Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jun 1.
Published in final edited form as: Neurobiol Aging. 2023 Jan 26;126:25–33. doi: 10.1016/j.neurobiolaging.2023.01.010

Multi-Omic Characterization of Brain Changes in the Vascular Endothelial Growth Factor Family during Aging and Alzheimer’s Disease

Mabel Seto a,b,+, Logan Dumitrescu a,b,+, Emily R Mahoney a,b, Annah M Sclafani a,b, Philip L De Jager c,d, Vilas Menon c, Mary Ellen I Koran e, Renã AS Robinson a,f, Douglas M Ruderfer b, Nancy J Cox b, Nicholas T Seyfried g, Angela L Jefferson a, Julie A Schneider h, David A Bennett h, Vladislav A Petyuk i,#, Timothy J Hohman a,b,#,*
PMCID: PMC10106439  NIHMSID: NIHMS1872235  PMID: 36905877

Abstract

The Vascular Endothelial Growth Factor (VEGF) signaling family has been implicated in neuroprotection and clinical progression in Alzheimer’s disease (AD). Previous work in postmortem human dorsolateral prefrontal cortex demonstrated that higher transcript levels of VEGFB, PGF, FLT1, and FLT4 are associated with AD dementia, worse cognitive outcomes, and higher AD neuropathology. To expand prior work, we leveraged bulk RNA sequencing data, single nucleus RNA (snRNA) sequencing, and both tandem mass tag and selected reaction monitoring mass spectrometry proteomic measures from the post-mortem brain. Outcomes included AD diagnosis, cognition, and AD neuropathology. We replicated previously reported VEGFB and FLT1 results, whereby higher expression was associated with worse outcomes, and snRNA results suggest microglia, oligodendrocytes, and endothelia may play a central role in these associations. Additionally, FLT4 and NRP2 expression were associated with better cognitive outcomes. This study provides a comprehensive molecular picture of the VEGF signaling family in cognitive aging and AD and critical insight towards the biomarker and therapeutic potential of VEGF family members in AD.

1. Introduction

The vascular endothelial growth factor (VEGF) signaling family is involved in neuroprotection and may contribute to the neurodegenerative process in Alzheimer’s disease (AD, (Harris et al., 2018; Hohman et al., 2015; Huang et al., 2013; Mahoney et al., 2019; Moore et al., 2020; Provias and Jeynes, 2014; Religa et al., 2013; Wang et al., 2011). This signaling family—including five ligands (VEGFA, VEGFB, VEGFC, VEGFD, and PGF), three tyrosine kinase receptors (FLT1, FLT4, and KDR), and two modulating receptors (NRP1 and NRP2)—plays a critical role in angiogenic, neurotrophic, and microglial proliferation pathways in the brain (Lange et al., 2016; Ruiz de Almodovar et al., 2009; Ryu et al., 2009; Tillo et al., 2012). Notably, interactions between this diverse set of ligands and receptors drive vastly different signaling cascades. For example, VEGFA can signal through a KDR-NRP1 complex on endothelial cells to modulate endothelial cell survival, migration, and proliferation (Koch and Claesson-Welsh, 2012; Lee et al., 2011; Lee et al., 1996) while VEGFB signals through FLT1 to promote neuroprotection (Dhondt et al., 2011; Sun et al., 2006). Given the ubiquity of the biological pathways modulated by VEGF signaling, it is difficult to disentangle which family members, and which biological effects, are most relevant to AD and offer promise as therapeutic targets.

VEGFA is the most well-studied member of the VEGF family. Mounting model system and human evidence suggests that both peripheral and central VEGFA levels protect against memory impairment and brain atrophy in AD (Cao et al., 2004; Garcia et al., 2014; Hohman et al., 2015; Leung et al., 2015; Mahoney et al., 2019; Moore et al., 2020; Religa et al., 2013; Spuch et al., 2010; Wang et al., 2011). We have also characterized transcriptional changes of all VEGF family members in the prefrontal cortex in relation to AD neuropathology and clinical disease (Mahoney et al., 2019). We found that higher expression levels of VEGFB, PGF, FLT1, and FLT4 related to faster cognitive decline and greater AD neuropathological burden (Mahoney et al., 2019).

Despite growing evidence that the VEGF family has relevance to AD, several questions remain unanswered. Currently, it is unclear whether the transcriptomic differences we observed in prefrontal cortex extend to other brain regions. Further, it is unknown what cell types in the brain contribute to associations between VEGF family member expression and AD-relevant outcomes. Finally, it is unclear whether previously observed transcriptomic differences result directly in proteomic changes or reflect a compensatory change in transcription to respond to a reduction in protein levels.

Here, we leverage bulk RNA sequencing from prefrontal cortex, posterior cingulate cortex, and head of the caudate nucleus as well as single nucleus RNA (snRNA) sequencing and mass spectrometry measures of protein abundance from prefrontal cortex to address these three gaps and provide the most comprehensive picture of brain VEGF family member associations with AD outcomes.

2. Materials and methods

2.1. Participants

Data were obtained from the Religious Orders Study and the Rush Memory and Aging Project (ROS/MAP). These well-characterized cohort studies enrolled older adults without dementia who agree to annual clinical evaluations and brain donation at death (Bennett et al., 2018). Written informed consent was obtained from all participants. Institutional Review Board (IRB) approval was granted for all protocols. The Vanderbilt University Medical Center IRB approved these secondary analyses. ROS/MAP data are available online on the Accelerating Medicines Partnership – Alzheimer’s Disease (AMP-AD) Knowledge Portal (syn3219045) or through the Rush Alzheimer’s Disease Center Resource Sharing Hub (https://www.radc.rush.edu/).

2.2. Neuropsychological Composites

Cognition was operationalized as a global cognition composite, an average of z-scores from 17 tests across 5 domains of cognition (episodic, semantic, and working memory, perceptual orientation, and perceptual speed). Cognition measures in ROS/MAP have been described elsewhere (Wilson et al., 2015).

2.3. Bulk mRNA Sequencing

Bulk RNA sequencing was performed in 3 brain regions: dorsolateral prefrontal cortex (DLPFC), posterior cingulate cortex (PCC), and the head of the caudate nucleus (CN). Data is available on the AMP-AD Knowledge Portal (syn23650893). Regions were selected by the parent study based on tissue availability and biological relevance to a number of disease conditions. Processing and alignment followed a published protocol (Logsdon et al., 2019). Briefly, reads were aligned to Ensembl (GRCh38, v99) reference genome using STAR (v2.5.2b), gene counts were computed using featureCounts from Subread (v2.0.0), and Picard metrics were calculated (v2.18.27). Samples with RNA integrity number (RIN) <4 or post-mortem interval (PMI) >24 hours were removed, and genes with missing gene length or GC content were removed prior to normalization. Gene counts were quantile-normalized, after which expression values >3 standard deviations (SD) from the mean were excluded. Sample outliers in principal component analysis or who were missing RIN or demographic variables were removed. Samples whose last visit was >2 years before death or who had non-AD dementia were also excluded from analyses.

To ensure observed effects were not due to technical differences, expression underwent further iterative normalization adjusting for batch, sex, death age, PMI, RIN, and percentage of coding, intronic, and intergenic bases, and sensitivity models were rerun with this expression data.

2.4. Single Nucleus RNA Sequencing

Nuclei were extracted and sequenced previously as described elsewhere (Mathys et al., 2019). Briefly, single nuclei were isolated from frozen DLPFC for 48 ROS/MAP samples and sequenced using droplet-based single-nucleus RNA-seq. Expression values were normalized and clustered into 20 pre-clusters. Cell-type labels were assigned based on statistical enrichment of marker genes and manual evaluation of known marker gene expression. Cell-type clusters were identified as all pre-clusters of the same type. This process resulted in single nucleus expression for all VEGF genes, except VEGFD, in up to 8 cell-types (astrocytes, endothelial cells, excitatory neurons, inhibitory neurons, oligodendrocytes, oligodendrocyte precursor cells (OPCs), microglia, and pericytes; Supplementary Table 1). It should be noted that analyses examining expression in endothelial and pericyte cells have low sample sizes (<40 individuals with quantifiable expression).

2.5. Cell-type fraction quantification

For correction of bulk DLPFC cell composition, we leveraged previously estimated cellular proportions calculated with single nucleus data (Cain et al., 2020). Briefly, an iterative method was used to select gene sets representing the predicted proportion of each cell type. Selected genes were then used for calculating cell fraction as proportion of total nuclei per sample. Single nucleus data were not available in PCC or CN, so region-specific expression of 5 cell-specific markers was leveraged for statistical adjustments in cell-type sensitivity analyses: ENO2 (neurons), GFAP (astrocytes), CD34 (endothelial cells), OLIG2 (oligodendrocytes), and CD68 (monocytes).

2.6. Tandem Mass Tag Mass Spectrometry Based Proteomics

VEGF receptor protein expression was quantified using isobaric tandem mass tag mass spectrometry (TMT-MS) on DLPFC tissue from 400 ROS/MAP samples (syn17015098) using the UniProtKB human proteome database containing both Swiss-Prot and TrEMBL reference sequences (downloaded on 21 April 2015, processed data available at syn21266454) as reported previously (Johnson et al., 2020). Samples whose last visit was >2 years before death or who had non-AD dementia were excluded. This quantification yielded measurements for FLT1 (2 peptides), NRP1 (1), and NRP2 (1) in 375 samples.

2.7. Selected Reaction Monitoring Mass Spectrometry Based Proteomics

Selected reaction monitoring (SRM) mass spectrometry was used to quantify VEGF protein expression in DLPFC from 1203 ROS/MAP samples (syn10468856). A standard protocol was used to prepare tissue samples (Andreev et al., 2012; Petyuk et al., 2010). Two micro-liters per sample was used for each measurement, which was performed on a nano ACQUITY UPLC coupled to TSQ Vantage MS instrument. First, peak assignment and boundaries were manually inspected for accuracy. Then, Skyline software was used to calculate peak area ratios of endogenous light peptides to synthetic heavy isotope-labeled peptides and to determine the best transition without matrix interference for accurate quantification. Relative protein abundances were then log2-transformed and median-centered, with the median shifted to 0. Protein abundances were adjusted for differences in total protein amounts within each sample. Expression values >4 SD from the sample mean were removed. This process resulted in quantifications for 4 VEGF family members: VEGFB (1 peptides), FLT1 (2), NRP1 (5), and NRP2 (4) in 1084 ROS/MAP samples, 371 of whom overlap with TMT samples. See Supplementary Table 2 for details on each protein.

2.8. Measures of Alzheimer’s Disease Pathology

ROS/MAP AD pathology measures have been described elsewhere.(Bennett et al., 2018) Briefly, β-amyloid load and tau tangle density at autopsy were quantified as the average percent area occupied across 8 regions (hippocampus, angular gyrus, and entorhinal, midfrontal, inferior temporal, calcarine, anterior cingulate, and superior frontal cortices), measured by immunohistochemistry. Values were square-root transformed to approximate a normal distribution.

2.9. Statistical Analyses

Statistical analyses were completed using R (v4.0.2; https://www.r-project.org/). Linear regression (covarying for age at death, sex, PMI, and interval to death) tested VEGF associations with cross-sectional cognition, β-amyloid load, tau tangle density, and last diagnosis before death. Linear mixed-effects regression tested VEGF associations with cognitive decline, with death age, sex, PMI, interval between last visit and death, interval from each visit to death, and expression entered as fixed effects and the intercept and interval to death entered as random effects. These models were used for VEGF expression measured by SRM, TMT, bulk RNA-seq, and single-nucleus RNA-seq.

Significance was set a priori to α=0.05, and all nominally significant associations (uncorrected p<0.05) are reported in the text. Additionally, p-values were corrected for all VEGF predictors across modalities and outcomes (560 models) using the FDR procedure. These FDR-corrected p-values are included in the Supplementary Tables alongside their corresponding uncorrected p-values, and uncorrected p-values reported in the text that passed correction are indicated by an asterisk. All models were restricted to individuals whose last cognitive assessment was within two years of death.

Sensitivity analyses included covarying for levels of AD pathology at death (both amyloid and tau burden), evaluating sex-stratified and sex interaction effects, evaluating APOE-ε4 stratified and interaction analyses, and rerunning models leveraging more stringent normalization procedures for bulk analyses (refer to Bulk mRNA Sequencing in Methods above).

3. Results

See Table 1 for participant descriptive statistics by data type. Fig. 1 summarizes results for VEGF genes with associations with cognition, diagnosis, or AD pathology passing FDR corrections (i.e., VEGFB, FLT1, NRP2, and FLT4), with all other genes presented in Supplementary Fig. 1. A detailed overview of results from each analysis for each VEGF family member is presented below, and full analysis results are presented in Supplementary Tables 3–11.

Table 1.

Participant Characteristics

Bulk RNA snRNA DLPFC SRM DLPFC TMT DLPFC
DLPFC PCC CN
N 208 490 673 48 1084 375
Age at death (yrs)1 90±6 89±7 89±6 87±6 89±6 89±6
Education (yrs) 16±4 16±4 16±4 19±3 16±4 16±4
Global Cognition −0.8±1.2 −0.67±1.03 −0.74±1.08 −1.12±1.37 −0.84±1.13 −0.53±1.05
Average Follow-up (yrs) 9.38±5.87 7.51±4.86 7.69±4.84 7.9±4.68 8.12±5.11 8.79±4.48
% Male 36% 38% 36% 50% 33% 30%
% NHW2 95% 96% 97% 92% 96% 96%
% APOE4 positive 26% 25% 27% 27% 25% 20%
% AD 41% 38% 40% 46% 45% 33%
Open in a new tab
1

Values are mean ± SD or percentage as indicated.

2

NHW=non-Hispanic White

Fig. 1.

Fig. 1

Open in a new tab

This figure illustrates associations with each outcome for VEGF family members that had associations which passed correction for all tests. On the x-axis for each gene are the datatypes, with the general type on the bottom and the specific region or method on the top. The outcomes are listed on the y-axis. The color of each square represents the strength of the association with more red colors indicating positive associations, blue colors indicating negative associations, and grey indicating a beta closer to zero. Boxes where there was not measured expression for a given gene are left white, without color. For proteins measured by multiple probes, the strongest association is presented. The numbers indicated on the color bar represent scaled values rather than raw betas. Associations which reached nominal significance are indicated with a black dot, while associations which passed correction for all 560 tests are indicated with a black asterisk.

3.1. VEGF Inter-tissue Correlations

To aid in the interpretation of region-specific RNA analyses discussed below, we first assessed the correlation of bulk RNA expression of each VEGF gene across the three available brain tissues in overlapping samples (n=127). Most VEGF genes were moderately positively correlated across tissues (r>0.30), although VEGFD inter-tissue correlations between DLPFC and the other two tissues were weaker (PCC r=0.21, CN r=0.09). Interestingly, while the correlation between FLT4 DLPFC and PCC expression was positive (r=0.60), CN FLT4 expression was negatively correlated with PCC and DLPFC expression (r=−0.11 and −0.12, respectively; Supplementary Fig. 2).

3.2. VEGF Associations with Cognition and Diagnosis

Corresponding to our previous study, VEGFB was associated with worse cognitive outcomes in DLPFC (Fig. 1), and this negative effect was also observed in two additional brain regions, CN and PCC. Specifically, higher VEGFB mRNA expression associated with lower cognition (DLPFC p=0.01; PCC p=0.01) and with faster cognitive decline (DLPFC p=0.01 [Fig. 2C]; PCC p=0.02; and CN p=0.02). Results remained significant in sensitivity analyses except for PCC associations when including technical covariates (p>0.11, all other regions p<0.05). Interestingly, snRNA data in DLPFC suggested that cognitive associations with VEGFB were driven by microglia, OPCs, oligodendrocytes, and endothelial cells (p<0.03, endothelial p=5.50×10−4*, Fig. 2D). No significant effects on cognition were observed in pericytes. At the protein level, VEGFB SRM-determined abundance was robustly related to cognition in the same direction as RNA results (cross-sectional β=−0.31, p=3.50×10−5*; longitudinal β=−0.04, p=3.70×10−8*, Fig. 3D). Higher VEGFB protein was also associated with higher risk of clinical AD diagnosis (p=0.004).

Fig. 2.

Fig. 2

Open in a new tab

This figure illustrates the congruence between snRNA results and previously reported associations between VEGFB and FLT1 DLPFC expression. Panels A and C have bulk DLPFC RNA expression (n=531) of FLT1 and VEGFB, respectively, on the x-axis while panels B and D have normalized snRNA expression of each gene on the x-axis. All plots have global cognition decline on the y-axis. Notes: In Panel B, FLT1 expression has been scaled and normalized for illustrative purposes. In Panel D, each participant is represented by three data points, one data point for each of the four cell types presented.

Fig. 3.

Fig. 3

Open in a new tab

Panels A, B, and C are illustrating FLT1 DLPFC protein expression associations with annual change in global cognition, AD diagnosis, and amyloid load, respectively. Panels D, E, and F are illustrating VEGFB DLPFC protein expression associations with annual change in global cognition, amyloid load, and tau tangle density, respectively. Abbreviations: NC = normal cognition, MCI = mild cognitive impairment, AD = Alzheimer’s disease

In general, cognitive results for FLT1 mirrored those of VEGFB, with some showing even stronger evidence of association. Higher FLT1 mRNA expression related to faster cognitive decline across all three regions (DLPFC p=0.02 [Fig. 2A]; PCC p=0.003; and CN p=0.01) and with worse cognition at death in the PCC (p=0.003). Like VEGFB, cognitive associations remained significant in sensitivity analyses, although only DLPFC longitudinal cognitive association survived adjustment for technical covariates (p=0.04). snRNA results indicated that, like VEGFB, FLT1 cognitive associations were present in microglia (p<0.003, all other cell types p>0.07, Fig. 2B), with no significant associations in other tested cell types. Lastly, FLT1 proteomic results recapitulated those in mRNA, with strong negative associations with cognitive decline in both SRM- and TMT-MS determined protein levels (SRM p=2.00×10−4*; TMT p=1.90×10−4*, Fig. 3A). These cognitive associations extend to associations with diagnosis, with higher levels of both mRNA (PCC p=0.03; DLPFC inhibitory neurons p=0.01) and protein (SRM p=0.01*; TMT p=0.001*, Fig. 3B) associated with higher AD risk.

Another VEGF ligand showed evidence of a detrimental effect on cognition. Higher PCC and CN VEGFD expression was associated with faster cognitive decline (PCC p=0.01; CN p=0.02). Associations remained significant in sensitivity analyses. As in previous analyses, no VEGFD associations were observed with cross-sectional cognition in DLPFC (p>0.23). Neither snRNA nor proteomic data were available for VEGFD.

In contrast to the aforementioned VEGF genes, four family members (NRP2, NRP1, FLT4, and KDR) displayed beneficial associations with cognitive outcomes. Specifically, higher NRP2 expression was associated with better cognition (CN p=0.01), slower cognitive decline (CN p=0.02; PCC p=0.03), and less clinical AD (CN p=0.001*; PCC p=0.045). Cognitive associations in the PCC were no longer statistically significant after correction for technical covariates (p=0.42). NRP2 snRNA models did not identify any cell-specific associations with cognition (p>0.10). In contrast to mRNA results, higher NRP2 protein levels were associated with faster cognitive decline (SRM p=0.048; TMT p=0.03).

Similarly, higher SRM-measured NRP1 protein levels were associated with better cognition (p=0.02) and slower decline before death (p=0.003). The longitudinal association remained when covarying for AD pathology (p<0.01). Further, snRNA NRP1 expression was associated with AD diagnosis, such that MCI and AD participants had lower NRP1 expression in excitatory (p=0.01) and inhibitory neurons (p=0.04) and lower NRP1 protein levels compared to cognitively normal participants (p=0.02).

For FLT4 and KDR, positive associations with both cross-sectional (FLT4 p=2.40×10−5*; KDR p=0.01) and longitudinal cognition (FLT4 p=1.10×10−6*, Fig. 4A; KDR p=0.02) were observed in the CN, with FLT4 showing the strongest mRNA association with cognition among all VEGF family members and a strong association with AD diagnosis in the CN (β=0.37, p=2.40×10−5*, Fig. 4B). All associations remained significant in sensitivity analyses, except KDR when adjusting for oligodendrocyte cell fraction (p=0.08, all other p<0.05). We were unable to replicate previous associations between FLT4 mRNA and cognition in DLPFC, perhaps due to a smaller sample size, but the direction of effect was consistent (cross-sectional β=−0.12; longitudinal β=−0.01). Protein-level data were unavailable for these two genes and no significant snRNA associations were observed (p>0.08).

Fig. 4.

Fig. 4

Open in a new tab

Panel A shows the association between mRNA FLT4 expression in the caudate nucleus and annual change in global cognition. Panel B illustrates FLT4 expression in the caudate nucleus with diagnostic status. Abbreviations: NC = normal cognition, MCI = mild cognitive impairment, AD = Alzheimer’s disease

No associations with cognition or diagnosis were observed for PGF, VEGFA, or VEGFC in any brain region, though there was very limited snRNA and proteomic data available for these genes.

3.3. VEGF Associations with AD Pathology

Next, we explored VEGF associations with post-mortem amyloid and tau pathology. VEGFB and FLT1 proteins showed the strongest associations with AD pathology. Specifically, higher VEGFB and FLT1 SRM-measured protein abundances were robustly related to higher amyloid (p=3.90×10−7* [Fig. 3E] and p=5.30×10−7*, respectively) and tau (p=1.10×10−10* [Fig. 3F] and p=2.30×10−4*, respectively) at autopsy. Moreover, these robust FLT1 associations were replicated in TMT as well (amyloid p=3.30×10−25*, Fig. 3C; tau p=8.20×10−10*). A few weak associations with AD pathology for these two were also observed in snRNA data. Specifically, VEGFB and FLT1 oligodendrocyte levels were associated with higher amyloid (VEGFB p=0.03; FLT1 p=0.003) as was FLT1 CN mRNA (p=0.04). VEGFB OPC expression was associated with higher tau levels (p=0.03). Higher VEGFD CN expression also related to more amyloid pathology (p=0.01), remaining significant in all sensitivity analyses (p<0.03).

VEGF family members that showed evidence of protective associations with pathology included FLT4, KDR, VEGFA, NRP1, and PGF. Specifically, higher FLT4 CN expression related to lower amyloid and tau (p<0.03). Further, higher KDR PCC expression was associated with lower amyloid and tau (p<0.01). All remained significant when in sensitivity analyses, except CN FLT4 on amyloid when accounting for cell-type composition (p=0.13, all others p<0.05). In snRNA-seq analyses, KDR inhibitory neuron expression was negatively associated with amyloid (p=0.01) and tau (p=0.03). Similarly, higher astrocytic VEGFA expression was related to lower tau (p=0.04), and higher SRM-measured NRP1 related to lower amyloid (p=0.002). In contrast to previous findings, higher PGF expression in pericytes was nominally associated with lower amyloid burden (p=0.05) though it may be due to the small sample size of pericyte snRNA analyzed. No associations with pathology were observed for VEGFC or NRP2.

3.4. Sex and APOE Interaction Effects

No sex interaction effects or APOE-ε4 interaction effects on any outcome survived correction for multiple comparisons. Comprehensive stratified and interaction results are presented in Supplementary Tables 12–17.

4. Discussion

This multi-omic analysis of the VEGF family in human brain provides strong evidence that higher VEGFB and FLT1 – at both the transcript and protein level – are associated with more AD neuropathology and faster cognitive decline, and our analyses suggest that microglial and endothelial gene expression appear to contribute to these associations. Further, we observed region-specific effects of FLT4 and NRP2, with protective effects observed in deep brain regions but not in the cortex. Together, our findings add to a growing body of literature suggesting VEGF family proteins are viable targets for biomarker development and clinical intervention in AD.

We previously identified the upregulation of VEGFB and FLT1 transcripts in prefrontal cortex of AD patients but did not have sufficient data to clarify the cell type, regional heterogeneity, or proteomic context (Mahoney et al., 2019). Here, we have replicated our original results in an independent prefrontal cortex sample from the same cohort and expanded our findings to overcome some of the previous limitations. Specifically, we illustrate that these associations are present at the protein level, provide evidence that they are not restricted to the frontal cortex, and indicate that these effects in the DLPFC are attributed to microglia and endothelial cells. Though we observed that higher VEGFB expression in endothelial cells is significantly associated with cognitive decline, one significant caveat of this finding is the low sample size of endothelial cell and pericyte snRNA data used in our analyses. Thus, the biological function of FLT1 and VEGFB in AD should be further characterized in these cell types.

Our results provide an interesting picture of changes along the VEGFB/FLT1 signaling cascade in aging and AD. While VEGFB does have known signaling roles that contribute to neurogenesis and neurotrophic activity (Li et al., 2008; Yue et al., 2014), our single nucleus results suggest that alternative roles in microglial proliferation are particularly relevant to AD clinical outcomes. Recent work from others suggests there is increased FLT1 endothelial expression in AD and that this upregulation may be related to increased angiogenesis and immune activation (Lau et al., 2020). We did not observe any significant FLT1 effects in endothelial cells, though the strong microglial signal supports the hypothesis of immune involvement. Additionally, while basal microglial FLT1 expression is low, upon exposure to β-amyloid peptides, microglial expression is dramatically upregulated and facilitates an inflammatory chemotactic response (Ryu et al., 2009). VEGFB/FLT1 signaling is also thought to regulate microglial control of astrocytes in neurodegenerative conditions (Rothhammer et al., 2018). Together, our results and previous findings appear to support a compensatory response in the context of neurodegeneration, though additional experimental confirmation is needed. Moreover, while VEGFB is beneficial for neuronal survival (Caballero et al., 2017; Falk et al., 2011), the overwhelming evidence of detrimental associations of high FLT1 and VEGFB expression suggests that off-target effects by signaling through FLT1 in microglia and oligodendrocytes must be considered if the beneficial effects of VEGFA or VEGFB in neurons are to become viable treatment targets.

Consistent with our previous work, associations for the classic VEGF angiogenic signaling pair (VEGFA-KDR) remain somewhat elusive. Neither were successfully measured at the protein level. We did observe nominal protective KDR associations, specifically with higher CN transcript abundance associated with better cognition, and both higher PCC transcript abundance and DLPFC excitatory neuron expression associated with lower neuropathology. Given that VEGFA can signal through KDR or FLT1, and the robust FLT1 associations outlined above, these two receptors may represent distinct signaling cascades operating in opposing directions. KDR results are consistent with neuroprotective effects of VEGFA in neurons (Cao et al., 2004), while the lack of VEGFA effects remains somewhat perplexing. It may be possible to differentiate the role of VEGFA in the neuron from microglial and endothelial cells as snRNA sample sizes increase. In our previous bulk analyses, we identified an APOE x VEGFA interaction whereby higher levels of VEGFA were associated with worse outcomes among carriers and better outcomes among non-carriers (Moore et al., 2020). In the larger analysis here, we show weak evidence of such an effect, with a nominal interaction in the CN on AD diagnosis that shows a comparable direction of effect, but overall do not find support for APOE x VEGFA interactions. Unfortunately, a comprehensive understanding of VEGFA brain expression changes in aging and AD remain elusive.

Higher neuropilin receptor expression (i.e., NRP1 and NRP2) was associated with better outcomes, although evidence at the protein, transcript, and single cell levels was mixed. While both are co-receptors with well-established roles in angiogenesis (Favier et al., 2006; Lampropoulou and Ruhrberg, 2014), their signaling cascades are quite diverse. For example, NRP2 is involved in lymphangiogenesis signaling through VEGFC-binding (Xu et al., 2010), a process regulated by FLT4. Thus, it is notable that both FLT4 and NRP2 in the CN showed beneficial associations with diagnosis and FLT4 with cognitive performance. This appears to be region-specific, as the opposite was observed in DLPFC mRNA previously (Mahoney et al., 2019) and in DLPFC protein abundance here. Further, CN FLT4 expression is anti-correlated with expression in the two cortical regions, an observation replicated in an independent dataset. Together, these results suggest that lymphatics may be an interesting target for further evaluation specifically in deep brain regions. Further, protective NRP1 associations with diagnosis and pathology were observed in protein. Previous work from our group suggested that NRP1-VEGFA-driven angiogenesis may be beneficial in ε4 non-carriers (Moore et al., 2020). While APOE interactions were not evaluated here, future studies should clarify whether these interactions persist at the protein level. Mechanistic studies will be needed to clarify the role of neuropilin receptors in aging and AD, but present results suggest that higher expression may be beneficial in some contexts.

Though we did not observe any sex interactions in our analyses, sex differences in VEGF biology have been observed in other fields. Sex-specific expression of VEGF has been shown to affect endothelial progenitor differentiation (Randolph et al., 2019) and angiogenic responses (Baggio et al., 2022), which may have implications in AD and vascular dementia (Tarkowski et al., 2002; Trigiani et al., 2022). Future studies further examining sex and VEGF family biology may provide additional insights to the AD field as more data about sex as a biological variable in AD is discovered.

This comprehensive study of the VEGF family in aging and disease includes many strengths. ROS/MAP is a well-characterized cohort with extensive follow-up, detailed longitudinal cognitive data, and comprehensive neuropathological data with multi-level measures of VEGF. In addition, this study reports both novel discoveries and independent replication of VEGF family effects on AD pathology and cognition at both the transcriptomic and proteomic level. We were also able to examine cell-type specific expression of VEGF family genes, providing important hypotheses about cellular context which can then be explored with more comprehensive data (e.g. snRNA with larger sample sizes) and histopathological confirmation.

Although post-mortem data is clearly a strength of this study, it is also a limitation, precluding statements of causality or directionality. ROS/MAP is also enriched for highly educated, non-Hispanic white individuals, limiting generalizability to more representative populations. Further, while we were able to examine VEGF family genes across multiple brain regions leveraging bulk RNAseq data, one significant limitation is that our cell-type specific analyses and proteomic analyses are limited to the dorsolateral prefrontal cortex. Additionally, while we measured multiple peptides, we were unable to differentiate functional proteoforms that might have provided new insights into protein function. Future work comprehensively assessing splicing isoforms and proteoforms holds promise for additional biological insight. Finally, while the present manuscript adds to growing evidence that the VEGFB/FLT1 axis is critical in AD, it is notable that genome-wide association studies and rare variant analyses have not uncovered genetic evidence for these loci. Future work should also comprehensively assess genetic variants that modulate these genes to gain further insight into the mechanistic pathway.

In sum, this study has provided the most extensive characterization of the VEGF signaling family in cognitive aging and AD. Our results provide support for the for the role of the VEGFB-FLT1 signaling axis in AD, particularly in glial cells. Future studies exploring differential effects of VEGF family proteoforms, leveraging larger snRNA samples, and exploring APOE interactions will be critical to moving toward using VEGF family members as biomarkers or treatment targets in AD.

Supplementary Material

Supplementary Figures
Supplementary Tables

Acknowledgements and Funding

The results published here are in whole or in part based on data obtained from the AD Knowledge Portal (https://adknowledgeportal.synapse.org). Study data were provided by the Rush Alzheimer’s Disease Center, Rush University Medical Center, Chicago. Data collection was supported through funding by NIA grants P30AG10161 (ROS), R01AG15819 (ROSMAP; genomics and RNAseq), R01AG17917 (MAP), R01AG30146, R01AG36042 (5hC methylation, ATACseq), RC2AG036547 (H3K9Ac), R01AG36836 (RNAseq), R01AG48015 (monocyte RNAseq) RF1AG57473 (single nucleus RNAseq), U01AG32984 (genomic and whole exome sequencing), U01AG46152 (ROSMAP AMP-AD, targeted proteomics), U01AG46161(TMT proteomics), U01AG61356 (whole genome sequencing, targeted proteomics, ROSMAP AMP-AD), the Illinois Department of Public Health (ROSMAP), and the Translational Genomics Research Institute (genomic). Additional phenotypic data can be requested at www.radc.rush.edu.

The Genotype-Tissue Expression (GTEx) Project was supported by the Common Fund of the Office of the Director of the National Institutes of Health, and by NCI, NHGRI, NHLBI, NIDA, NIMH, and NINDS. The data used for the validation analyses described in this manuscript were obtained from dbGaP accession number phs000424.GTEx.v8.p2 on 11/13/2019.

Additional support includes K01-AG049164, R01-AG059716, R01-AG061518, R21-AG05994, K12-HD043483, K24-AG046373, HHSN311201600276P, S10-OD023680, R01-AG034962, R01-NS100980, R01-AG056534, P30-AG010161, R01-AG15819, R01-AG17917, U01-AG46152, Vanderbilt Clinical Translational Science Award (UL1-TR000445), the Vanderbilt Memory and Alzheimer’s Center, and the Vanderbilt Alzheimer’s Disease Research Center (P20-AGAG068082).

Footnotes

Conflicts of Interest

Dr. Hohman serves on the advisory board for Vivid Genomics.

References

  1. Andreev VP, Petyuk VA, Brewer HM, Karpievitch YV, Xie F, Clarke J, Camp D, Smith RD, Lieberman AP, Albin RL, Nawaz Z, El Hokayem J, Myers AJ, 2012. Label-free quantitative LC-MS proteomics of Alzheimer’s disease and normally aged human brains. Journal of proteome research 11(6), 3053–3067. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Baggio C, Boscaro C, Oliviero F, Trevisi L, Ramaschi G, Ramonda R, Bolego C, Cignarella A, 2022. Gender differences and pharmacological regulation of angiogenesis induced by synovial fluids in inflammatory arthritis. Biomedicine & Pharmacotherapy 152, 113181. [DOI] [PubMed] [Google Scholar]
  3. Bennett DA, Buchman AS, Boyle PA, Barnes LL, Wilson RS, Schneider JA, 2018. Religious Orders Study and Rush Memory and Aging Project. Journal of Alzheimer’s disease : JAD 64(s1), S161–S189. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Caballero B, Sherman SJ, Falk T, 2017. Insights into the Mechanisms Involved in Protective Effects of VEGF-B in Dopaminergic Neurons. Parkinson’s Disease 2017, 4263795. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Cain A, Taga M, McCabe C, Hekselman I, White CC, Green G, Rozenblatt-Rosen O, Zhang F, Yeger-Lotem E, Bennett DA, Yang H-S, Regev A, Menon V, Habib N, De Jager PL, 2020. Multi-cellular communities are perturbed in the aging human brain and with Alzheimer’s disease. bioRxiv, 2020.2012.2022.424084. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Cao L, Jiao X, Zuzga DS, Liu Y, Fong DM, Young D, During MJ, 2004. VEGF links hippocampal activity with neurogenesis, learning and memory. Nature Genetics 36, 827. [DOI] [PubMed] [Google Scholar]
  7. Dhondt J, Peeraer E, Verheyen A, Nuydens R, Buysschaert I, Poesen K, Van Geyte K, Beerens M, Shibuya M, Haigh JJ, Meert T, Carmeliet P, Lambrechts D, 2011. Neuronal FLT1 receptor and its selective ligand VEGF-B protect against retrograde degeneration of sensory neurons. FASEB journal : official publication of the Federation of American Societies for Experimental Biology 25(5), 1461–1473. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Falk T, Yue X, Zhang S, McCourt AD, Yee BJ, Gonzalez RT, Sherman SJ, 2011. Vascular endothelial growth factor-B is neuroprotective in an in vivo rat model of Parkinson’s disease. Neuroscience Letters 496(1), 43–47. [DOI] [PubMed] [Google Scholar]
  9. Favier B, Alam A, Barron P, Bonnin J, Laboudie P, Fons P, Mandron M, Herault JP, Neufeld G, Savi P, Herbert JM, Bono F, 2006. Neuropilin-2 interacts with VEGFR-2 and VEGFR-3 and promotes human endothelial cell survival and migration. Blood 108(4), 1243–1250. [DOI] [PubMed] [Google Scholar]
  10. Garcia KO, Ornellas FLM, Martin PKM, Patti CL, Mello LE, Frussa-Filho R, Han SW, Longo BM, 2014. Therapeutic effects of the transplantation of VEGF overexpressing bone marrow mesenchymal stem cells in the hippocampus of murine model of Alzheimer’s disease. Frontiers in aging neuroscience 6, 30–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Harris R, Miners JS, Allen S, Love S, 2018. VEGFR1 and VEGFR2 in Alzheimer’s Disease. Journal of Alzheimer’s Disease(Preprint), 1–12. [DOI] [PubMed] [Google Scholar]
  12. Hohman TJ, Bell SP, Jefferson AL, 2015. The Role of Vascular Endothelial Growth Factor in Neurodegeneration and Cognitive Decline: Exploring Interactions With Biomarkers of Alzheimer Disease. JAMA Neurology 72(5), 520–529. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Huang L, Jia J, Liu R, 2013. Decreased serum levels of the angiogenic factors VEGF and TGF-β1 in Alzheimer’s disease and amnestic mild cognitive impairment. Neuroscience letters 550, 60–63. [DOI] [PubMed] [Google Scholar]
  14. Johnson ECB, Dammer EB, Duong DM, Ping L, Zhou M, Yin L, Higginbotham LA, Guajardo A, White B, Troncoso JC, Thambisetty M, Montine TJ, Lee EB, Trojanowski JQ, Beach TG, Reiman EM, Haroutunian V, Wang M, Schadt E, Zhang B, Dickson DW, Ertekin-Taner N, Golde TE, Petyuk VA, De Jager PL, Bennett DA, Wingo TS, Rangaraju S, Hajjar I, Shulman JM, Lah JJ, Levey AI, Seyfried NT, 2020. Large-scale proteomic analysis of Alzheimer’s disease brain and cerebrospinal fluid reveals early changes in energy metabolism associated with microglia and astrocyte activation. Nat Med 26(5), 769–780. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Koch S, Claesson-Welsh L, 2012. Signal transduction by vascular endothelial growth factor receptors. Cold Spring Harb Perspect Med 2(7), a006502. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Lampropoulou A, Ruhrberg C, 2014. Neuropilin regulation of angiogenesis. Biochem Soc Trans 42(6), 1623–1628. [DOI] [PubMed] [Google Scholar]
  17. Lange C, Storkebaum E, de Almodóvar CR, Dewerchin M, Carmeliet P, 2016. Vascular endothelial growth factor: a neurovascular target in neurological diseases. Nature Reviews Neurology 12, 439. [DOI] [PubMed] [Google Scholar]
  18. Lau S-F, Cao H, Fu AKY, Ip NY, 2020. Single-nucleus transcriptome analysis reveals dysregulation of angiogenic endothelial cells and neuroprotective glia in Alzheimer’s disease. Proceedings of the National Academy of Sciences 117(41), 25800. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Lee HK, Chauhan SK, Kay E, Dana R, 2011. Flt-1 regulates vascular endothelial cell migration via a protein tyrosine kinase-7-dependent pathway. Blood 117(21), 5762–5771. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lee J, Gray A, Yuan J, Luoh SM, Avraham H, Wood WI, 1996. Vascular endothelial growth factor-related protein: a ligand and specific activator of the tyrosine kinase receptor Flt4. Proc Natl Acad Sci U S A 93(5), 1988–1992. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Leung YY, Toledo JB, Nefedov A, Polikar R, Raghavan N, Xie SX, Farnum M, Schultz T, Baek Y, Deerlin VV, Hu WT, Holtzman DM, Fagan AM, Perrin RJ, Grossman M, Soares HD, Kling MA, Mailman M, Arnold SE, Narayan VA, Lee VM, Shaw LM, Baker D, Wittenberg GM, Trojanowski JQ, Wang LS, 2015. Identifying amyloid pathology-related cerebrospinal fluid biomarkers for Alzheimer’s disease in a multicohort study. Alzheimers Dement (Amst) 1(3), 339–348. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Li Y, Zhang F, Nagai N, Tang Z, Zhang S, Scotney P, Lennartsson J, Zhu C, Qu Y, Fang C, 2008. VEGF-B inhibits apoptosis via VEGFR-1–mediated suppression of the expression of BH3-only protein genes in mice and rats. The Journal of clinical investigation 118(3), 913. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Logsdon BA, Perumal TM, Swarup V, Wang M, Funk C, Gaiteri C, Allen M, Wang X, Dammer E, Srivastava G, Mukherjee S, Sieberts SK, Omberg L, Dang KD, Eddy JA, Snyder P, Chae Y, Amberkar S, Wei W, Hide W, Preuss C, Ergun A, Ebert PJ, Airey DC, Carter GW, Mostafavi S, Yu L, Klein H-U, Collier DA, Golde T, Levey A, Bennett DA, Estrada K, Decker M, Liu Z, Shulman JM, Zhang B, Schadt E, De Jager PL, Price ND, Ertekin-Taner N, Mangravite LM, 2019. Meta-analysis of the human brain transcriptome identifies heterogeneity across human AD coexpression modules robust to sample collection and methodological approach. bioRxiv, 510420. [Google Scholar]
  24. Mahoney ER, Dumitrescu L, Moore AM, Cambronero FE, De Jager PL, Koran MEI, Petyuk VA, Robinson RAS, Goyal S, Schneider JA, Bennett DA, Jefferson AL, Hohman TJ, 2019. Brain expression of the vascular endothelial growth factor gene family in cognitive aging and Alzheimer’s disease. Mol Psychiatry. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Mathys H, Davila-Velderrain J, Peng Z, Gao F, Mohammadi S, Young JZ, Menon M, He L, Abdurrob F, Jiang X, Martorell AJ, Ransohoff RM, Hafler BP, Bennett DA, Kellis M, Tsai L-H, 2019. Single-cell transcriptomic analysis of Alzheimer’s disease. Nature 570(7761), 332–337. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Moore AM, Mahoney E, Dumitrescu L, De Jager PL, Koran MEI, Petyuk VA, Robinson RA, Ruderfer DM, Cox NJ, Schneider JA, Bennett DA, Jefferson AL, Hohman TJ, 2020. APOE epsilon4-specific associations of VEGF gene family expression with cognitive aging and Alzheimer’s disease. Neurobiol Aging 87, 18–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Petyuk VA, Qian WJ, Smith RD, Smith DJ, 2010. Mapping protein abundance patterns in the brain using voxelation combined with liquid chromatography and mass spectrometry. Methods 50(2), 77–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Provias J, Jeynes B, 2014. Reduction in vascular endothelial growth factor expression in the superior temporal, hippocampal, and brainstem regions in Alzheimer’s disease. Current neurovascular research 11(3), 202–209. [DOI] [PubMed] [Google Scholar]
  29. Randolph LN, Bao X, Oddo M, Lian XL, 2019. Sex-dependent VEGF expression underlies variations in human pluripotent stem cell to endothelial progenitor differentiation. Scientific Reports 9(1), 16696. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Religa P, Cao R, Religa D, Xue Y, Bogdanovic N, Westaway D, Marti HH, Winblad B, Cao Y, 2013. VEGF significantly restores impaired memory behavior in Alzheimer’s mice by improvement of vascular survival. Nature Scientific Reports 3(2053). [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Rothhammer V, Borucki DM, Tjon EC, Takenaka MC, Chao CC, Ardura-Fabregat A, de Lima KA, Gutiérrez-Vázquez C, Hewson P, Staszewski O, Blain M, Healy L, Neziraj T, Borio M, Wheeler M, Dragin LL, Laplaud DA, Antel J, Alvarez JI, Prinz M, Quintana FJ, 2018. Microglial control of astrocytes in response to microbial metabolites. Nature 557(7707), 724–728. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Ruiz de Almodovar C, Lambrechts D, Mazzone M, Carmeliet P, 2009. Role and therapeutic potential of VEGF in the nervous system. Physiol Rev 89(2), 607–648. [DOI] [PubMed] [Google Scholar]
  33. Ryu JK, Cho T, Choi HB, Wang YT, McLarnon JG, 2009. Microglial VEGF receptor response is an integral chemotactic component in Alzheimer’s disease pathology. J Neurosci 29(1), 3–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Spuch C, Antequera D, Portero A, Orive G, Hernández RM, Molina JA, Bermejo-Pareja F, Pedraz JL, Carro E, 2010. The effect of encapsulated VEGF-secreting cells on brain amyloid load and behavioral impairment in a mouse model of Alzheimer’s disease. Biomaterials 31(21), 5608–5618. [DOI] [PubMed] [Google Scholar]
  35. Sun Y, Jin K, Childs JT, Xie L, Mao XO, Greenberg DA, 2006. Vascular endothelial growth factor-B (VEGFB) stimulates neurogenesis: evidence from knockout mice and growth factor administration. Developmental biology 289(2), 329–335. [DOI] [PubMed] [Google Scholar]
  36. Tarkowski E, Issa R, Sjögren M, Wallin A, Blennow K, Tarkowski A, Kumar P, 2002. Increased intrathecal levels of the angiogenic factors VEGF and TGF-beta in Alzheimer’s disease and vascular dementia. Neurobiol Aging 23(2), 237–243. [DOI] [PubMed] [Google Scholar]
  37. Tillo M, Ruhrberg C, Mackenzie F, 2012. Emerging roles for semaphorins and VEGFs in synaptogenesis and synaptic plasticity. Cell adhesion & migration 6(6), 541–546. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Trigiani LJ, Bourourou M, Lacalle-Aurioles M, Lecrux C, Hynes A, Spring S, Fernandes DJ, Sled JG, Lesage F, Schwaninger M, Hamel E, 2022. A functional cerebral endothelium is necessary to protect against cognitive decline. J Cereb Blood Flow Metab 42(1), 74–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Wang P, Xie ZH, Guo YJ, Zhao CP, Jiang H, Song Y, Zhu ZY, Lai C, Xu SL, Bi JZ, 2011. VEGF-induced angiogenesis ameliorates the memory impairment in APP transgenic mouse model of Alzheimer’s disease. Biochemical and Biophysical Research Communications 411(3), 620–626. [DOI] [PubMed] [Google Scholar]
  40. Wilson RS, Boyle PA, Yu L, Barnes LL, Sytsma J, Buchman AS, Bennett DA, Schneider JA, 2015. Temporal course and pathologic basis of unawareness of memory loss in dementia. Neurology 85(11), 984–991. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Xu Y, Yuan L, Mak J, Pardanaud L, Caunt M, Kasman I, Larrivée B, Del Toro R, Suchting S, Medvinsky A, Silva J, Yang J, Thomas J-L, Koch AW, Alitalo K, Eichmann A, Bagri A, 2010. Neuropilin-2 mediates VEGF-C-induced lymphatic sprouting together with VEGFR3. The Journal of cell biology 188(1), 115–130. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Yue X, Hariri DJ, Caballero B, Zhang S, Bartlett MJ, Kaut O, Mount DW, Wullner U, Sherman SJ, Falk T, 2014. Comparative study of the neurotrophic effects elicited by VEGF-B and GDNF in preclinical in vivo models of Parkinson’s disease. Neuroscience 258, 385–400. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figures
NIHMS1872235-supplement-Supplementary_Figures.docx (126.5KB, docx)
Supplementary Tables
NIHMS1872235-supplement-Supplementary_Tables.pdf (321.5KB, pdf)

RESOURCES