage was measured by calculating the average Olive Tail Moment of 50 randomly chosen cells.^31,32  Significantly, primary CD34⁺ AML cells treated with PKRI displayed a more rapid and complete repair of DSBs 24 hours after irradiation than untreated control cells (Figure 5A-B). Similarly, 24 hours after IR, REH sh-PKR cells also displayed significantly more rapid resolution of Comet tails compared with control cells (Figure 5C-D). In addition, Lin⁻ BM cells from TgPKR mice had a significant delay in DSB repair, whereas Lin⁻ BM cells from PKRKO mice demonstrated faster kinetics of DSB repair after IR compared with WT cells (Figure 5E). Furthermore, treatment of TgPKR cells with PKRI promoted an increased efficiency of DSB repair similar to the level of PKRKO cells (Figure 5E). These data indicate that inhibition of PKR expression/activity can promote more rapid DSB repair following IR.

To determine whether increased PKR may accelerate leukemia progression, we crossed TgPKR or PKRKO mice with the NHD13 mouse model of leukemia to produce NHD13-TgPKR and NHD13-PKRKO mice.^3,33,34  The highly penetrant NHD13 model simulates high-risk human MDS that evolves to acute leukemia in ∼60% of mice by 14 months.^33,34  Our laboratory and others have reported that leukemic transformation of NHD13 mice is accompanied by progressive genomic instability and the accumulation of collaborating oncogenic mutations.^35,36  NHD13, NHD13-TgPKR, and NHD13-PKRKO mice were aged until death or physical deterioration requiring euthanasia. Acute leukemia was identified by high PB white blood count (≥20 × 10³/μL), ≥20% BM blasts, increased BM cellularity with loss of differentiation, and a clonal population of immature cells detected by flow cytometry (supplemental Figure 4A).

The OS of NHD13-TgPKR mice were significantly shorter, whereas the survival of NHD13-PKRKO mice were significantly longer than that of NHD13 mice (Figure 6A). Furthermore, NHD13-TgPKR mice had a higher incidence of leukemia than the NHD13 mice (Figure 6B). Conversely, KO of PKR delayed but did not prevent leukemia development (Figure 6A-B). In addition, endogenous PKR expression was significantly increased (RQ >2) with age in NHD13 mice that developed acute leukemia and had shortened survival (Figure 6C and supplemental Figure 4B). When BM of mice was examined at a time prior to the development of acute leukemia, NHD13 mice exhibited an age-associated increase in BM blasts that was further elevated by the PKR transgene (Figure 6D). In contrast, at 6 months of age, the NHD13-PKRKO mice had significantly reduced BM blasts and increased colony-forming unit (CFU)-granulocyte, erythrocyte, monocyte, megakaryocyte (GEMM) frequency compared with NHD13 mice (Figure 6E). Consequently, PKRKO mice demonstrated increased PB cell counts and reduced inhibition of myeloid differentiation (supplemental Figure 4C-K). Furthermore, endogenous PKR expression in Lin⁻ BM cells of NHD13 mice was inversely proportional to p-ATM and γ-H2AX after IR (Figure 6F and supplemental Figure 4L). Collectively, these results indicate that increased PKR expression may cooperate with the NHD13 oncogene to accelerate the accumulation of mutations that lead to more rapid MDS evolution to acute leukemia.

Because PKR inhibition promoted DDR signaling and DSB repair in vitro, we tested whether PKR may affect the frequency of somatic mutations in vivo. Mutation frequency was measured using the PIG-A assay that detects loss of CD24 on murine reticulocytes and is representative of spontaneous mutations that occur in hematopoietic stem and progenitor cells.^35,37,38 

At all ages examined, reticulocytes from TgPKR mice had a significantly increased PIG-A mutation frequency compared with WT mice (Figure 7A). Furthermore, the age-associated increase in mutation frequency from 3 to 12 months was enhanced by TgPKR but significantly reduced by KO of PKR (Figure 7A). Similarly, PB cells of NHD13-TgPKR mice demonstrated a significantly higher age-associated increase of the PIG-A mutation frequency, whereas those of NHD13-PKRKO mice were significantly reduced compared with NHD13 mice (Figure 7B). In addition, reticulocytes of TgPKR mice had a significantly greater mutation frequency than WT mice at Week 1 after IR and did not recover to the low mutation frequency observed in WT mice at Week 4 after IR (Figure 7C). In contrast, PKRKO mice displayed a reduced mutation frequency compared with WT mice following IR (Figure 7C). Thus, increased PKR expression promoted a mutator phenotype, whereas KO of PKR expression reduced the frequency of mutations in vivo.

Next, we tested whether in vivo pharmacologic inhibition of PKR can reduce the mutation frequency in hematopoietic cells of mice. The administration of PKRI to mice inhibited PKR activity, but not total PKR expression or activity of the related kinase PERK, in Lin⁻ BM cells (supplemental Figure 5A-C). Significantly, PKRI treatment reduced the PIG-A mutation frequency in PB reticulocytes from IR-treated mice (Figure 7D). Furthermore, when PKRI was administered continuously to NHD13 mice for 28 days, the mutation frequency in PB reticulocytes were reduced compared with vehicle-treated mice, and BM from PKRI-treated NHD13 mice had significantly improved CFU-GEMM activity (Figure 7E-F). In addition, PKRI treatment of NHD13 mice promoted a significant improvement in PB cell counts and animal weight (supplemental Figure 5D-F and data not shown). Thus, pharmacologic inhibition of PKR activity seemingly prevents the accumulation of potentially deleterious mutations in the NHD13 model.

Our results demonstrate that nuclear PKR has an unexpected pro-oncogenic function to inhibit DDR signaling and DSB repair that can lead to the accumulation of age-associated or irradiation-induced somatic mutations. PKR expression was inversely proportional to both ATM autophosphorylation/activation and phosphorylation of ATM downstream targets following IR in primary CD34⁺ cells, Lin⁻ BM cells from mouse models, and established leukemia cell lines. Furthermore, inhibition of PKR expression or activity in all of these cells promoted a more rapid DDR signaling and DSB repair following IR. Consistent with these observations, TgPKR but not PKRKO mice demonstrated a mutator phenotype characterized by increased radiation-induced or age-associated mutation frequency in hematopoietic cells. Furthermore, the accumulation of somatic mutations that occurs in the NHD13 mouse model of MDS/acute leukemia upon aging was significantly elevated by co-expression of the PKR transgene, whereas KO of PKR expression or pharmacologic inhibition of PKR activity reduced the frequency of spontaneous mutations in vivo. Thus, PKR cooperated with the NHD13 transgene to accelerate leukemia progression leading to shortened survival. Significantly, high PKR expression was associated with poor OS and shortened remission duration for AML patients, and correlated with worse survival in breast, lung, and ovarian cancer patients (supplemental Figure 5).^13-15,39-41  Although it is unclear how PKR may be upregulated in cancer patients and mice, chronic inflammatory stress may contribute since PKR has been well characterized as an IFN-inducible gene. Significantly, findings here suggest that patients with relatively high PKR expression may have inferior clinical outcomes due to the cooperating oncogenic role of PKR to promote the accumulation of potentially deleterious mutations and accelerate leukemogenesis. These findings suggest that PKR inhibition may represent a novel therapeutic strategy to prevent/delay leukemia progression and potentially tumorigenesis of other cancers.

Our findings demonstrate that PKR has an important and previously unrecognized nuclear function to antagonize ATM activation. Significantly, ATM and PKR co-precipitated in nuclear extracts, and PKR activity was required for this interaction. Furthermore, the PKR-ATM association was lost upon irradiation. Our findings suggest that PKR antagonizes ATM function by activating PP2A. Other reports suggest that PP2A maintains ATM in a catalytically quiescent state in undamaged cells.^28,42  Significantly, we observed that PKR inhibition reduced nuclear PP2A activity and nuclear localization of the PP2A B55α regulatory subunit. Whether B55α is a direct downstream target of PKR remains undetermined. In addition, knockdown of PKR resulted in reduced co-precipitation of ATM with PP2A, implying that PKR may regulate PP2A-ATM interaction. Thus, we propose that PKR promotes nuclear B55α localization necessary for nuclear PP2A activity and PP2A-ATM association, which functionally suppresses ATM autophosphorylation in undamaged cells. In response to genotoxic stress including IR, PP2A and ATM rapidly dissociate, allowing ATM autophosphorylation and initiation of the DDR signaling cascade (Figure 4I). These results have now prompted us to reinterpret previous findings that inhibition of PKR promotes cell survival following treatment with genotoxic agents, such as doxorubicin, exclusively due to decreased apoptosis. We now propose that inhibition of PKR expression/activity may also promote increased cell survival due to an improved DNA repair capacity that results in a more effective response to genotoxic stress. Significantly, these results suggest that increased PKR expression/activity, such as observed in acute leukemia and other cancers, may cooperate with other more potent oncogenes to accelerate tumorigenesis.

The authors thank Dr Ying Li for flow cytometry assistance and Dr Christopher Carter for immunohistochemistry interpretation.

This study was supported by the National Institutes of Health (NIH) National Heart, Lung, and Blood Institute (R01 HL054083), the Florida Department of Health Bankhead-Coley award (09BW-06), a University of Florida Health Cancer Center Team Science award, and University of Florida Gatorade funding. X.C. was supported in part by the NIH National Institute of Diabetes and Digestive and Kidney Diseases (T32 DK074367). M.B. was supported in part by the NIH National Center for Research Resources Clinical and Translational Science Awards grant (UL1 TR000064).

Contribution: M.B. and X.C. performed research, analyzed data, and wrote the manuscript; K.D.B. designed experiments and analyzed results; M.Y.K. provided key reagents; S.M.K. performed experiments and analyzed results; and R.L.B. and W.S.M. designed experiments, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: W. Stratford May, Division of Hematology and Oncology, Department of Medicine, University of Florida, 2033 Mowry Rd, Box 103633, Gainesville, FL 32610; e-mail: smay@ufl.edu.