Abstract
Neuronal damage induced by ongoing human immunodeficiency virus type 1 (HIV-1) infection was investigated in humanized NOD/scid-IL-2Rγcnull mice transplanted at birth with human CD34-positive hematopoietic stem cells. Mice infected at 5 months of age and followed for up to 15 weeks maintained significant plasma viral loads and showed reduced numbers of CD4+ T-cells. Prospective serial proton magnetic resonance spectroscopy tests showed selective reductions in cortical N-acetyl aspartate in infected animals. Diffusion tensor imaging revealed structural changes in cortical gray matter. Postmortem immunofluorescence brain tissue examinations for neuronal and glial markers, captured by multispectral imaging microscopy and quantified by morphometric and fluorescence emission, showed regional reduction of neuronal soma and synaptic architectures. This was evidenced by loss of microtubule-associated protein 2, synaptophysin, and neurofilament antigens. This study is the first, to our knowledge, demonstrating lost neuronal integrity after HIV-1 infection in humanized mice. As such, the model permits studies of the relationships between ongoing viral replication and virus-associated neurodegeneration.
Introduction
Human immunodeficiency virus 1-associated neurocognitive disorders (HAND) remain the most common CNS disease linked to advanced viral infection and present despite the widespread use of antiretroviral therapy (Antinori et al., 2007). Although considerable research is directed towards understanding disease pathobiology, little is known about disease onset and progression. How advanced viral infection elicits neuronal damage is linked to immune, neural cell, and tissue pathobiological events. These are critical in ultimately producing cognitive, behavioral, and motor disorders and causing disabling morbidity in infected people (McArthur et al., 2010).
Virus invades the CNS within 1–2 weeks after viral infection coincident with the peak of viremia and seen within perivascular macrophages and lymphocytes (Davis et al., 1992; Horn et al., 1998; Williams et al., 2001). During the seroconversion reaction, mild aseptic meningitis is observed in up to 20% of infections (Atwood et al., 1993; Newton, 1995; Gray et al., 1996; del Saz et al., 2008). The long-term sequelae of such events to the brain are incompletely understood (Boufassa et al., 1995; Wallace et al., 2001). Although such pathobiological events are best studied in animal models of human disease (Kraft-Terry et al., 2009), gaps in understanding human immunodeficiency virus 1 (HIV-1) neuropathogenesis parallel the limited availability of relevant model systems. As lentiviral infections are host cell specific, the mainstay for studies of human disease have centered on the use of rhesus macaques infected with simian immunodeficiency virus (SIV) (Lackner et al., 1991; Narayan et al., 1995). Although this model has yielded insights into viral pathobiology and can, in part, mimic human disease, existing limitations in terms of cost, conduct, and availability to investigators have hampered research progress. This has galvanized efforts to develop disease-relevant rodent models (Santoro et al., 1994; Toggas et al., 1994; Persidsky et al., 1996; Reid et al., 2001; Kim et al., 2003; Potash et al., 2005). Nonetheless, such models, while demonstrating relevance for studies of human disease diagnosis and therapies, have also met with limitations. These limitations include impaired graft survival and the inability to mimic progressive HIV infection and neuroAIDS, in particular. The recent development of humanized NOD/scid-IL-2Rγcnull (NSG), BALB/c-Rag2−/−γc−/−, and NOD/scid mice by transplanting with human cord blood isolated CD34+ hematopoietic stem cells (HSCs), or with fetal liver and thymus, have overcome many of these limitations. Indeed, humanized mice can facilitate studies of chronic HIV-1 infection in immune grafts that are sustained for the lifetime of the animal (for review, see Van Duyne et al., 2009). Importantly, in the field of neuroAIDS, this model may identify relationships between HIV-1 disease progression and neuropathobiology.
To these ends, NSG mice were transplanted at birth with CD34+ HSCs (CD34-NSG), and then infected with HIV-1. During progressive viral infection, we investigated when and to what extent neuronal injuries occur after infection onset and how such changes correspond to immune, neuronal, and glial aberrations. The data showed that progressive viral infection correlated with loss of neuronal integrity. As such, this model can uncover relationships between ongoing viral replication and its associated neural dysfunction, and is useful for studies of HIV-1 neuropathogenesis and therapeutic development.
Materials and Methods
Generation of CD34-NSG mice.
Animals were obtained from The Jackson Laboratory and bred under specific pathogen-free conditions in accordance with the ethical guidelines for care of laboratory animals at the University of Nebraska Medical Center and set forth by the National Institutes of Health.
CD34+ cells were obtained from fetal liver. Tissues were provided by the University of Washington, Laboratory of Developmental Biology, supported by National Institutes of Health Award 5R24HD000836 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. CD34+ cells were enriched using immunomagnetic beads (CD34+ selection kit; Miltenyi Biotec). Purity of CD34+ cells was >90% by flow cytometry. Cells were transplanted into newborn mice irradiated at 1 Gy using a C9 cobalt 60 source (Picker Corporation). CD34+ cells were injected intrahepatically at 105 cells/mouse in 20 μl of PBS using a 30 gauge needle. The levels of engraftment and number of human cells in peripheral blood were analyzed by flow cytometry.
Flow cytometry.
Peripheral blood was collected from the submandibular vein in EDTA-coated tubes by using lancets (MEDIpoint) or by cardiocentesis at the study end. Blood leukocytes and spleen cell suspensions (collected at end point of the experiment) were tested for human pan-CD45, CD3, CD4, CD8, CD14, and CD19 markers as six-color combinations. Antibodies and isotype controls were obtained from BD Biosciences Pharmingen, and staining was analyzed with a FACSDiva (BD Biosciences Immunocytometry Systems). Results were expressed as percentages of total number of gated lymphocytes. The percentages of CD4- and CD8-positive cells were obtained from the gate set on human CD3+ cells. Four HIV-1-infected and six control-uninfected animals (all male) were analyzed.
HIV-1 infection and peripheral viral load.
At 22 weeks of age, male mice were injected intraperitoneally with HIV-1ADA using a single dose of 105 50% tissue culture infectious doses/ml. Levels of viral RNA copies/ml were analyzed with the automated COBAS Amplicor System V1.5 (Roche Molecular Diagnostics). For the assay, 10 μl of mouse sera were diluted to 500 μl with sterile filtered normal human sera. The detection limit of the assay after dilution is 2500 viral RNA copies/ml.
Quantitative localized 1H-magnetic resonance spectroscopy.
Single voxel localized spectra were acquired using point resolved spectroscopy (PRESS) with outer volume suppression and high bandwidth pulses to optimize sequence performance. Acquisitions were acquired with a repetition time of 4 s, echo time of 33 ms, 256 averages, using volume coil transmit and surface coil receive on a 7 tesla/16 cm Bruker Pharmascan magnetic resonance imaging (MRI)/magnetic resonance spectroscopy (MRS) system. Single-scan-localized unsuppressed water signals were acquired as a reference for metabolite quantification. Spectroscopic data were processed by removal of residual water signal using the HLVSD filter. Spectra from 1H-MRS data sets were curve fitted in the time domain. Time domain analyses prevent the uncertainties associated with fitting the baseline to remove baseline intensity contributions to peak areas. Signals were fit in the time domain by the QUEST algorithm (Ratiney et al., 2004, 2005) using spectra (basis set) obtained from phantoms acquired by the same pulse sequence used during in vivo acquisition. Data for basis sets were used to calibrate signal intensities. These were normalized to water without correction for relaxation. Metabolite concentrations reported were semiquantitative. To preclude concentration corrections for relaxation, water normalized signal amplitudes were presented in institutional units (IU). IU approximates millimolar concentrations of metabolites.
Diffusion tensor imaging acquisition.
Mice were anesthetized by inhalation of 1–2% isoflurane in a nitrous oxide/oxygen mixture, during MRI data acquisition. Diffusion tensor imaging (DTI) data were obtained using a 7 tesla/16 cm Bruker Pharmascan, an actively decoupled 72 mm volume coil transmitter and a laboratory built 1.25 × 1.5 cm surface coil receiver. Reference coronal mouse brain images were obtained using three-dimensional FLASH with acquisition parameters of 24 × 24 × 16 mm field of view, a 128 × 128 × 32 matrix, 35° flip angle, 50 ms repetition time (TR), 3 ms echo time (TE), and two averages for a total acquisition time of 7.5 min. DTI data were acquired using single-shot diffusion-weighted spin-echo echo-planar imaging (TE, 43 ms). Respiratory gating was used to prevent motion artifacts. Acquisition was repeated at TR of 10–15 s depending on respiratory rate and one shot per breath. Echo-planar imaging acquisition parameters included the following: 14 slices, 200 kHz bandwidth, 96 × 96 in-plane acquisition zero-filled to 256 × 256, and a 0.5 mm slice thickness. The diffusion encoding used was a balanced, rotationally invariant, and alternating polarity icosahedral scheme (12 directions) (Boska et al., 2007). The encoding scheme was designed to reduce background-diffusion gradient coupling (Neeman et al., 1991a,b). Diffusion weighting b factor = 800 s/mm, d = 4 ms, D = 15 ms, Gdmax = 40 G/cm, 200 ms rise time, seven averages for b = 0 acquisition, three averages for each b = 800 encoding direction, for a total acquisition time of 20–40 min, depending on respiratory rate. Respiratory monitoring of animals, gating of the MRI system, and temperature monitoring of phantoms were done using an SA Instruments model 1025 MRI-compatible physiological monitoring system (model 1025; SA Instruments). Animal temperature control was maintained using a circulating water bath (Gaymar Industries) with heat exchange coils surrounding the mouse body on a custom-constructed MRI-compatible stereotactic mouse holder to immobilize the head. Water phantom studies using identical acquisition parameters and coils as the in vivo data were acquired monthly for quality assurance of instrument stability as well as to validate the lack of direct effect of HIV-1 on Dav and fractional anisotropy (FA). Phantoms were chosen as they closely matched the coil loading of a mouse. Phantom temperature was monitored by the thermocouple of the MRI-compatible monitoring system (SA Instruments). These studies show consistent temperature-corrected Dav (mean ± SD, 2.29 ± 0.30 × 10−3 mm2/s at 24°C) and low FA (0.065 ± 0.035).
Image analyses.
Analyses of the diffusion-weighted data were performed using custom programs written in IDL as previously described (Basser et al., 1994a,b; Hassan et al., 2001). Analyses produced maps of the tensor diffusivities (λ1, λ2, λ3), average diffusivity (Dav) where Dav = 1/3*(λ1 + λ2 + λ3), and FA. Transverse (λ⊥ = (λ2 + λ3)/2) and longitudinal (λ|| = λ1) components of the diffusion tensor were obtained as described (Hasan and Narayana, 2006).
Statistical analysis.
Values of Dav, FA, λ⊥, and λ|| were acquired from both hemispheres of each mouse studied. The values from both regions of interest were averaged to arrive at the value of each variable as a single observation within each mouse, as paired t test demonstrated no difference between results in the two hemispheres. Changes of FA, Dav, λ⊥, and λ|| after infection were compared with preinfection values by an unpaired t test as a preliminary analysis. Differences were considered significant at p < 0.05. Statistical comparisons are performed using SPSS 14 (SPSS, Inc.).
Histopathological evaluation.
Brains were perfused with PBS followed by 4% paraformaldehyde and then postfixed overnight and embedded in paraffin. Sections, 5 μm thick, were cut from the paraffin blocks, mounted on glass slides, and labeled with mouse monoclonal antibodies (Dako) for HLA-DQ/DP/DR (clone CR3/43; 1:100), HIV-1 p24 (clone Kal-1; 1:10), and glial fibrillary acidic protein (GFAP) (1:1000). Microglia and macrophages were stained with rabbit polyclonal antibodies to ionized calcium binding adaptor molecule 1 (Iba-1) (1:500; Wako Chemicals). The polymer-based HRP-conjugated anti-mouse and anti-rabbit Dako EnVision systems were used as secondary detection reagents and developed with DAB (3,3′-diaminobenzidine). All paraffin-embedded sections were counterstained with Mayer's hematoxylin. Deletion of primary antibodies or mouse IgG served as controls. Images were obtained with a Nikon DS-Fi1 camera fixed to a Nikon Eclipse E800 (Nikon Instruments) using NIS-Elements F 3.0 software.
Immunofluorescence labeling, multispectral imaging, and image analyses.
Slide specimens (5 μm thick) of paraffin-embedded mouse brain were selected from the corresponding regions analyzed by DTI. Brain sections were treated with different paired combinations of primary mouse monoclonal antibodies to 200 kDa plus 68 kDa neurofilaments (NFs) (clone 2F11; 1:200; Dako) and synaptophysin (SYN) (1:1000; Novus Biologicals), and rabbit polyclonal antibodies to MAP2 and GFAP. Primary antibodies were labeled with secondary anti-mouse and anti-rabbit antibodies conjugated to the fluorescent probes Alexa Fluor 488 and Alexa Fluor 594, and nuclei were labeled with 4′,6-diamidino-2-phenylindole (DAPI). Slides were coverslipped with ProLong Gold anti-fade reagent (Invitrogen), allowed to dry for 24 h at room temperature, and then stored at −20°C for future use. Images were captured at wavelengths encompassing the emission spectra of the probes, with a 20× objective. The fluorescence emission of each probe and autofluorescence of the tissue samples were analyzed by multispectral imaging/image analysis, with an automated fluorescence microscope (Leica DMRXA2; North Central Instruments), Nuance EX multispectral imaging system (Cambridge Research Instruments), and ImagePro Plus image analysis software (Media Cybernetics). These spectra were incorporated into a spectral unmixing algorithm (Nuance system) that quantitatively separated the grayscale images representing each spectral component. The grayscale images representing each fluorescent probe were exported from the Nuance system to the ImagePro Plus environment where fluorescent events, in brain regions of interest, were partitioned, and the fluorescence signal (same exposure times for controls and infected samples) was quantified as mean pixel intensity (12 bit grayscale) and total area (in square micrometers) of positive labeling. Area-weighted average fluorescence intensity was calculated for MAP2 and SYN expression by dividing the sum of the product of area and mean fluorescence intensity, for each partitioned area, by the sum of the partitioned areas. Fluorescence emission, across areas of hundreds of micrometers, were captured by wide-field microscopy. Loss of light due to light scatter at the perimeters was judged as negligible as partitions included areas of interest up to 200,000 μm2 and individual events of 25–50 μm2.
Results
Human cell reconstitution and HIV-1 infection of mice
Four humanized animals were monitored prospectively for numbers of CD4+ and CD8+ cells in circulation and, at the end point of observation, in spleen. At 22 weeks of age, mice were infected with HIV-1ADA and viral load (VL) was assayed at 2, 4, 8, 12, and 15 weeks postinfection (wpi) (Fig. 1A–D); this was performed at the same time as MRS/DTI imaging. VL peaked at 4 weeks after viral infection and all infected animals had a sustained VL (1.7 × 104 to 7.6 × 105 RNA copies/ml). The decline of CD4+ cells progressed rapidly because of disease between 8 and 15 weeks (Fig. 1B) but did not change in control noninfected mice observed over the same time period (n = 30 weeks) (Fig. 1C). The rapid decrease in the CD4 count during the course of infection, forced us to terminate the study at 15 wpi. The human cell profiles in the spleens of infected animals at the termination point were also analyzed (Fig. 1E). Animals were stratified according to their immune status and peripheral VL. Animals m969 and m973 had 21.1 and 36.4% of CD3+CD4+ cells in splenic lymphoid tissue, CD4:CD8 cell ratios of 1.0 and 0.8 (lower compared with 2:1–3:1 in uninfected animals), and peripheral VLs of 1.71 × 105 and 0.65 × 105 viral RNA copies/ml, respectively. They were considered less affected by HIV-1 infection compared with two other mice m983 and m985, which had higher VLs (7.6 × 105 and 4.81 × 105 viral RNA copies/ml) and inverted CD4:CD8 ratios (0.1 and 0.4), and m983 had the lowest number (3.4%) CD3+CD4+ cells in spleen.
Time frame of observation, immune and viral load status of experimental humanized mice. A, Human CD34+ cell reconstitution, MRSI/DTI analysis, HIV-1 infection, and splenic histological analysis of NSG mice are mapped to a study timeline. B, C, Analyses of CD4 and CD8 cell populations in peripheral blood of four HIV-1-infected mice and six control mice. D, The dynamics of VL in peripheral blood of infected animals are shown. Mice were bled once every 4 weeks starting the second week after HIV infection. E, Percentage of human T- and B-lymphocytes and macrophages in spleen were determined at the end point of the study and mapped with end point VL. B–E, Mean and SEM are shown (B–D) and individual numbers for each mouse (E).
Brain immunopathobiology
Brain pathologies, which included influx of activated HLA-DR+ human cells, activation of microglial cells (Iba-1), and astrogliosis (GFAP) were determined by immunohistology of paraffin-embedded 5-μm-thick sections (Fig. 2). Brain regions including frontal cortex, whisker barrel (WB) somatosensory cortex (data not shown), cerebellum, corpus callosum, hippocampus, and brainstem were investigated. Mouse m983 (with highest VL and lowest CD3+CD4+ cell percentage in spleen) had the most severe brain pathology. The presence of scattered lymphocytes in brain parenchyma, accumulation of human cells in meninges, and formation of perivascular cuffs around brain microvessels were found (Fig. 2). In the pons/medulla, activation of microglial cells with nodule formation not directly associated with human cell infiltration was noted. The same markers in control uninfected humanized mice of the same age group were analyzed, and the data of one representative mouse (m975) are shown in Figure 2. Based on the CD4 cell profiles, viral load profiles, and brain pathologies observed, the animals were grouped into two categories as severe brain pathology (m983 and m985) and moderate brain pathology (m969 and m973).
Histopathology of brain subregions. Representative brain regions in control uninfected (HIV-1−) m975 and infected (HIV-1+) m983 animals. Paraffin sections (5 μm) of brain were labeled with anti-HLA-DR to identify human activated cells, anti-Iba-1 to detect macrophages and microglia, and anti-GFAP antibodies for astrocytes (all positive labeling is brown). Sections were counterstained with hematoxylin. Images of cortex, cerebellum, hippocampus, brainstem, and corpus collosum were captured with a 40× objective. Accumulation of human cells in meninges, perivascular spaces with activation of microglial cells (pons/medulla) and astrocytes (cortex, hippocampus, and pons/medulla) in infected mouse were noted. The arrows indicate perivascular cuff of HLA-DR+ cells; red circle indicates microglial nodule formation.
1H-MRS measures
1H-MRS tests were used to determine whether changes in neuronal integrity could be present. These results were validated by immunocytochemical tests for neuronal integrity. The regions selected for spectroscopic examination using the PRESS volume selective spectroscopic acquisitions included cerebral cortex, hippocampus, cerebellum, and brainstem/pons (Fig. 3A). 1H-MRS was collected from four humanized NSG mice at different time points (0, 4, 8, 12, and 15 weeks after infection with HIV). Preinfection reconstituted (time, 0 weeks), unmanipulated humanized NSG mice aged 3 months (n = 5), and NSG mice aged 6 months (n = 3), showed no significant differences in N-acetyl aspartate (NAA) concentrations among the different brain regions studied (data not shown). The NAA concentration of uninfected mice of equivalent age (36 weeks) and time (equal to 15 weeks after infection) was 10–12 IU as measured in the cerebral cortex. The nervous system-specific metabolite NAA, which is synthesized in neurons and appears to be a key link in distinct biochemical features of CNS metabolism, was measured at the time points described above. The means ± SEMs of NAA concentration normalized to water for quantitative assessments are presented in Figure 3B. It can be clearly seen that NAA is significantly reduced over the course of infection in the cerebral cortex (Fig. 3B) but not in the cerebellum, hippocampus, or brainstem/pons regions. Moreover, one mouse (m983) showed a substantial reduction in NAA in the cerebellum (NAA, 6.85 IU) at the 15 wpi. As NAA is commonly presented as a ratio to creatine, we compared NAA levels normalized to creatine or to water. Both measures were nearly identical. Creatine remains constant over time in the animals (data not shown). Hence, presentation of NAA as either concentrations or metabolite ratios to creatine provides equivalent results in the animals.
1H-MRS analyses. A, Brain regions of interests. B, N-Acetyl aspartate concentrations in HIV-1-infected humanized mice. The means ± SEMs of [NAA] from selected regions of brain are shown. The values were normalized to water for quantitative assessment. The values were drawn at *p < 0.05 versus preinfection and +p < 0.1 versus preinfection. It can be clearly seen that NAA is significantly reduced over the course of infection in the cerebral cortex but not in the cerebellum, hippocampus, or brainstem/pons region.
DTI tests
The DTI measures for FA reflect changes of the tissue organization at a microscopic level by measuring water diffusion in different directions, which changes as a consequence of structural neuronal geometry. DTI examinations were performed over the same time period as 1H-MRS measurements. DTI results from the selected regions measured bilaterally from the four animals were divided based on peripheral immune status, VL, and brain immunopathology analyzed independently. Results from the severely affected mice (m983 and m985) are presented in Figure 4. Notable and consistent changes include increased FA in the frontal cortex and hippocampus as well as decreased Dav, λ||, and λ⊥ in the WB along with other cortical areas and hippocampus. Multiple gray matter regions show statistically significant changes over the course of disease development, more than were found from moderately affected mice m969 and m973 (data not shown). In the WB, there was a strong linear correlation between VL and FA, specifically when correlated magnitude of FA increase (Δ) from preinfection to 15 wpi (R2 = 0.9987). The same strong trends were found for decline of Dav diffusivity (Fig. 4E,F). Similar correlations of FA and VL were observed in the pons/medulla (R2 = −0.9811) (data not shown).
DTI tests. A, Fractional anisotropy (FA). B, Mean diffusivity (Dav). C, Longitudinal diffusivity (λ||). D, Transverse diffusivity (λ⊥). Notable and consistent changes include increased FA in the frontal cortex (FC) and hippocampus (Hip), as well as decreased diffusivity in the whisker barrels (WB), cerebral cortex (CCx), and hippocampus of humanized mice group with more severe brain pathology. Statistically significant changes were not observed in the corpus callosum (CC), substantia nigra (SN), medulla (M), and cerebellum (C). E, G, Correlation analyses of endpoint viral load with endpoint DTI metrics (FA and Dav) in WB. F, H, Correlation analysis of magnitude of changes of DTI metrics (Δ) before and after HIV infection with viral load in WB region.
Quantitative multispectral fluorescence microscopy
To determine whether in vivo reduction of NAA concentration detected by 1H-MRS and changes in DTI values correspond to changes of brain architecture, at termination of the experiment, immunofluorescence neuronal and glial antigens were analyzed by multispectral fluorescence microscopy of replicate brain regions. The images captured with the Nuance multispectral imaging system (20× objective) were converted to 12 bit grayscale images and quantified using ImagePro Plus software. Representative images of the labeled neuronal markers captured (40× objective) with the Nuance multispectral imaging system are presented in Figure 5. The cortical neurons of control animals showed normal MAP2-immunolabeled dendrites, whereas the HIV-1-infected humanized mice showed decreased expression of MAP2-immunolabeled dendrites with irregular and wavy contours. SYN labeling in cortical areas of control animals was observed as diffusely punctate, whereas SYN expression was irregularly shaped in all infected animals. Reduction of NF-positive fibers was observed in the cortex of animals with significant immune and brain pathology (m983 and m985) compared with less affected (m969 and m973) and control animals. In WB of all infected animals, the amount of NF was significantly reduced (Fig. 5). However, in the pons/medulla, these changes were not as significant as seen in cortex and WB (Fig. 5, bottom right panel).
Immunofluorescence labeling of astrocytes, neurofilaments, and synapses in brain regions of both HIV-infected and noninfected CD34-NSG mice. Slide-mounted sections (5 μm) of brain from a noninfected mouse with the least brain damage (control m975) and an infected mouse with a markedly elevated viral load and exhibiting the most brain damage (m983) were stained for GFAP (an astrocyte-specific marker; red) and Neurofilament (green). Images of cerebral cortex, whisker barrel, and brainstem were captured at 40× objective. Dendritic and synaptic morphology of cortical neurons are illustrated. Sections of mouse m975 and mouse m983 were labeled for MAP2 (red), Synaptophysin (green), and nuclei (DAPI; blue). The panels to the far left map tissue labeling to regions in the context of whole-brain sections.
For MAP2 and SYN, the quantitative data were expressed as total positive area/total area analyzed in square micrometers and their corresponding mean pixel density intensity (Fig. 6). For NF and GFAP, data were expressed as total area labeled. There was a diminution in the cerebral cortex and WB of MAP2 and SYN levels compared with uninfected animal but not in the pons/medulla. The most dramatic reduction of NF labeling was found in WB (Fig. 6). Observed changes allowed us to suggest that in addition to dendritic/synaptic protein aberrations, the damage represented by NF also contributed to the changes of [NAA] and MRI/DTI metrics.
Quantitative fluorescence microscopy of neuronal protein expressions in brains of CD34-NSG mice. A, Schematic representation of the strategy of analysis. Slide-mounted brain sections (5 μm thick) that included the corresponding regions (mouse brain atlas) analyzed by DTI were analyzed by quantitative fluorescence microscopy with an automated fluorescence microscope (Leica), multispectral imaging system (Nuance EX), and ImagePro Plus image analysis software. Image cubes consisting of a Z-stack of images over a wide range of wavelengths were captured from each microscope field with a 20× objective. Cubes were unmixed by spectra corresponding to each fluorescence label and to autofluorescence into quantitative grayscale images for analysis of intensity and area of fluorescence emission. B, Individual mouse data represent expression of the neuronal proteins MAP2 and SYN (fluorescence intensity) and NF and GFAP (area of positive fluorescence labeling). The reduction of MAP2 and SYN were detected in cortical regions, but not in pons/medulla. The significant drop of NF was found in WB in all analyzed mice and in frontal cortex of two mice with significant immune pathology (m983, m985). It was not clearly observed in pons/medulla region. Variable changes in GFAP-labeled areas were noted for different regions.
Discussion
HIV-1-associated neuropathology seen during progressive viral infection of humanized CD34-NSG mice was linked to systemic viral infection. The sequence of events consisted of viral invasion of the nervous system, glial neuroinflammatory responses, and neurodegeneration. Early development of aseptic meningitis and CD4+ T-lymphocyte depletion was also observed, as reported (Gorantla et al., 2007, 2010a,b). After the initiation of viral infection, development of neuronal abnormalities in cerebral cortex were seen in the animals for up to 15 weeks by spectroscopic and diffusion tensor imaging and confirmed by immunohistological tests. These revealed decreased expression of SYN, an integral protein in presynaptic terminals, and MAP2, a marker for neuronal cell bodies and dendrites, and axonal NF in cerebral cortex. Such changes reflect substantive neurodegeneration and are also found in the brains of HIV-1-infected individuals.
Human neuropathology was previously studied in some depth during the later stages of viral infection in the absence of combination antiretroviral therapy. This is where encephalitis was commonly seen and was a result of virus-infected mononuclear phagocytes (blood-derived perivascular macrophages and microglia) and formation of multinucleated giant cells, astrogliosis, myelin pallor, and neuronal morphological changes. The latter was seen as decreased synaptic density and neuronal dropout (Price et al., 1988; Budka, 1991; Gelman, 1993; Wiley, 1995; Kraft-Terry et al., 2009). These morphological features of HIV-1-induced neurodegeneration have diminished in severity as a consequence of antiretroviral medicines (Masliah et al., 1996; Everall et al., 1999; Zheng et al., 2001; McCarthy et al., 2006).
In recent years, clinical investigators have used MRS and DTI to assess the underlying basis for cognitive dysfunction during progressive HIV-1 infection in its human host (Pomara et al., 2001; Lim and Helpern, 2002; Schifitto et al., 2009; Müller-Oehring et al., 2010). In the present study, we used both 1H-MRS and DTI, to monitor biochemical and structural alterations, respectively, in brain regions during the progression of HIV infection in humanized mice. Our 1H-MRS tests showed significant decreases in NAA concentration in cerebral cortex. NAA, the most abundant neuron-specific metabolite reflects the degree of neuronal injury (Chong et al., 1994; Meyerhoff et al., 1994; Laubenberger et al., 1996; Simone et al., 1998; Moffett et al., 2007). Moreover, NAA concentration is often used as a marker of neuronal viability in a range of neurodegenerative disorders that includes HAND, Alzheimer's disease, epilepsy, multiple sclerosis, and spinal cord injury (Meyerhoff et al., 1993, 1994; Marcus et al., 1998; Suwanwelaa et al., 2000; Tarasów et al., 2003; Chang et al., 2004; Soares and Law, 2009; Mohamed et al., 2010). Reductions in NAA were also seen in 1H-MRS studies of SIV-infected monkeys (Lentz et al., 2005, 2008a,b; Ratai et al., 2009). Although previous studies showed changes in NAA concentrations as reflective of human cognitive impairment, their use has been hampered by variations in levels over time, comorbid conditions, and the inability to coregister histopathological correlates of affected brain regions. The decrease in NAA concentration in the cerebral cortex occurred in all four HIV-1-infected humanized mice and correlated with diminution of MAP2, SYN expression. However, in two mice with the highest levels of viral infection (m983 and m985), NF expression was also reduced. In contrast, astrogliosis, as measured by GFAP expression, was not uniformly altered. The diminished NAA concentration, seen in the cortex of the HIV-1-infected humanized mice, may not be directly related to neuronal apoptosis but rather may be an indicator of neuronal abnormalities and is supported by the findings that the biosynthetic enzyme for NAA, N-acetyl transferase, is localized in neuronal mitochondria and synaptosomal cell fractions (Moffett et al., 2007). More work certainly remains to be done to quantify the extent of neuronal damage and validate biomarkers of virus-associated neurodegenerative disease in this model system.
In addition to biochemical changes, we found structural brain tissue alterations by DTI. We observed an increase of FA and a decrease of diffusivity in several gray matter regions starting 4 wpi in animals with higher viral load and established immunopathology. DTI can assess displacement of water on the micrometer scale, yielding information about white matter fiber integrity (Takahashi et al., 2000; Assaf and Pasternak, 2008; Y. Chen et al., 2009; Gongvatana et al., 2009) as well as gray matter during CNS development (McKinstry et al., 2002; Mukherjee et al., 2002; Mukherjee and McKinstry, 2006; Neil and Inder, 2006; Yassa et al., 2010) and neurodegeneration such as Alzheimer's disease (Head et al., 2004; Rose et al., 2008; T. F. Chen et al., 2009; Cherubini et al., 2010; Kantarci et al., 2010; Sexton et al., 2010). FA is highly sensitive to microstructural changes but is not very specific to the type of changes. If in the developing brain a decrease in FA represents ongoing establishment of cortical neuronal interconnectivity, then in earlier stages of HIV-1 infection in humanized mice the changes observed could be associated with structural alteration of synaptic and dendritic processes. The reduction in expression of MAP2 and SYN and the abnormal appearance of SYN-positive structures in the cortex support our findings of FA and diffusivity values.
We also systematically studied, and in replicate manner to the bioimaging and histopathological tests, cognitive and sensorimotor functions studying NSG mice. Others' and our previous experiences in short-term rodent models of HIV-1 encephalitis were founded on Morris water maze testing (Avgeropoulos et al., 1998; Zink et al., 2002) and a water escape radial arm maze test series (Sas et al., 2007). In these works, cognitive deficits (spatial memory development and retention) were shown as was seen in aged HIV-1 gp120 transgenic mice (D'Hooge et al., 1999). However, water maze could not be easily used for our current model as NSG animals are substantively immunologically impaired and any exposure to cold water significantly increases the risk of infection and subsequent death of the animal. To overcome such impediments and to test learning and memory, we selected the Barnes circular platform task (Markowska et al., 1989). This system was used to perform several basic testing. We found that NSG mice had gender-dependent differences as males had significantly increased exploratory behavior on the platform compared with females. We also observed age-dependent decline in learning and memory in 12-month-old compared with 6-month-old males. We then compared learning and memory in unmanipulated versus animals irradiated at birth. Here, females were reconstituted at 6 months of age. The unmanipulated and irradiated/reconstituted mice did not show statistically significant differences in cognition (our unpublished observations) and suggested that such an approach would not readily yield changes in learning and memory after HIV infection. Currently, we are in the process of standardizing tests for behavioral testing in this unique model.
The only other animal system that shows the changes observed herein in humanized mice is the SIV-infected rhesus macaque (Bissel et al., 2002; Lentz et al., 2005; Scheller et al., 2005; Thompson et al., 2009). Importantly, and also operative in SIV-infected monkeys, few numbers of virus-infected and immune-competent perivascular mononuclear phagocytes can affect widespread neuronal dysfunction (Williams et al., 2001, 2005; Marcario et al., 2004; Crews et al., 2008). Nonetheless, expense and species specificity preclude the widespread use of SIV models for the study of the early stages of infection. Thus, relevant rodent models that mimic human disease are sorely needed. To our knowledge, this is the first report to show associations between 1H-MRS and DTI evidence of deficits in neuronal integrity. These deficits were confirmed by immunohistochemical evaluation. This study highlights the relevance of humanized mouse models for research on the pathobiology of human disease and underlies the future applications of noninvasively monitoring the systemic effects of antiretrovirals and adjunctive therapeutics.
Footnotes
-
This work was supported by National Institutes of Health Grants 1P01 DA028555, P20 RR15635, 1 P01 NS043985-01, 2R37 NS36126, 5 P01 DA026146, and 5 P01 MH64570-03. We acknowledge Deepa Roy, Sidra Akhter, and Tanuja Gutti for technical assistance.
- Correspondence should be addressed to Dr. Howard E. Gendelman, Department of Pharmacology and Experimental Neuroscience, University of Nebraska Medical Center, Omaha, NE 68198-5215. hegendel{at}unmc.edu
References
In this issue
