Central nervous system
Although deposition of amyloid β-peptide (Aβ) as Aβ plaques involves activation of microglia-mediated inflammatory responses, activated microglia ultimately fail to clear Aβ plaques in the brains of either Alzheimer's disease (AD) patients or AD mouse models. Mounting evidence suggests that chronic microglia-mediated immune response during Aβ deposition etiologically contributes to AD pathogenesis by promoting Aβ plaque formation. However, the mechanisms that govern microglia response in the context of cerebral Aβ/β-amyloid pathology are not well understood. We show that ligation of CD40 by CD40L modulates Aβ-induced innate immune responses in microglia, including decreased microglia phagocytosis of exogenous Aβ1–42 and increased production of pro-inflammatory cytokines. CD40 ligation in the presence of Aβ1–42 leads to adaptive activation of microglia, as evidenced by increased co-localization of MHC class II with Aβ. To assess their antigen-presenting cell (APC) function, cultured microglia were pulsed with Aβ1–42 in the presence of CD40L and co-cultured with CD4+ T cells. Under these conditions, microglia stimulate T cell-derived IFN-γ and IL-2 production, suggesting that CD40 signaling promotes the APC phenotype. These data provide a mechanistic explanation for our previous work showing decreased microgliosis associated with diminished cerebral Aβ/β-amyloid pathology when blocking CD40 signaling in transgenic Alzheimer's mice.
Alzheimer's disease (AD) is a neurodegenerative disease with primary neuropathological hallmarks of intracellular neurofibrillary tangles and extracellular senile plaques comprised primarily of amyloid β-peptide (Aβ). Additionally, the AD brain is also characterized by marked microglia-mediated immune responses, the role of which has become potentially important, yet unclear. On the one hand, activated microglia have been implicated in AD pathogenesis by promoting Aβ plaque formation as well as pro-inflammatory cytokine production 1, 2; on the other hand, studies have shown that microglia are involved in phagocytic Aβ plaque clearance 3, 4. Microglia, as resident macrophages of the brain, form the first line of defense against invading pathogens and provide a key link between the central nervous system (CNS) and the immune system. In the normal adult brain, microglia are relatively quiescent, but in response to CNS injury they actively phagocytose cellular debris and dying cells 5. This phagocytic uptake of apoptotic cells by microglia results in reduced pro-inflammatory cytokine production and minimizes damage in the inflamed brain 6.
In both AD and animal models thereof, there is increased activation and recruitment of microglia to areas of cerebral amyloidosis, but these activated microglia fail to clear Aβ deposits in Alzheimer's mice 7. However, the employment of therapeutic strategies that boost microglia phagocytic activity is able to reduce cerebral Aβ load in mouse models of AD. This is exemplified by a report showing that transforming growth factor-β overexpression promotes phagocytic clearance of cerebral Aβ 8. Similarly, microglia Fc-receptor-mediated removal of Aβ is proposed as one mechanism by which Schenk and colleagues’ 9 peripheral immunization of the PDAPP mouse model of AD with Aβ1-42 resulted in reduced Aβ plaque burden. Interestingly, macrophage colony-stimulating factor (M-CSF) has also been shown to promote microglia phagocytic removal of Aβ, and osteopetrotic mice which lack functional M-CSF develop CNS fibrillar plaques 10. Taken together, it seems that microglia phagocytic ability can be mobilized to promote Aβ removal but endogenous mechanisms fail to effectively initiate or maintain this process. We have previously identified that microglia CD40 receptor (CD40) potentiates Aβ-induced tumor necrosis factor (TNF)-α production, and subsequent in vivo studies demonstrated that blockage of CD40 signaling – either by antibody depletion of the ligand for CD40 (CD40L) or by ablating the CD40L gene – was able to reduce cerebral Aβ/β-amyloid load in an AD mouse model 11, 12.
CD40 is a ∼45–50-kDa cell surface molecule and a member of the TNF-α/nerve growth factor receptor superfamily 13. A variety of professional and non-professional antigen-presenting cells (APC) express CD40, including dendritic cells, B cells, monocytes/tissue macrophages, and microglial cells 13. Its cognate ligand, CD40L, is typically found on the surface of activated T cells, but it is also expressed in the CNS by astrocytes. Moreover, the expression of CD40 and CD40L has been found in and around Aβ plaques in the AD brain 14, 15. Effects of CD40 ligation on maturation, activation, and survival of dendritic cells suggest that the CD40-CD40L interaction plays an important role in APC-mediated immune responsiveness 16, 17.
Increasing evidence in AD mouse models suggests that there are multiple forms of microglia activation, some manifestations of which may act to promote Aβ/β-amyloid pathology, while other forms may act to remove this pathology. We hypothesize that two types of activated microglia exist in the context of Aβ; specifically, a phagocytic phenotype and an APC phenotype. Promotion of the phagocytic phenotype would be associated with anti-inflammatory cytokine production and would inhibit Aβ plaque formation, while the APC phenotype, given the rare presence of T cells in the AD brain, would essentially result in unchecked production of APC-associated pro-inflammatory cytokines that would act to exacerbate AD-like pathology. Given their pleiotrophic presentation once activated, we suggest that microglia phenotypes are determined in part by their local brain milieu. Thus, in the context of the evolving Aβ plaques of AD, we hypothesize that engagement of microglia CD40 represents one mechanism to shift activated microglia response away from phagocytic and towards antigen presentation, thus rendering these cells ineffective at Aβ removal. To test these hypotheses, we have employed primary mouse microglia to investigate the role of CD40 signaling in phagocytosis, antigen presentation, and cytokine response of microglia to Aβ.
The CD40-CD40L interaction retards microglia phagocytosis and clearance of Aβ1–42
Microglia phagocytosis of Aβ has been deemed a principle mechanism governing the removal of Aβ from the brain parenchyma 18, and the CD40-CD40L interaction is a critical modulator of immune cell activation. We wished to evaluate whether CD40 ligation could inhibit microglia uptake of Aβ. Thus, we added “aged” FITC-tagged Aβ1–42 to primary cultured microglia for 15, 30, and 60 min in the absence or presence of anti-Aβ antibody (clone BAM-10 as a positive control), isotype-control IgG, CD40L, or heat-inactivated CD40L. As a control for non-phagocytic incorporation of Aβ by microglia, microglia were incubated at 4°C in parallel cell culture plates under the same treatment conditions described above. Cell supernatants and lysates were analyzed for extracellular and cell-associated FITC-Aβ using a fluorometer. As shown in Fig. 1, data indicate that the CD40-CD40L interaction results in decreased microglia uptake of Aβ1–42. Consistent with a previous report 4, treatment with the positive control (anti-Aβ antibody) markedly increased microglia uptake of Aβ1–42 at 60 min.
It should be noted that there is not always a direct correlation between microglia phagocytosis and removal of Aβ 19, 20. Therefore, we also determined whether retarded microglia phagocytosis of Aβ following CD40 ligation also produced decreased removal of the peptide. We treated cultured microglia with “aged” Aβ1–42 in the presence or absence of CD40L protein for 1 h. Microglia culture media were collected and immunoprecipitated with anti-Aβ antibody, and data showed that CD40L protein treatment prevented Aβ removal from the culture media relative to the absence of CD40L (Fig. 2a, b). In addition, microglia were extensively washed and lysed for Western blot analysis of Aβ, and densitometry showed that CD40L protein treatment resulted in a threefold reduction in the amount of cell-associated Aβ (Fig. 2c, d). Taken together, these data suggest that CD40 ligation retards the ability of microglia to remove “aged” exogenously added human Aβ1–42.
Microglia-secreted APC-associated cytokines TNF-α, IL-6, and IL-1β are significantly increased in the presence of Aβ following CD40 ligation
We have previously shown that ligation of CD40 leads to the synergistic activation of microglia by Aβ, as evidenced by TNF-α release and bystander neuronal cell injury 11. In addition, it is well known that the CD40-CD40L interaction is critical for promotion of APC differentiation of monocytes, macrophages, and dendritic cells, as evidenced by the production of pro-inflammatory cytokines such as TNF-α, IL-6, and IL-1 16. To investigate whether CD40 ligation promotes microglial APC differentiation in the presence of Aβ1–42, we examined cytokine profiles in the supernatants of microglia co-treated with CD40L protein and Aβ1–42 for 48 h. Data show that co-treatment of microglia with CD40L protein and Aβ1–42 synergistically increased production of APC-associated TNF-α, IL-6, and IL-1β cytokines (Fig. 3).
CD40 signaling-associated Th1, but not Th2, cytokines inhibit microglia phagocytosis of Aβ1–42
It is well known that the CD40-CD40L interaction promotes Th1 cell differentiation as evidenced by increased production of pro-inflammatory cytokines such as IFN-γ by the Th1 cell and TNF-α by the engaged APC 21. In order to examine putative modulation of microglia phagocytosis of Aβ by CD40 signaling-associated Th1 cytokines, microglia (in 24-well tissue culture plates, 1×105 cells/well) were treated with 300 nM of “aged” FITC-tagged Aβ1–42 in the presence or absence of recombinant Th1 cytokines (IFN-γ and/or TNF-α) or Th2 cytokines (IL-4 and/or IL-10 as controls for Th1 cytokines). One hour after treatment, cell culture supernatants were collected and cell lysates were prepared to measure Aβ by fluorometry. In parallel experiments, microglia were treated with 3 μM of “aged” synthetic Aβ1–42, for 1 h, in the presence or absence of the Th1 or Th2 cytokines mentioned above. Cell lysates were then prepared and subjected to Western blot analysis using monoclonal antibody BAM-10, which specifically recognizes human Aβ. Both the fluorometric and immunoblot analysis (Fig. 4) of the microglia-conditioned media and lysates consistently showed that CD40 signaling-associated Th1 cytokines inhibit whereas Th2 cytokines boost microglia uptake of Aβ1–42.
CD40 ligation promotes co-localization of microglia MHC class II and Aβ peptide
Having shown that both CD40 ligation and Th1 cytokines retarded microglia uptake and clearance of Aβ, we wondered whether CD40L might shift microglia activation towards an APC phenotype. MHC class II plays a critical role in loading and transporting of extracellular pathogens and toxins to the APC surface, where they are recognized by CD4+ T cells 22. MHC class II is particularly interesting as its expression levels are markedly increased on microglia associated with Aβ plaques in AD 23, 24. The impairment of MHC class II function results in a significant reduction of microglia-associated CNS inflammation 23, 24, suggesting that MHC class II plays an important role in regulation of microglia immune responses. In addition, effects of CD40 ligation on promotion of the microglia MHC class II-mediated APC processes have been also reported 25–28.
We therefore wished to investigate whether the CD40-CD40L interaction could promote co-localization of MHC class II with Aβ peptide in cultured microglia. We treated microglia with CD40L protein (2 μg/ml) in the presence or absence of “aged” Cy3-Aβ1–42 peptide (50 nM) for 60 min (in pilot studies, we evaluated various time points from 30 to 180 min, and found that the 60-min time point allowed optimal co-localization of MHC class II and Aβ as measured by immunoprecipitation/Western blot and observed by fluorescence microscopy; data not shown). To examine the co-localization of MHC class II and Aβ, we performed immunofluorescence staining on these cells using FITC-conjugated anti-mouse MHC class II antibody.
Results show that CD40 ligation promotes MHC class II-Aβ co-localization as detected by fluorescence microscopy (Fig. 5a, b). To further determine whether co-localization of MHC class II and Aβ peptide was due to fact that they were physically interacting, we treated primary cultured microglia in parallel with CD40L protein (2 μg/ml) in the presence or absence of “aged” Aβ1–42 (3 μM) for 60 min. Cell lysates were then prepared, immunoprecipitated with anti-mouse MHC class II antibody, and subjected to Western blotting using anti-human Aβ antibody (BAM-10). As shown in Fig. 5c, d, data show that CD40 ligation results in recovery of Aβ from MHC class II. Taken together, these data demonstrate that CD40 signaling on microglia promotes MHC class II-Aβ association.
Microglia MHC class II-Aβ co-localization is functional
In order to further evaluate the functionality of CD40L-stimulated MHC class II-Aβ co-localization, we established primary cultured splenocytes from C57BL/6 mice immunized with Aβ1–42. We then isolated CD4+ T cells from these splenocytes and performed an antigen presentation assay, as detailed in Materials and methods. Briefly, microglia were pretreated with Aβ1–42 (1 μM) in the presence or the absence of CD40L protein (2 μg/ml) for 6 h prior to co-culture with Aβ-immunized mouse-derived CD4+ T cells. Thirty-six hours after co-culture, the resulting co-cultured supernatants were collected for analysis of cytokines. As shown in Fig. 6, levels of T cell-derived cytokines (including IFN-γ and IL-2) and microglia-derived cytokines (TNF-α and IL-6) were significantly elevated in the co-treated condition (CD40L plus Aβ1–42) compared to controls, demonstrating that this condition stimulates both microglia APC function and T cell activation.
Furthermore, to test whether this effect could be specifically dependent on the CD40-CD40L interaction, we pre-incubated microglia with antagonist anti-CD40 antibody (2.5 μg/ml) for 1 h. Results show that stimulation of microglia and T cell activation is significantly attenuated as evidenced by decreased secretion of IFN-γ, IL-2, TNF-α, and IL-6 (Fig. 6), suggesting that CD40 signaling is necessary for microglia APC phenotype in the context of Aβ.
We have previously shown that interaction of CD40 with CD40L enables microglia activation in vitro in response to Aβ 11. This finding was validated in vivo by the observation of decreased microgliosis in transgenic mice overproducing Aβ but deficient in CD40L 12. Furthermore, reduced microgliosis in these mice was associated with marked reduction in cerebral Aβ levels and Aβ plaque load. However, not all forms of microglia activation are deleterious, as microglia activated strongly in an acute fashion may subserve a protective role in immunotherapeutic strategies for AD 4, 8, 9.
Here, we suggest a theory for microglia activation that addresses its seemingly controversial role as both potentially therapeutic as well as pathogenic in AD. This theory finds many parallels with the activation of innate immune cells (i.e. macrophages and dendritic cells) in the periphery and is in consonance with numerous studies that have already described the plasticity of CNS microglia as well as their pleiotrophic presentation once activated 29, 30. We suggest that microglia exist in at least two functionally discernable states once “activated” in the context of the AD brain; namely, a phagocytic phenotype or an antigen-presenting phenotype, as governed by costimulatory (i.e. CD40L) environment. In the phagocytic phenotype, microglia function to engulf and degrade Aβ peptide and thereby provide therapeutic benefit for AD. Conversely, switching to the APC phenotype blocks this phagocytic ability and elicits the release of pro-inflammatory cytokines leading to chronic microgliosis, which may act to exacerbate AD-like pathology 31.
We have previously demonstrated that co-activation of microglia with CD40L and Aβ potentiates stimulation of TNF-α production 11. In this study, we extend our earlier finding by showing that the CD40-CD40L interaction can regulate switching of microglia activation in vitro from the phagocytic to the antigen-presenting phenotype. In particular, we observe that co-incubation with Aβ and CD40L inhibits microglia uptake and removal of exogenous Aβ while evoking the release of Th1-type cytokines. We further demonstrate that these Th1-cytokines can directly inhibit microglia phagocytosis of exogenous Aβ peptide. This suggests a positive feedback mechanism whereby microglia products of the APC phenotype drive activation away from anti-inflammatory phagocytosis and towards pro-inflammatory antigen presentation.
Recently, it has been shown that many of the reported Aβ-binding receptors not only mediate signal transduction mechanisms but are most probably also involved in the initial steps of microglia phagocytosis 32, 33. It is a well-documented phenomenon that phagocytosis requires actin filament elongation and that agents like cytochalasin D, which disrupts actin filament polymerization, also prevents phagocytosis. Therefore, to demonstrate that the uptake of Aβ as modulated by Th1/2 cytokines actually involved phagocytosis, microglia were pretreated with 5 μM cytochalasin D for 30 min prior to various cytokine challenge. The data from these studies show that, in the presence of cytochalasin D, about 90% of immunofluorescent localization of Aβ with the microglia was ablated (data not shown), suggesting that the Th1/Th2 cytokine effects we reported on microglia are dependent on microglia phagocytosis per se.
Consistent with the above data, we report increased microglia MHC class II-Aβ co-localization following CD40 ligation in the presence of exogenous Aβ. This suggests that, following CD40 ligation, the phagocytosed Aβ is being directed away from clearance mechanisms (i.e. lysosomal pathway) towards MHC class II loading and presentation. Moreover, we demonstrate that co-localization of MHC class II-Aβ is functional, as it results in T cell activation. This may be of particular importance in immunotherapeutic strategies for AD, as preliminary evidence from the recently halted Aβ1–42 vaccination clinical trails suggests that Aβ presentation by APC may have contributed to the invading T cell-induced meningoencephalitis seen in some of the patients 34.
Interestingly, besides the established signaling cascade initiated by CD40 ligation, CD40 has been shown to directly ensure surface binding and uptake of heat shock protein (HSP)-70-associated proteins 35. In fact, HSP-70-associated proteins have previously been shown to bind Aβ and facilitate their phagocytosis and clearance by microglia 36. However, as we have not directly assessed the role of endogenous HSP-associated proteins in this study, it remains to be determined to what extent they impact the phenomenon that we have described here, specifically microglia CD40 signaling-induced switching from phagocytic to APC phenotype in the context of exogenous Aβ.
In summary, we suggest that some of the confusion about the role of microglia activation in AD may be remedied if one considers that the term “activation” when it is applied to microglia probably represents at least two functionally distinct states. Moreover, we provide data that suggest CD40 signaling pathway stimulation as a mechanism by which “activated” microglia are converted from the phagocytic phenotype to the APC phenotype. With respect to the microgliosis occurring in AD, we suggest that therapeutic interventions that promote the phagocytic phenotype of “activated” microglia might thus avoid the unwanted inflammation sequelae of the APC phenotype and mitigate AD-type pathology.
Materials and methods
Mouse anti-human Aβ monoclonal antibody (BAM-10) was purchased from Sigma (St. Louis, MO). Aβ1–42 and FITC-conjugated Aβ1–42 were obtained from Biosource International (Camarillo, CA). Cy3, an orange fluorescing cyanine dye, was purchased from Amersham (Piscataway, NJ) for conjugation with Aβ1–42 peptide. Human soluble recombinant CD40L protein was purchased from Alexis Biochemicals (San Diego, CA). DuoSet™ enzyme-linked ELISA kits (including TNF-α, IL-1β, IL-6, IL-2, and IFN-γ) were obtained from R&D Systems (Minneapolis, MN). Recombinant proteins (IFN-γ, TNF-α, IL-4, and IL-10) were purchased from R&D Systems. Purified rat anti-mouse MHC class II antibody was obtained from PharMingen (San Diego, CA). The anti-mouse and anti-rabbit horseradish peroxidase-conjugated secondary antibodies were purchased from New England Biolabs (Beverly, MA). Immun-Blot™ polyvinylidene difluoride membranes and the Immun-Star™ chemiluminescence substrate were purchased from Bio-Rad (Hercules, CA).
Mouse primary cell culture
Breeding pairs of C57BL/6 mice were purchased from Jackson Laboratory (Bar Harbor, ME) and housed in the animal facility at the University of South Florida Health Science Center. Mouse primary cultured microglia were isolated from mouse cerebral cortices and grown in complete RPMI 1640 medium according to previously described methods 37. Briefly, cerebral cortices from newborn mice (1–2 days old) were isolated under sterile conditions and were kept at 4ºC prior to mechanical dissociation. Cells were plated in 75-cm2 flasks, and complete medium was added. Primary cultures were kept for 14 days so that only glial cells remained, and microglial cells were isolated by shaking flasks at 200 rpm in a Lab-Line™ Incubator-Shaker. More than 98% of these glial cells stained positive for MAC-1 (Boehringer Mannheim, Indianapolis, IN). Additionally, between 85% and 95% of microglia stained positive for CD45 by fluorescence-activated cell sorter (FACS) analysis as described previously 37.
Microglial phagocytosis assay
Primary mouse microglia were seeded at 1×105 cells/well (n=6 for each condition) in 24-well tissue culture plates containing 0.5 ml of complete RPMI 1640 medium. These cells were treated for 15, 30, and 60 min with “aged” Aβ1–42 conjugated with FITC (Biosource International) 38. In the presence of FITC-Aβ1–42, microglia were then co-treated with CD40L or control (heat-inactivated CD40L protein, 2 μg/ml). Some of these cells were treated with “aged” FITC-Aβ1–42 in the presence of cytokines (including TNF-α, IFN-γ, IL-4, and IL-10) for 60 min. Microglia were then rinsed three times in Aβ-free complete medium, and the media were exchanged with fresh Aβ-free complete medium for 10 min both to allow for removal of non-incorporated Aβ and to promote concentration of the Aβ into phagosomes.
Extracellular and cell-associated FITC-Aβ were quantified using an MSF (SpectraMax®, Molecular Devices) with an emission wavelength of 538 nm and an excitation wavelength of 485 nm. A standard curve from 0 to 500 nM of FITC-Aβ was run for each plate. Total cellular proteins were quantified using the Bio-Rad protein assay. The mean fluorescence values for each sample at 37ºC and 4ºC at the indicated time points were determined by fluorometric analysis. Relative fold change values were calculated as: mean fluorescence value for each sample at 37ºC / mean fluorescence value for each sample at 4ºC. In this manner, both extracellular and cell-associated FITC-Aβ were quantified. Considering nonspecific adherence of Aβ to plastic surface of cultured plates, an additional control without cells was carried out through all of experiments above. We found that incubation time of less than 4 h did not change the amount of Aβ peptide detected in the supernatant, which is consistent with a previous report 39. In order to determine the extent to which cell death might have influenced the phagocytic activity in the various treatment groups, we performed the LDH assay on the relevant supernatant. Data showed that there was no significant cell death occurring over the 3-h time frame in any of the treatment groups (p>0.05).
Fluorescence microscope examination
Cy3-Aβ1–42 was prepared according to methods described in 38. Primary microglia (seeded in 24-well tissue culture plates) were treated with pre-aggregated Cy3-Aβ1–42 (50 nM) in the presence or absence of CD40L (2 μg/ml) at 37°C for 60 min. In addition, in parallel 24-well tissue culture plates, microglia were incubated at 4°C with Cy3-conjugated Aβ (50 nM) with or without CD40L (2 μg/ml) for 60 min. Following treatment, these cells were washed five times with ice-cold PBS to remove extracellular Aβ and then stained with FITC-anti-mouse MHC class II antibody (2 μg/ml) at 4°C for 15 min. After washing three times with ice-cold PBS, these cells were fixed in 4% paraformaldehyde, mounted, and then viewed under an Olympus IX71/IX51 microscope equipped with a digital camera system.
Western blot analysis
Primary cultured microglia were treated with pre-aggregated Aβ1–42 (3 μM) in the presence or absence of CD40L (2 μg/ml) for 1 h. Cell lysates were prepared and immunoprecipitated using anti-mouse MHC class II antibody (1:50), and then subjected to Western blot using BAM-10 (1:1,500) 37. To ensure equal loading, cell lysate immunoblots were probed in parallel for actin as well as human Aβ.
We intraperitoneally immunized C57BL/6 mice at 4 weeks of age with a single dose of Aβ1–42 or vehicle control (PBS) emulsified 1:1 (v/v) in complete Freund's adjuvant. Seven to ten days later, these mice were sacrificed; their spleens were isolated, homogenized individually in 10 ml of complete medium 37, centrifuged at 5,000×g, resuspended in 3 ml of RBC lysis buffer, and allowed to stand on ice for 10 min. Seven milliliters of complete medium was then added to spleen homogenates, the mixture was centrifuged again at 5,000×g, the supernatants were removed, and the resulting splenocytes were resuspended in 10 ml of complete medium. CD4+ T cells were isolated from these splenocytes as previously described 40 and distributed at a concentration of 1×106 cells/well in 24-well plates in the presence or absence of co-cultures of Aβ1–42 pre-treated microglia at a concentration of 1×105/well in 1 ml of medium. Microglia had been pre-treated with Aβ1–42 either in the presence or absence of CD40L protein for 6 h prior to co-culture 41.
The resulting co-cultured supernatants were collected for analysis of cytokines. IFN-γ, IL-2, TNF-α, and IL-6 were measured by ELISA as previously reported 40, 41. Total cellular proteins of each sample were quantified using the Bio-Rad protein assay, enabling to express the results as pg of each cytokine/mg of total cellular protein. Data are represented as relative fold mean cytokine, defined as pg of each cytokine/mg of total cellular protein in the co-cultures normalized to CD4+ T cells alone (in the absence of microglia).
This work was supported by grants from the NIH/NINDS (to J. Tan) and the Alzheimer's Association (to J. Tan). We thank Dr. Y. B. Zhao (Surgery Branch, NIH/NCI) for helpful discussion. We also thank N. Sun for his assistance in primary cultured microglia. Dr. T. Town is a recipient of an NIH/NRSA/NIA post-doctoral fellowship.