Gary J. Vanasse, MD, Division of Haematology, Brigham and Women’s Hospital, Harvard Medical School, 75 Francis Street, Karp 5.216, Boston, MA 02115, USA. E-mail: email@example.com
Overexpression of pro-inflammatory cytokines, including tumour necrosis factor alpha (TNFα), has been implicated in the pathogenesis of anaemia of inflammation. TNFα suppresses erythroid colony formation via both direct and indirect effects on haematopoietic progenitors, often involving activation of nuclear factor (NF)-κB signalling resulting in downregulation of transcription factors critical for erythropoiesis. There is a dearth of effective and safe therapies for many patients with inflammatory anaemia. Resveratrol is a flavanol found in red wine grapes that possesses potent anti-inflammatory properties, but studies of its impact on human erythropoiesis have proven contradictory. We investigated whether resveratrol ameliorates TNFα-mediated suppression of erythropoiesis in human CD34+ haematopoietic progenitors. We found that resveratrol partially reverses the erythroid suppressive effects of TNFα, leading to significant recovery in burst forming unit-erythroid colony formation in human CD34+ cells. CD34+ cells pre-incubated with resveratrol for 72 h in the presence of TNFα inhibited NF-κB activation via decreased NF-κB nuclear localization without altering total NF-κB protein levels and independent of IκB degradation. Resveratrol also significantly restored the baseline expression of erythroid transcription factors NFE2 and the GATA1/GATA2 ratio in CD34+ cells treated with TNFα. In conclusion, resveratrol may inhibit TNFα-mediated NF-κB activation and promote erythropoiesis in primary human CD34+ cells.
Anaemia of inflammation (AI) has historically been termed the ‘anaemia of chronic disease’ and is most commonly seen in association with infection, rheumatologic disorders, malignancy and other chronic illnesses. On a biochemical level it is classically characterized by low serum iron and low iron binding capacity in the setting of an elevated serum ferritin. Classical AI has been attributed to a combination of decreased red cell survival, disordered iron-limited erythropoiesis and progressive erythropoietin (EPO) resistance of erythroid progenitors, but the relative role and interplay of these three mechanisms in the development of anaemia remains unknown, as are the potential common pathways that may link them. Our understanding of AI has been transformed by the discovery of the antimicrobial peptide hepcidin (Andrews, 2004). Hepcidin is potently induced by the inflammatory mediator interleukin (IL)-6, and this acute phase reactant is hypothesized to be responsible for mediating iron-limited erythropoiesis in patients with acute and chronic inflammatory states (Ganz, 2003). However, recent studies have failed to show a correlation between IL-6 and urinary hepcidin (Ferrucci et al, 2010), suggesting that increased hepcidin synthesis occurs only in situations of overt inflammation or that anaemia may also be mediated through hepcidin-independent inflammatory pathways.
Consistent with this latter postulate, tumour necrosis factor alpha (TNFα), as well as other pro-inflammatory cytokines (Roubenoff et al, 1998; Ershler & Keller, 2000; Bruunsgaard, 2002; Bruunsgaard & Pedersen, 2003; Krabbe et al, 2004; Ferrucci et al, 2005), may directly suppress erythropoiesis by signalling through TNFα receptors expressed on haematopoietic progenitors (Rusten & Jacobsen, 1995). Although TNFα has been demonstrated to inhibit erythroid differentiation of K562, HEL and TF1 cells through activation of p38 mitogen-activated protein kinase (MAPK) signalling pathways and regulation of erythroid specific transcriptional factors, including GATA binding factor-1 (GATA1), GATA binding factor-2 (GATA2) and zinc finger protein, multi-type 1 [ZFPM1, previously termed friend of GATA1 (FOG1)] (Buck et al, 2008), a detailed understanding of the mechanisms by which TNFα suppresses erythroid differentiation and proliferation of primary human haematopoietic progenitors remains elusive.
Reflecting the marked biological and clinical heterogeneity found among patients with AI, there are presently few effective and safe therapies for this large patient population. Resveratrol, a phytoalexin polyphenolic compound found in red wine, berries and peanuts, has been implicated in a wide range of biological and pharmacological activities and is felt to possess potent anti-inflammatory properties (de la Lastra & Villegas, 2005). Resveratrol has been demonstrated to decrease IL-6 expression (Kang et al, 2009), reduce vascular inflammation (Smolen et al, 2007), increase circulating endothelial progenitor cells (Balestrieri et al, 2008) and reverse inflammatory anaemia in a rat model of colitis (Larrosa et al, 2009). Resveratrol has been shown to inhibit TNFα-mediated p38 MAPK activation in endothelial cells and c-Jun NH2-terminal kinase (JNK) and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) signalling in myeloid (U-937), lymphoid (Jurkat) and epithelial (HeLa and H4) cells (Manna et al, 2000; Balestrieri et al, 2008). However, studies examining resveratrol’s effects on human erythropoiesis have proven contradictory. Although resveratrol induces fetal haemoglobin synthesis in the K562 erythroleukaemia cell line and in patients with sickle cell disease (Rodrigue et al, 2001), moderate doses were found to inhibit erythroid colony formation in normal human CD34+ cells (Ferry-Dumazet et al, 2002). Its ability to reverse TNFα-mediated erythroid colony suppression in primary human haematopoietic progenitor cells remains to be examined.
In the present study, we sought to determine whether lower doses of resveratrol ameliorate TNFα-mediated suppression of erythropoiesis in human CD34+ haematopoietic progenitors. Our results indicate that resveratrol partially reversed the erythroid suppressive effects of TNFα via pathways that decrease NF-κB nuclear localization, resulting in increased expression of erythroid-specific transcription factors and significant recovery of normal erythroid colony formation in human CD34+ cells, warranting future clinical study of resveratrol or derivative compounds for the treatment of patients with AI.
Materials and methods
Cell culture and reagents
Cryopreserved human CD34+ haematopoietic progenitor cells isolated from bone marrow (Lonza, Basel, Switzerland) or mobilized peripheral blood (Yale Center of Molecular Hematology, New Haven, CT, USA) from healthy adult donors aged 30–60 years were cultured for 1 d in StemSpan serum-free media (StemCell Technologies, Vancouver, BC, Canada) supplemented with 10 ng/ml IL-3 (IL-3; PeproTech, Rocky Hill, NJ, USA), 50 ng/ml stem cell factor (SCF; PeproTech) and antibiotics (50 U/ml penicillin, 50 ng/ml streptomycin and 250 ng/ml fungizone; Invitrogen, Carlsbad, CA, USA). The following day (day 0) cells were induced toward the erythroid lineage by treatment with 1 U/ml erythropoietin (EPO; R & D Biosystems, Minneapolis, MN, USA). Cells were treated with varying doses of resveratrol (0–50 μmol/l; Sigma, St. Louis, MO, USA) on days 1–4 and incubated at 37°C in an atmosphere of 5% CO2 for 0–72 h prior to TNFα (PeproTech) treatment on day 4.
Human CD34+ cells were washed at 4°C in phosphate-buffered saline (PBS) containing 1% fetal bovine serum (FBS; Invitrogen) and labelled with the following human monoclonal antibodies: fluorescein isothiocyanate (FITC)-CD34, phycoerythrin (PE)-glycophorin A (GlyA), FITC-CD71, FITC-CD36 and PE-CD45 (all purchased from BD Biosciences, San Diego, CA, USA). Nonviable cells were excluded by counterstaining with 7-AAD (BD Biosciences). Isotype-matched FITC- or PE-conjugated antibodies were used as controls. Cells were analysed on a FACScaliber cytometer (BD Biosciences) using Cellquest Pro Software (BD Biosciences).
Erythroid progenitor cell assays
Erythroid colony formation in CD34+ cells was analysed using a methylcellulose colony-forming assay (StemCell Technologies). On day 4, an equal number of cells (1000 cells/dish) was plated in triplicate in 35-mm dishes (StemCell Technologies) in a semisolid methylcellulose medium consisting of 80% methocult H4230 media (StemCell Technologies), 20% IMDM (Invitrogen) + 2% FBS (Invitrogen) and supplemented with cytokines (10 ng/ml IL-3, 50 ng/ml SCF and 1 U/ml EPO) and antibiotics in the presence or absence of 0–50 μmol/l resveratrol, 10 μg/ml infliximab and/or 25 ng/ml TNFα and burst-forming unit-erythroid (BFU-E) colonies scored using an inverted microscope (Nikon, Tokyo, Japan) on day + 14.
RNA isolation and cDNA synthesis
Total RNA was isolated from human CD34+ cells at different time points in liquid culture under conditions described above. Cells were lysed in Trizol (Invitrogen) and RNA was separated from solution by chloroform addition. Total RNA was isolated using the mirVana miR Isolation kit (Invitrogen) according to the manufacturer’s instructions. cDNA was synthesized from 2 μg of total RNA using the RETROscript first strand synthesis kit for RT-PCR (Invitrogen) according to the manufacturer’s instructions.
Quantitative real-time PCR (qRT-PCR)
qRT-PCR was performed using SYBR Green (BioRad, Hercules, CA, USA) in triplicate and analysed on a Bio-Rad iQ5 machine with Bio-Rad iQ5 software using β-actin gene (ACTB) expression as a normalization control. The primer sets for each gene are as follows:
Nuclear and cytoplasmic extracts were prepared using NE-PER reagents (Pierce, Rockford, IL, USA) according to the manufacturer’s instructions. NF-κB activity was assayed by using the NF-κB EZ-TFA transcription factor assay kit (Millipore, Billerica, MA, USA), a 96-well format non-radioactive transcription factor assay system. Briefly, nuclear extracts were incubated with the biotinylated oligonucleotide capture probe containing the consensus sequence for NFKB1 (5′-GGGACTTTCC) in streptavidin-coated wells. The immobilized NF-κB was probed by sequential incubation with antibody recognizing NF-κB p65 or NF-κB p50 subunits and a horseradish peroxidase-conjugated secondary antibody. Chemiluminescent signals were read in the DTX880 Multimode Detector (Beckman Coulter, Brea, CA, USA). The signals were normalized with amounts of nuclear protein used in each assay.
Whole cell lysates or nuclear and cytoplasmic extracts were prepared with 2× sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer containing Protease Inhibitor Cocktail Set III (CalBiochem/EMD Chemicals, Gibbstown, NJ, USA) and Phosphatase Inhibitor Cocktail Set III (CalBiochem) and boiling for 5 min. Equal protein amounts were subjected to SDS-PAGE and analysed by immunoblotting using antibodies specific for phospho-NF-κB p65 (Cell Signaling Technology, Danvers, MA, USA), total NF-κB p65 (Santa Cruz Biotechnology, Santa Cruz, CA, USA), total NF-κB p105/50, total IκB, phospho-p38, total p38, phospho-JNK1/2/3, total JNK1/2/3, phospho-ERK1/2 or total ERK1/2 (all from Cell Signaling Technology). Anti-lamin A/C was used for nuclear loading control and anti-GAPDH was used for cytoplasmic loading control (both from Cell Signaling Technology). Band intensities were quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA).
Data are presented as mean ± standard error (SE). Differences between groups were analysed using the Student’s t-test. The level of significance was set at P value < 0·05.
Resveratrol significantly reverses TNFα-mediated erythroid colony suppression in human CD34+ cells
Human CD34+ cells from four healthy donors (two bone marrow-derived and two peripheral blood mobilized, ≥98% CD45+/CD34+ by flow cytometry) were incubated in serum-free liquid culture medium containing 50 ng/ml SCF, 10 ng/ml IL-3 and 1 U EPO for 72 h, plated in methylcellulose media supplemented with similar concentrations of SCF, IL-3 and EPO in the presence of varying concentrations of TNFα (0–100 ng/ml) and BFU-E colonies enumerated on day + 14. At the time of plating in methylcellulose, ≥95% of cells were found by fluorescence-activated cell sorting (FACS) analysis to express CD45+/CD34+/CD71+medium/CD36−/GlyA− (data not shown), consistent with an early erythroid progenitor population (Okumura et al, 1992). At day + 14, the number of BFU-E colonies was decreased by 75% in the presence of 25 ng/ml or 100 ng/ml TNFα when compared to control (P < 0·01, t-test), and this effect was reversed to baseline with addition of the neutralizing anti-TNFα antibody infliximab (P < 0·01, t-test), suggesting that TNFα-specific pathways were responsible for quantitative reduction in the number of BFU-E colonies (Fig 1A). TNFα also appeared to disrupt the normal BFU-E architecture, with TNFα-treated colonies appearing smaller, more dispersed and composed of fewer erythroblasts (Fig 1B).
To determine whether resveratrol could inhibit TNFα-mediated erythroid colony suppression, human CD34+ cells were next grown in liquid culture as above but were pre-incubated in the presence of varying concentrations of resveratrol (0–50 μmol/l) for 72 h and then plated in methylcellulose media containing similar concentrations of SCF, IL-3 and EPO in the presence or absence of 25 ng/ml TNFα. Pre-incubation with resveratrol did not alter the phenotype of CD34+ cells, as FACS analysis again revealed ≥95% of cells expressed CD45+/CD34+/CD36−/CD71+medium/GlyA− (data not shown). Whereas TNFα treatment resulted in a 59% decrease in BFU-E colonies on day + 14, pre-incubation with both 2 and 10 μmol/l resveratrol significantly restored BFU-E colony formation by 41% and 51%, respectively (P < 0·02 for each dose of resveratrol compared to TNFα alone, t-test) (Fig 1C). Fifty micromolar resveratrol was found to be toxic to BFU-E colony formation, resulting in >90% reduction in BFU-E colonies. These data indicate that lower concentrations of resveratrol are effective at ameliorating TNFα-mediated suppression of normal human erythropoiesis.
Resveratrol reverses TNFα-mediated changes in erythroid transcription factor expression
To determine whether resveratrol may inhibit the ability of TNFα to suppress erythroid-specific transcription factors critical for early erythroid development in human CD34+ cells, we next investigated its possible effect on the expression levels of erythroid specific nuclear factor erythroid derived 2 (NFE2), GATA1, GATA2, ZFPM1 and Erythroid Krüppel-like Factor (KLF1) in the presence of TNFα. CD34+ cells from the same four individual donors were grown in liquid culture as above with and without 72 h pre-incubation with 10 μmol/l resveratrol and in the presence or absence of 25 ng/ml TNFα for an additional 12 h on culture day + 4. As expected, TNFα treatment was found to decrease the expression levels of NFE2, ZFPM1, KLF1, and the ratio of GATA1 to GATA2 by 20–35%. However, pre-incubation with resveratrol significantly restored the expression of NFE2 (P < 0·02, t-test), ZFPM1 (P < 0·05, t-test), KLF1 (P < 0·05, t-test) and the ratio of GATA1 to GATA2 (P < 0·01, t-test) to baseline levels (Fig 2A–D). Whereas resveratrol restored baseline expression of NFE2 and the GATA1/GATA2 ratio in the presence of TNFα treatment, increased ZFPM1 and KLF1 expression appeared independent of TNFα treatment. Pre-incubation with resveratrol was also found not to alter the expression of endogenous EPO receptors on either CD34+ cells or later-stage CD34− erythroid precursors (Figure S1).
Numerous studies implicate induction of NF-κB as an important signalling intermediary in TNFα-mediated suppression of erythropoiesis, with activation of the canonical NF-κB p50/p65 pathway contributing to inhibition of erythroid-specific gene expression in globin genes (La Ferla et al, 2002; Imagawa et al, 2003; Liu et al, 2003). To determine whether resveratrol altered TNFα-induced NF-κB activation in primary human cells, an initial timecourse experiment was performed using peripheral blood mobilized CD34+ cells from one individual donor. CD34+ cells were pre-incubated with 10 μmol/l resveratrol over the course of 0–72 h in serum-free liquid culture medium containing 50 ng/ml SCF, 10 ng/ml IL-3 and 1 U EPO, treated with 5 ng/ml TNFα for 30 min on culture day 4, and nuclear localization of NF-κB activity determined. Based on our combined studies incorporating both liquid and semi-solid medium cultures, a TNFα dose of 5 ng/ml rather than 25 ng/ml was used for these studies so as to avoid a saturating concentration of TNFα. The oligonucleotide binding activities of both the p65 (Fig 3A) and p50 subunits of NF-κB (Fig 3B) were markedly increased by TNFα treatment and binding was completely blocked by co-incubation with competitor oligonucleotide, confirming the specificity of the assay. Pre-incubation with resveratrol for 1–24 h had little effect on NF-κB activation. However, significant inhibition of TNFα-induced NF-κB activation was observed with pre-incubation with resveratrol for 48 h (32% and 51% inhibition for p65 and p50, respectively) and 72 h (43% and 72% inhibition for p65 and p50, respectively) (Fig 3A, B).
Next, to investigate whether resveratrol inhibited TNFα-mediated NF-κB activation in a dose-dependent manner, CD34+ cells from this same donor were pre-incubated for 72 h with increasing concentrations of resveratrol in liquid culture as above and treated with 5 ng/ml TNFα for 30 min on culture day 4. Whereas 0·4 μmol/l resveratrol had little effect on activated NF-κB levels in response to TNFα, resveratrol concentrations of 2, 10 and 50 μmol/l exhibited 11%, 36% and 64% inhibition of TNFα-mediated NF-κB activation, respectively (Fig 3C). Finally, in combined analyses with optimized conditions using CD34+ cells from four individual donors (two bone marrow-derived and two peripheral blood mobilized), we found that 72 h pre-incubation with 10 μmol/l resveratrol exhibited 46% inhibition of TNFα-mediated NF-κB activation (P = 0·004, t-test) (Fig 3D). As noted before, 50 μmol/l resveratrol was found to be toxic to BFU-E colony formation in CD34+ cell progenitor assays.
Immunoblotting of cytoplasmic and nuclear fractions and probing for p65 and p50 revealed that the decreased NF-κB activity we observed was due to inhibition of nuclear translocation of NF-κB by resveratrol. Paralleling the results of the oligonucleotide binding assay, nuclear translocation of p65 and p50 was unaffected by pre-incubation with resveratrol for 1–24 h but was decreased in the setting of either 48 or 72 h pre-incubation with resveratrol (Fig 4A), with 72 h pre-incubation with 10 μmol/l resveratrol exhibiting 54% inhibition of TNFα-mediated NF-κB activation (P = 0·004, t-test) (Fig 4B). We then examined whether resveratrol inhibited the nuclear translocation of NF-κB by regulating phosphorylation of p65 and/or stabilization of IκB, which is known to sequester NF-κB in the cytoplasm. However, pre-incubation with 10 μmol/l resveratrol for 72 h failed to alter p65 phosphorylation or IκB degradation mediated by varying concentrations of TNFα (25, 5 and 1 ng/ml) and did not modulate total NF-κB protein levels (Fig 4C). We also found that resveratrol exhibited no significant effects on other TNFα-mediated signalling pathways, including phosphorylation of p38, ERK-1/2 or JNK under similar conditions (Fig 4C). In summary, the mechanism by which resveratrol blocks translocation of NF-κB remains to be elucidated.
Accumulating evidence suggests that resveratrol, a flavanol found in high concentrations in red wine, may possess important anti-inflammatory properties, partially explaining its link to the ‘French paradox’ that the French population have significantly decreased rates of heart disease and obesity in the face of markedly higher dietary saturated fat intake. Studies examining resveratrol’s effects on normal human erythroid colony development are limited. Resveratrol concentrations of 50 μmol/l or greater have been shown to potently induce haemoglobin and fetal haemoglobin (Rodrigue et al, 2001) and to stimulate apoptosis in human leukaemia cell lines (Clement et al, 1998; Ferry-Dumazet et al, 2002; Liu et al, 2010). However, this appears to come at the expense of toxic effects on normal erythropoiesis. To our knowledge, our results demonstrate for the first time that resveratrol, when used at concentrations below those that appear to mediate its anti-proliferative effects, is non-toxic and capable of promoting significant restoration of erythroid colony formation in normal human CD34+ cells exposed to TNFα. Consistent with prior studies (Ferry-Dumazet et al, 2002), we found that the 50 μmol/l dose of resveratrol was toxic to the formation of BFU-E colonies regardless of the presence or absence of TNFα. Our results also suggest that resveratrol does not directly modulate EPO responsiveness of haematopoietic progenitors, as expression of endogenous EPO receptors remained unchanged in the presence or absence of resveratrol during the course of erythroid differentiation. Although the sample size of our study is small, the results were highly reproducible with little variability.
TNFα indirectly inhibits erythropoiesis primarily through upregulation of transcription factors NF-κB and GATA2, resulting in subsequent downregulation of GATA1 and depressed EPO signalling (La Ferla et al, 2002; Imagawa et al, 2003). NF-κB is active during the early stages of normal erythroid development (Zhang et al, 1998), and is responsible for the suppression of a number of transcription factors critical for normal erythropoiesis, including MYB, MYC (Zhang et al, 1998) and NFE2 (Labbaye et al, 1995; Shivdasani & Orkin, 1995; Liu et al, 2003; Zhou et al, 2010). Overexpression of the p65 NF-κB subunit in K562 cells blocks their maturation in response to decreased NFE2 levels and NFE2/DNA complex formation (Chenais, 1998; Liu et al, 2003). Our data suggest that resveratrol’s inhibitory effects on TNFα in early erythroid precursors derived from human CD34+ cells are mediated at least in part by modulation of NF-κB activity, leading to increased NFE2 levels and subsequent recovery of erythroid colony forming capacity. We note that resveratrol appears to modulate NF-κB activity in human CD34+ cells via novel mechanisms compared with known common NF-κB inhibitors in other cell types, which typically involve sequestration of the NF-κB subunit by a complex consisting of IκB proteins. The polyphenol chrysin was reported to inhibit NF-κB activity through inhibition of IκBα phosphorylation in human intestinal Caco-2 cells (Romier et al, 2008). In contrast, resveratrol-treated U-937 cells revealed no change in phospho- and total IκBα but did result in decreased phosphorylation of the p65 subunit of NFκB (Manna et al, 2000). Although our data strongly indicate that resveratrol inhibits the NF-κB pathway specifically through prevention of the nuclear translocation of both the NF-κB p50 and p65 subunits in human CD34+ cells, we detected no change in either IκBα levels or phospho-NF-κB levels in response to resveratrol. Therefore, further investigations are required to identify the mechanism by which resveratrol inhibits NF-κB nuclear translocation in primary human erythroid precursors. Furthermore, contrary to studies in other cell types, resveratrol did not appear to block the erythroid suppressive effects of TNFα via modulation of ERK, JNK, or p38 cellular pathways in human CD34+ cells.
Direct effects of TNFα on erythropoiesis are mediated primarily through TNF receptors I and II. TNFα has been previously demonstrated to inhibit maturation of EPO-induced TF1 cells through inhibition of GATA1 and ZFPM1 (FOG-1) expression and through an increase in GATA2 expression (Buck et al, 2008). Studies using GATA2 knock-out cell lines revealed that GATA2 is required for early erythropoiesis but has negligible influence over late erythroid maturation (Tsai & Orkin, 1997). In contrast, GATA1 in association with ZFPM1 is required for terminal erythroid differentiation (Tsang et al, 1997; Weiss et al, 1997). The ratio of GATA1 and GATA2 is critical for normal erythroid maturation due in part to regulation of globin gene expression (Labbaye et al, 1995; Ikonomi et al, 2000). Our results reveal that in addition to reversing TNFα-induced downregulation of NFE2, resveratrol also results in normalization of the ratio of GATA1 to GATA2, primarily by reversing TNFα-mediated suppression of GATA1 without having appreciable effects on GATA2 expression. Although we found that resveratrol treatment also increased the levels of both ZFPM1 and KLF1, a transcription factor involved in HBB (β-globin gene) regulation (Wijgerde et al, 1996; Tewari et al, 1998), these changes occurred independent of TNFα stimulation. Taken together, resveratrol treatment appears to promote a recovery in the baseline expression levels of erythroid-specific transcription factors in a population of early erythroid precursors.
In summary, our study suggests that a lower dose of resveratrol ameliorates TNFα-mediated suppression of erythroid colony formation in human CD34+ haematopoietic progenitor cells via inhibition of NF-κB-mediated inflammatory pathways and regulation of specific erythroid transcription factors that together promote recovery of erythroid colony-forming capacity in early erythroid precursors. TNFα is a prime mediator of inflammatory-associated anaemia, particularly in patients with rheumatoid arthritis (RA). Although TNFα-modulating therapies are effective in alleviating anaemia in a percentage of patients with RA, toxicities have prevented their widespread use in other patient cohorts. Our preclinical data suggests that resveratrol or derivative compounds warrant further preclinical and clinical study to determine whether they may provide effective adjunctive therapy for subgroups of patients with inflammatory anaemia.
Both Jee-Yeong Jeong and Matthew Silver contributed equally as first authors in the writing of this manuscript. Jee-Yeong Jeong performed the research study, analysed the data and wrote the paper. Matthew Silver performed the research study, analysed the data and wrote the paper. Aric Parnes performed the research study. Sarah Nikiforow performed the research study and analysed the data. Nancy Berliner analysed the data and assisted in writing the manuscript. Gary Vanasse oversaw all aspects of the project and designed the research study, analysed the data and wrote the manuscript.
This work was supported by National Institutes of Health Grant RO1AG29154 (N.B.).