Alzheimer’s Disease Progression in the 5×FAD Mouse Captured with a Multiplex Gene Expression Array

Abstract

Background:

Alzheimer’s disease (AD) is an incurable complex neurodegenerative condition with no new therapies licensed in the past 20 years. AD progression is characterized by the up- and downregulation of distinct biological processes that can be followed through the expression level changes of associated genes and gene networks.

Objective:

Our study aims to establish a multiplex gene expression tracking platform to follow disease progression in an animal model facilitating the study of treatment paradigms.

Methods:

We have established a multiplex platform covering 47 key genes related to immunological, neuronal, mitochondrial, and autophagy cell types and processes that capture disease progression in the 5×FAD mouse model.

Results:

We show that the immunological response is the most pronounced change in aged 5×FAD mice (8 months and above), and in agreement with early stage human disease samples, observe an initial downregulation of microglial genes in one-month-old animals. The less dramatic downregulation of neuronal and mitochondrial gene sets is also reported.

Conclusion:

This study provides the basis for a quantitative multi-dimensional platform to follow AD progression and monitor the efficacy of treatments in an animal model.

Keywords

Alzheimer’s disease biomarker 5×FAD mouse transcriptional profiling

INTRODUCTION

Alzheimer’s disease (AD), like other neurodegenerative disorders, is a multifactorial disease which is thought to be triggered by a combination of genetic, epigenetic, and environmental factors. The complexity of AD and long-time scale of disease progression has made therapeutic intervention particularly challenging. The state of drug discovery in AD is highlighted by there being no new therapies entering the market over the last 20 years. The relatively little success in targeting specific genes associated with the disease, such as anti-amyloid antibody therapies [1, 2] and BACE inhibitors [3], has led many groups to explore non-target-based techniques [4]. One such approach is founded on the disease transcriptome with the simple hypothesis that drugs tending to perturb gene expression in a reverse sense to that seen in disease may consequently act to ameliorate pathology [5, 6]. This methodology has met with some success [7 –13] since its first publication centering on the connectivity map (CMAP) database of compound associated expression profiles [14]. The application of this approach to the area of complex disorders such as AD and neurodegeneration in general is bolstered by the observation that drugs with established neuroprotective activities tend to have transcriptional profiles anti-correlating with those observed in AD [15]. It is therefore likely that transcription-based approaches may offer an alternative avenue for drug discovery in AD, especially given the wealth of transcriptional data available [16 –19].

The initial testing ground for compound candidates is facilitated by animal models and there has been a long and prolific endeavor in this regard resulting in multiple alternative designs usually consisting of the introduction of gene cassettes encoding AD predisposing mutations [20, 21]. A simple model search on Alzforum (http://www.alzforum.org) identifies 168 currently used research murine models of AD (164 mouse and 4 rat models). The introduced mutations in these models vary from APP to APOE, PSEN, tau, and TREM2 among others, reflecting distinct aspects of the pathology of the disease. The expressed isoforms of these genes also vary, as do the promoters driving expression and the genetic background of the strain.

In this work, we show that transcriptional analysis points to the five familial AD (5×FAD) mutant mouse model [21] being the closest to human disease. 5×FAD transgenic mice co-overexpress and co-inherit FAD mutant forms of human amyloid precursor protein (APP) (the Swedish mutation: K670N, M671L; the Florida mutation: I716V; the London mutation: V717I) and presenilin 1 (PS1) (M146L; L286V) transgenes under transcriptional control of the neuron-specific mouse Thy-1 promoter (Tg6799 line). These transgenic mice exhibit intra-neuronal Aβ accumulation at 1.5 months of age leading to amyloid deposition and gliosis shortly thereafter. Progressive neurodegeneration leads to loss of pyramidal neurons and loss of synaptic markers with increasing age. This is further accompanied by an increased expression of the neurodegeneration pathway marker p25 from 3 months onward [21].

Previous transcriptomic analyses of the 5×FAD mouse model have consistently revealed that the strongest transcriptional alterations occur in the neuroinflammation and immune-related gene networks when compared to non-transgenic mice [22 –25]. This distinct differential gene expression persists up to 18 months of age [26, 27], is not related to normal aging and has been further validated in a recently generated 5×FAD mouse model with a mixed background strain [24].

The lack of rigorous biomarkers that reproducibly capture disease progression in mouse models of AD has motivated the present work. It is shown that tracking the expression levels of a relatively small number of genes offers a multi-dimensional quantitative readout of disease progression and hence a measure of the extent to which human pathology is recapitulated in these models. Further, once a promising model emerges, drug candidates can be assessed through their perturbation of the disease-associated changes in gene expression. This can be made more statistically robust by focusing on a representative set of genes that cover distinct features of pathology. Notably, the protracted decline in brain function characterized by catastrophic neuronal atrophy is also paired with pronounced inflammation [28, 29], mitochondrial malfunction [30], and dysregulation of autophagy [31]. Relatively small gene sets tracking these separate features should facilitate an effective quantitative readout of disease state and therapeutic drug effects.

In this study, gene expression changes seen in both the animal model and human disease were filtered into four distinct representative sets covering inflammation, neuronal loss, mitochondrial malfunction and autophagy. We custom designed a panel encompassing 47 genes, covering these four sets and tested the expression levels of these genes in homogenized cortical tissue from 5×FAD mice and age-matched non-transgenic mice. It is perhaps worth emphasizing that in this study, reproducibility is achieved through the transcriptional changes in gene sets, where significance is imparted through the collective expression changes of the gene set and individual genes need not pass statistical thresholds.

The expression level of the gene set is followed over a 15-month period starting with one-month-old animals and reveals an informative dynamic. Disease progression in AD is widely accepted to follow Braak staging [32] and we hypothesized that this staging facilitates a surrogate temporal measure and hence sought to link reported expression changes characteristic of Braak stage to our bead set data. The caveat here is that Braak staging is related to tau pathology and there is no tau pathology in the mutant mouse brains [33]. However, this does not invalidate Braak staging as being a temporal marker for general disease progression. An interesting observation from the 5×FAD progression is that the immune response encoded by microglial gene expression is suppressed at early stage before its later dramatic elevation. Analysis of early stage AD samples across independent cohorts appears to corroborate this finding. Whereas inflammatory responses are the most dramatic in the animal model we also find that neuronal loss follows inflammation and that mitochondrial and autophagy changes occur in the expected direction albeit with less magnitude.

The gene expression dynamics is followed with the Quantigene branched DNA technology which has proven to be efficient and reliable in direct measurement of gene expression in FFPE human tissue [34, 35] and whole blood [36]. The platform relies on direct hybridization and labelling of transcripts without the need of RNA purification or reverse transcription. The present work seeks to establish the utility of this platform as a ‘biomarker tool’ in animal models of disease via the temporal tracking of AD-related gene expression changes in the 5xFAD mouse brain tissue.

METHODS

AD transcriptional profiles

Transcriptional data was obtained from the NCBI GEO repository [37] using the series search portal (https://www.ncbi.nlm.nih.gov/geo/browse/?view=series) with the keywords: Alzheimer’s disease, 5×FAD, and 3×TG. The AD landscape was populated by 44 profiles based on 17 separate series. The 5×FAD set was based on 7 profiles from 4 series and the 3×TG profiles consisted of 8 profiles from 9 series (see Supplementary Table 1 for details). Profiles were defined by significant gene expression changes between disease and normal tissue see [15, 38]. Profiles were compared by a simple regression analysis over genes perturbed in both profiles and the significance measured with the associated Z score.

For the analysis of temporal AD progression, Braak and Clinical Dementia Rating (CDR) stages were used as time surrogates. Out of the 17 AD associated series, only four (GSE1297, GSE48350, GSE84422, and GSE106241) had explicit full range Braak assignments and one had CDR assignments (GSE84422). For each gene, a linear mixed model was generated with gender, ethnicity, and brain region as covariates, when present. Specifically, the model is $y_{i} = α x_{i} + \sum_{n} β^{n} z_{i}^{n} + ε_{i}$ , where y_i is the gene expression level for the ith sample, x_i is the disease state (Braak stage or CRD level), $z_{i}^{n} 0, 1$ is the matrix of the categorical calls (brain region, gender, and ethnicity) with unit entries when the ith sample has the nth categorical and ɛ _i is the remaining error term. The residual correlation of gene expression with disease stage marker constituted the temporal disease profile. Here, the amount a given gene contributes to the disease evolution is measured by the regression Z score.

Bead profiles

The expression levels were first normalized to the geometric mean of the housekeeper genes with ${\tilde{e}}_{i} = e_{i} / {(\prod_{j \in HK} e_{j})}^{\frac{1}{HKn}}$ , where e_i is the expression level of the ith gene and HK are the housekeeping set numbering HKn genes, ACTB and GAPDH in the present case. Expression changes were defined as $2 \frac{〈 d 〉 - 〈 n 〉}{〈 d 〉 + 〈 n 〉}$ , were $〈 d 〉$ and $〈 r 〉$ are the average age matched disease and normal levels. Note, the maximum and minimum fold values are 2 and -2 respectively. Significance at the individual gene and time level is measured by a student’s t-test. Regression significance is quantified by the associated Z score. The significance in the difference between pairs of regression slopes is measured by the Dupont Plummer test [39].

Animal husbandry

All animal care and experimental procedures conformed to the UK Animals (Scientific Procedures) Act, 1986 and every effort was made to minimize animal numbers and suffering. Experiments were approved by the King’s College London Ethics committee. All animals were subjected to normal animal housing conditions (21±1°C; 12 : 12 light/dark cycle) and had access to food and water ad libitum. 5×FAD lines (C57BL/6 genetic background) were purchased from Jackson Laboratory and were maintained by crossing heterozygous transgenic mice with wild-type (WT) C57BL/6 breeders.

Animal timepoints and sacrifice

5×FAD male mice and male WT littermates were sacrificed at these different age points: 1 month (3,3), 3 months (6,6), 8 months (10,10), 11 months (4,8), and 15 months (3,3). The numbers in brackets refer to the animal numbers (WT, 5×FAD). All males were sacrificed by cervical dislocation and both left and right cortices dissected out following hemi-section. Tissue was flash frozen in isopentane and dry ice and stored at –80°C until homogenized.

QuantiGene Plex assay

The QuantiGene Plex assay was carried out as described in the User Manual (QuantiGene 2.0 Plex Assay, ThermoFisher Scientific). Briefly, a working bead mixture containing the custom-designed capture beads and probe set was pipetted into wells of a hybridization plate and a sample of neat lysate was added (15 mg of frozen tissue, homogenized using the QuantiGene 2.0 Sample processing kit). An overnight incubation step allows for target hybridization. The next day the signal was amplified via a series of wash steps with amplifier solution using a hand-held magnetic separation plate. A label probe solution is added and a final wash with Streptavidin phycoerythrin allows for bead signal detection on a Luminex flow cytometer (Luminex Corporation). Bead counts were analyzed to ensure an adequate number of beads were used to obtain a reading (set to at least 100 beads per well) and Median fluorescence intensity (MFI) was measured as a signal readout, proportional to the amount of target RNA present in the well. To control for differences in total RNA, background-subtracted MFI were normalized to the geometric mean of two housekeeping genes, GAPDH and ACTB. The final data are presented as fold-changes of mean 5×FAD versus WT samples for each time point.

RESULTS

Bead set design

The transcriptional landscape of AD can serve as a quantitative phenotype for the disease in the sense that profiles from independent studies show a high degree of correlation and the distinct trajectories of gene pathway sets reveal the complex processes at play during degeneration. Great efforts have been made to generate an animal model of the condition and these generally involve the introduction of human gene cassettes harboring familial AD risk mutations. Transcriptional profiling offers one window through which these models can be gauged against human disease. In Fig. 1, sets of AD profiles are compared to each other and against the 5×FAD together with 3×TG mouse transcriptomes. It is clear that the AD profiles are highly consistent, and it is also apparent that only the 5×FAD model is in good agreement with AD. The 3×TG model profiles are neither self-consistent nor correlate with AD. On the basis of this analysis 5×FAD was the model of choice for this study.

Fig.1

Postmortem human AD brain samples show a high degree of consistency and are correlated with profiles derived from the 5×FAD mouse model but not with those from the 3×TG model. The heat maps are colored according to the Pearson correlation Z scores.

The expression changes tracking AD progression fall into distinct categories. AD is associated with neuroinflammation encoded in the expansion and activation of the microglial population, which can be followed via the expression changes in microglia associated genes. Neuronal loss can likewise be traced through the loss of corresponding gene expression. It is also expected that mitochondrial and autophagy dysregulation seen in AD will be revealed through specific gene expression changes.

The first stage in the gene panel design consists in defining four gene sets covering the biological processes underway in AD. The microglia and neuronal gene sets were defined based on RNAseq expression levels across a panel of brain cell types: astrocytes, neurons, oligodendrocytes, newly formed oligodendrocytes, myelinating oligodendrocytes, microglia, and endothelial cells [40]. Genes associated with the given cell type were defined such that their expression level was over two standard deviations away from the mean across the other cell types. It is perhaps worth emphasizing here that the genes constituting the microglial set are not absolutely exclusive to microglia and may be detected in other cell types, but it is expected that the expression level of the gene set is a surrogate marker for microglial activation state. The mitochondrial and autophagy sets were based on gene ontology [41] and pathway classification [42].

The second stage in panel selection is based on the observed expression change in the genes belonging to the given sets across the 5×FAD and AD profiles (Fig. 2 and Supplementary Table 2). In the first instance, genes were filtered based on their regulation in the GSE50521 series corresponding to 14/15-month-old 5xFAD mice, which has the greatest overlap with AD. The bead set was populated with genes either having the highest regulated rank in GSE50521 or having a consistent regulation across AD profiles.

Fig.2

The expression levels of genes populating the bead array in the 5×FAD mouse model of AD and postmortem human AD brain samples. The cells are colored according to the relative expression level of disease versus non-diseases samples ( $2 \frac{(d - n)}{(d + n)}$ , where d is the average disease level and n the average non-disease level, note that the range of fold values is then between 2 and –2). The gene sets are representative of the distinct processes characterizing disease progression. Notably, a microglial associated gene level increase is a surrogate for inflammation, neuron associated gene levels trace neurodegeneration and mitochondrial gene expression is expected to capture corresponding deficits seen within AD. Autophagy genes tend to be upregulated with disease, but there is a clear downregulated cohort.

The gene panel consists of: a 15-gene microglial set with a consistent upregulation in the other 5×FAD series as well as the AD profiles; a 10-member neuronal set showing a consistent downregulation; a 9-member downregulated mitochondrial set; an autophagy set with 8 upregulated and 5 downregulated genes. The panel is supplemented with two housekeeping genes ACTB and GAPDH together with a probe for human APP to confirm genotyping.

Expression changes are consistent with AD

The temporal pattern of expression level changes in the mouse model of AD was assayed over a 15-month period with mutant and WT mice sacrificed at 1, 3, 8, 11, and 15 months. The pattern of gene expression changes over the bead set and lifetime of the animal is shown in Fig. 3. The overall agreement with predicted changes can be scored with a simple 2×2 table Fisher exact test counting the number of times genes change in the expected sense. Note that all 47 genes apart from housekeepers are included in the contingency table. In Table 1, the temporal correlation with expected changes is shown with a correlation onset at 8 months getting stronger toward the 15-month stage. Interestingly, there appears to be a slight anti-correlation at early stage with 28 genes going in opposite direction to that expected in AD and only 19 going in the predicted direction. This will be discussed in the next section on the non-linear nature of the inflammatory response.

Fig.3

The gene expression level changes observed on the bead array are in general agreement with those expected in AD. The numbers correspond to the expression levels in 5×FAD relative to WT mouse brains Numbers in bold pass the student’s t-test 95% confidence interval. The agreement with AD is most striking within the inflammatory response. The correlation within the other gene sets is less strong. However, when the gene expression direction of change across all sets is scored against what is expected in AD significance is passed at 8 months (Fig. 4).

Fig.4

The temporal variation of the four gene sets tracking disease progression. The most pronounced changes are associated with the immune response captured by the expression changes in microglial genes (top left). Interestingly as a prelude to the dramatic upregulation of microglial genes, there is an early stage downregulation that is also seen in early stage AD pathology. Neuronal changes (top right) are less dramatic, but consistently down after 3 months with a more pronounced drop at 15 months. Mitochondrial changes (bottom left) are consistently down over the lifetime of the animal, but only moderately so, indicating that either a more sensitive gene set would capture these changes more effectively or that these changes are subtler in the animal mode. The autophagy gene set behaves as expected (bottom right). The separation between the up- and downregulated sets occurs after 11 months, but without reaching significance.

Table 1

The overall direction of gene expression changes observed on the bead array is in accord with those reported in AD samples and the 5×FAD model. The correlation is assessed with an exact Fisher test and significance is passed at 8 months. The numbers correspond to the p-value from the Fisher test with the expected and observed up- and downregulated counts indicated with U and D. In the four possible circumstances (UU, DD, UD, and DU), the first letter corresponds to the expected outcome whilst the second corresponds to the observed outcome

Age (months)	p	UU	DD	UD	DU
1	0.957348	6	13	17	11
3	0.337197	11	15	12	9
8	0.000733	19	16	4	8
11	0.011324	18	14	5	10
15	0.000001	20	20	3	4

Microglial response in man and mouse

The immune response is monitored by following the expression levels of microglial genes. The 15 microglial genes on the bead panel show the expected dramatic upregulation at the 8 to 15-month stage. Interestingly, there is a significant drop in expression at the one-month stage (Figs. 3 and 4). In the light of this unexpected finding, we investigated a possible correlate in AD progression. In human samples one must postulate surrogates for temporal AD progression, such as the Braak pathology measure [32] or the CDR [43]. Expression changes associated with early stage disease would then correlate with the low range Braak/CDR samples and those associated with full blown disease would correlate over the full range of pathology and dementia. In Fig. 5, the direction of microglial gene expression driving early stage and late stage disease are shown. It is clear that there is a significant downregulation of the microglial genes across all four sample/platform sets at the early disease stage and the reverse at the late disease stage. This result is in agreement with what is observed in the 5×FAD model, albeit with a more symmetric pattern in human disease in contrast to the mouse model where the late stage immune response is much more dramatic. The trajectories of four genes driving the immune response are illustrated in Fig. 6. CD84 is part of a five-protein plasma AD biomarker set [44], the inflammatory chemokine CCL3 has been shown to be elevated in AD patient microglia [45], CST7 is a biomarker for disease associated microglia [46], ITGAX or CD11c is a surface marker associated with microglial activation in conjunction with neurodegeneration in multiple animal models [47 –51].

Fig.5

The early stage downregulation of microglial gene expression is mirrored in human AD pathology. The temporal progression of AD can be captured with Braak and/or CDR stage surrogate markers. When the expression changes driving the early and late stage pathology are separated it becomes clear that microglial genes undergo a significant reversal. Early stage disease tends to be associated with a downregulation of microglial genes and late stage with an upregulation. This is the case across independent studies and using either Braak stage or the psychiatric CDR score as time surrogates.

Fig.6

The regression plots for four microglial genes showing a strong upregulation with age in the 5×FAD relative to WT mice. The leucocyte antigen CD84 is an established plasma AD biomarker; CCL3, an inflammatory chemokine, is elevated in AD sampled microglia; Leukocystatin (CST7) is another biomarker for disease associated microglia; and another leucocyte antigen ITGAX or CD11c is associated with microglial activation in conjunction with neurodegeneration in multiple animal models.

Neuronal response

The 10-gene neuronal tracker set average expression relative to WT is consistently negative from three months, but only reaches significance at the 15-month stage (Figs. 3 and 4). Only the ATP triggered channel P2RX5 reaches significance, and this is at 3 and 15 months. A regression analysis over the full animal lifespan reveals neuronal deficits in both the 5×FAD and WT animals (Fig. 7). For example, MAPK8 is strongly downregulated in both WT and 5×FAD. The temporal regression slope for the pituitary adenylate cyclase activating polypeptide (ADCYAP1), an established marker of AD progression [52], in the 5×FAD appears to significantly deviate from that of the WT (Figs. 7 and 8).

Fig.7

The temporal variation in gene expression for the 5×FAD and WT animals. The regression of gene expression levels against time highlight the ageing process seen in the WT and additional changes seen in 5×FAD mice. The numbers correspond to the Pearson regression coefficients, r, the associated Z-score, Z, and the p-value corresponding to the significance in the difference of the regression slopes. It is clear that the inflammatory response only occurs in the 5×FAD mice. Neuronal changes are less easily differentiable between WT and 5×FAD mice at the level of regression. Again, the overall regression sign is in excellent agreement with the expected changes.

Fig.8

Temporal regression plots for two selected genes from the neuronal tracker set. The adenylate cyclase activating polypeptide (ADCYAP1) shows a dramatic downregulation over the lifetime of the in the brains of 5×FAD mouse in contrast there is little age-related decline, see graph on the left. A gene showing a significant decline with age in both WT and 5×FAD mice is MAPK8, shown at right.

Mitochondrial changes

The 9-member mitochondrial tracker set is consistently down from 3 months. However, significance is not reached either at the single gene or group levels (Figs. 3 and 4). The results contribute to the overall correlation with expected changes but are not significant in themselves. This could be a reflection of the less pronounced effect that mitochondrial malfunction has on gene expression or it could be an indication that the gene set chosen is not the most sensitive in this context. These observations will inform the future evolution of the bead set.

Autophagy

The autophagy gene set comprises five genes that are expected to be downregulated and eight that are predicted to have enhanced expression as pathology sets in. Figure 4 shows that from the 11th month, the two sets separate as predicted, but without reaching a significant deviation from zero.

APP

The bead gene set also contained a probe for human APP which is expressed in the mutant mouse. This facilitates a significant advantage of the bead platform as it thus enables a simultaneous genotyping of the samples. In particular, the 5×FAD mutant is characterized by the presence of human APP, reflected in the signal from the human specific bead probe. A surprising observation is that in the 5×FAD animals the level of APP increases with age (Fig. 9).

Fig.9

The human APP gene expression goes up with age in the 5×FAD mouse. As expected, the human APP probe effectively distinguishes mutant from WT animals. Whereas the level of APP protein will build up in brain tissue over the lifetime of the animal, the elevation in transcript level is an unexpected observation.

DISCUSSION

We have generated a gene expression analysis platform which recapitulates transcription changes observed in four major processes and cell types altered in AD. The platform was tested on the 5×FAD model of disease which we established to be the most representative of human pathology using published transcriptional data. We observed changes in microglial and neuronal cell types as well as mitochondrial and autophagic processes over the lifetime of the mice (1 to 15 months) and found that these changes mirror changes observed in the human condition.

Bead platform profiling of the 5×FAD mouse model reveals the characteristic sustained expression of human APP together with up- and downregulation of gene sets representative of the key biological processes at play in disease progression. Most interestingly, microglial gene expression appeared to be downregulated at the earliest observed stage of one month before undergoing a dramatic upregulation characteristic of inflammation at 8 months through to 15 months. This immune switch is also characteristic of human disease samples when Braak stage or cognitive decline is used as a surrogate for disease stage. Whereas the immune response associated with overt disease is well established [28, 29], compromised microglial function may play a role in AD as it is known that microglia can protect the brain by clearing debris but can damage it by engulfing otherwise healthy synapses [53]. Immune-deficient 5×FAD mice have in fact been proven to exhibit overtly increased Aβ levels, concomitant with alterations in chemokine or cytokine signaling and microglial associated pathways, suggesting that the adaptive immune system has a direct role in AD progression which is likely mediated via alterations in Aβ clearance [25]. Interestingly, a recent study highlighted how a normal age-related loss in the ability to clear debris might contribute to the onset of AD [54]. Detailed single-cell transcriptional profiling of microglia in control and 5×FAD mice points to the presence of an AD-associated phagocytic cell whose activation is two-step; an initial downregulation of checkpoint genes is followed by an upregulation of genes associated with microglial activation [55]. A bi-phasic peak in microglial activation has also been observed via PET and MRI scans in AD patients and controls, where a reduction in microglial activation observed in MCI patients, followed by an increase in activation is observed in later stage AD [56]. In this light, our observation of a two-phase transcriptional signature may speak to the fidelity of the bead platform in its ability to capture this turning point in disease progression; however, further work is required to link these findings to those observed in human AD.

Individual gene expression changes in the neuronal, mitochondrial, and autophagic gene sets are much more modest in comparison to the microglial set. Neuronal deficits set in at three months but only pass significance at 15 months. Mitochondrial markers were consistently down but not to a significant extent at any of the timepoints. The autophagy response was as expected, with segregation into up- and downregulated gene sets at 15 months. This could be explained by differential gene expression in these processes occurring in normal aging, and therefore diluting the effective change observed in relation to AD. However, the overall direction of change in these three gene sets still mirror disease-associated changes.

This platform therefore effectively captures the gene expression changes underlying four main aspects of AD in the 5×FAD mouse model. Since this model is widely used in pre-clinical drug discovery studies, we expect this platform will facilitate high-throughput, cost-effective, and quick screening for multi-gene transcriptional analysis. Further, this assay offers distinct advantages in that it does not require RNA isolation as gene expression can be measured in prepared tissue homogenate, greatly diminishing the time required and chances of contamination.

While it was in the interest of this study to look at gene expression levels in the whole cortex, it is still possible to use this platform to investigate differential gene expression in a more delimited setting. It would be of great future interest to evaluate the age-course of gene expression changes in brain regions such as the hippocampus and entorhinal cortex which are sites of early pathological changes and neuronal loss. Further, as an APP overexpression model, the 5×FAD mouse is lacking tau-related pathology and therefore presents a number of inherent limitations in its representation of sporadic AD. Additionally, this study focused on the age-course transcriptional changes in male 5×FAD brains. Female mice are thought to have slightly altered transcriptomic changes in relation to estrogen levels [57] and therefore gender-related changes must be taken into account in future studies. Therefore, one could extend this study to compare gene expression in different genders, cell types, brain regions, and genotypes all in the same experiment, similarly to what we have exemplified by the comparison of different ages.

Importantly, this study highlights the need to investigate transcriptional changes early on in the disease as this could help identify stage-specific targets for novel therapeutics. Interestingly, a recent single-cell transcriptomic analysis has highlighted the fact that large-scale transcriptional changes occur in human AD prior to overt pathological onset, and that such early changes tend to be cell-type specific [58].

Overall, with low operating costs, minimal sample processing, and high-throughput, we expect this platform will prove to be a useful tool for gene expression studies in AD and offer a cost-effective method for measuring transcriptional changes in drug screening studies prior to more extensive and time-consuming behavioral assays.

Conclusions

The gene set expression platform recapitulates the main features of the AD transcriptome in the 5×FAD mouse. The single 96-well plate-based assay highlights progressive AD-associated changes in neuronal, microglial, autophagic, and mitochondrial gene sets over a 15-month period and could be highly applicable as a drug discovery screening platform aiding the identification of novel therapeutics. Further, we highlight that microglial gene expression is suppressed at young age in the 5×FAD mouse, prior to an overt upregulation at the later disease stages. In conclusion, we expect that this transcription-based descriptor provides a more robust and sensitive platform for tracking disease progression especially at the critical early stage of disease.

Footnotes

ACKNOWLEDGMENTS

This work was funded by the Wellcome foundation: A systematic programme to develop and evaluate the best candidate treatments for repositioning as therapies for Alzheimer’s Disease (SMART-AD) reference 102001/Z/13/Z.

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

van Dyck

(2018) Anti-amyloid-beta monoclonal antibodies for Alzheimer’s disease: pitfalls and promise. Biol Psychiatry 83, 311–319.

Honig

, Vellas

, Woodward

, Boada

, Bullock

, Borrie

, Hager

, Andreasen

, Scarpini

, Liu-Seifert

, Case

, Dean

, Hake

, Sundell

, Poole Hoffmann

, Carlson

, Khanna

, Mintun

, DeMattos

, Selzler

, Siemers

(2018) Trial of solanezumab for mild dementia due to Alzheimer’s disease. N Engl J Med 378, 321–330.

Egan

, Kost

, Voss

, Mukai

, Aisen

, Cummings

, Tariot

, Vellas

, van Dyck

, Boada

, Zhang

, Li

, Furtek

, Mahoney

, Harper Mozley

, Mo

, Sur

, Michelson

(2019) Randomized trial of verubecestat for prodromal Alzheimer’s disease. N Engl J Med 380, 1408–1420.

Corbett

, Williams

, Ballard

(2015) Drug repositioning in Alzheimer’s disease. Front Biosci (Schol Ed) 7, 184–188.

Marton

, DeRisi

, Bennett

, Iyer

, Meyer

, Roberts

, Stoughton

, Burchard

, Slade

, Dai

, Bassett

Jr. , Hartwell

, Brown

, Friend

(1998) Drug target validation and identification of secondary drug target effects using DNA microarrays. Nat Med 4, 1293–1301.

Hughes

, Marton

, Jones

, Roberts

, Stoughton

, Armour

, Bennett

, Coffey

, Dai

, He

, Kidd

, King

, Meyer

, Slade

, Lum

, Stepaniants

, Shoemaker

, Gachotte

, Chakraburtty

, Simon

, Bard

, Friend

(2000) Functional discovery via a compendium of expression profiles. Cell 102, 109–126.

Wei

, Twomey

, Lamb

, Schlis

, Agarwal

, Stam

, Opferman

, Sallan

, den Boer

, Pieters

, Golub

, Armstrong

(2006) Gene expression-based chemical genomics identifies rapamycin as a modulator of MCL1 and glucocorticoid resistance. Cancer Cell 10, 331–342.

Zhang

, Ciciriello

, Anjos

, Carissimo

, Liao

, Carlile

, Balghi

, Robert

, Luini

, Hanrahan

, Thomas

(2012) Ouabain mimics low temperature rescue of F508del-CFTR in cystic fibrosis epithelial cells. Front Pharmacol 3, 176.

Sirota

, Dudley

, Kim

, Chiang

, Morgan

, Sweet-Cordero

, Sage

, Butte

(2011) Discovery and preclinical validation of drug indications using compendia of public gene expression data. Sci Transl Med 3, 96ra77.

10.

Dudley

, Sirota

, Shenoy

, Pai

, Roedder

, Chiang

, Morgan

, Sarwal

, Pasricha

, Butte

(2011) Computational repositioning of the anticonvulsant topiramate for inflammatory bowel disease. Sci Transl Med 3, 96ra76.

11.

Chang

, Smith

, Thorpe

, Barratt

, Karim

(2010) Evaluation of phenoxybenzamine in the CFA model of pain following gene expression studies and connectivity mapping. Mol Pain 6, 56.

12.

Kunkel

, Suneja

, Ebert

, Bongers

, Fox

, Malmberg

, Alipour

, Shields

, Adams

(2011) mRNA expression signatures of human skeletal muscle atrophy identify a natural compound that increases muscle mass. Cell Metab 13, 627–638.

13.

Walf-Vorderwulbecke

, Pearce

, Brooks

, Hubank

, van den Heuvel-Eibrink

, Zwaan

, Adams

, Edwards

, Bartram

, Samarasinghe

, Ancliff

, Khwaja

, Goulden

, Williams

, de Boer

, Williams

(2018) Targeting acute myeloid leukemia by drug-induced c-MYB degradation. Leukemia 32, 882–889.

14.

Lamb

, Crawford

, Peck

, Modell

, Blat

, Wrobel

, Lerner

, Brunet

, Subramanian

, Ross

, Reich

, Hieronymus

, Wei

, Armstrong

, Haggarty

, Clemons

, Wei

, Carr

, Lander

, Golub

(2006) The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science 313, 1929–1935.

15.

Williams

(2012) A searchable cross-platform gene expression database reveals connections between drug treatments and disease. BMC Genomics 13, 12.

16.

Blalock

, Geddes

, Chen

, Porter

, Markesbery

, Landfield

(2004) Incipient Alzheimer’s disease: microarray correlation analyses reveal major transcriptional and tumor suppressor responses. Proc Natl Acad Sci U S A 101, 2173–2178.

17.

Wang

, Roussos

, McKenzie

, Zhou

, Kajiwara

, Brennand

, De Luca

, Crary

, Casaccia

, Buxbaum

, Ehrlich

, Gandy

, Goate

, Katsel

, Schadt

, Haroutunian

, Zhang

(2016) Integrative network analysis of nineteen brain regions identifies molecular signatures and networks underlying selective regional vulnerability to Alzheimer’s disease. Genome Med 8, 104.

18.

Webster

, Gibbs

, Clarke

, Ray

, Zhang

, Holmans

, Rohrer

, Zhao

, Marlowe

, Kaleem

, McCorquodale

3rd , Cuello

, Leung

, Bryden

, Nath

, Zismann

, Joshipura

, Huentelman

, Hu-Lince

, Coon

, Craig

, Pearson

, NACC-Neuropathology Group, Heward

, Reiman

, Stephan

, Hardy

, Myers

(2009) Genetic control of human brain transcript expression in Alzheimer disease. Am J Hum Genet 84, 445–458.

19.

Readhead

, Haure-Mirande

, Funk

, Richards

, Shannon

, Haroutunian

, Sano

, Liang

, Beckmann

, Price

, Reiman

, Schadt

, Ehrlich

, Gandy

, Dudley

(2018) Multiscale analysis of independent Alzheimer’s cohorts finds disruption of molecular, genetic, and clinical networks by human herpesvirus. Neuron 99, 64–82e67.

20.

Oddo

, Caccamo

, Shepherd

, Murphy

, Golde

, Kayed

, Metherate

, Mattson

, Akbari

, LaFerla

(2003) Triple-transgenic model of Alzheimer’s disease with plaques and tangles: intracellular Abeta and synaptic dysfunction. Neuron 39, 409–421.

21.

Oakley

, Cole

, Logan

, Maus

, Shao

, Craft

, Guillozet-Bongaarts

, Ohno

, Disterhoft

, Van Eldik

, Berry

, Vassar

(2006) Intraneuronal beta-amyloid aggregates, neurodegeneration, and neuron loss in transgenic mice with five familial Alzheimer’s disease mutations: potential factors in amyloid plaque formation. J Neurosci 26, 10129–10140.

22.

Paesler

, Xie

, Hettich

, Siwek

, Ryan

, Schroder

, Papazoglou

, Broich

, Muller

, Trog

, Garthe

, Kempermann

, Weiergraber

, Ehninger

(2015) Limited effects of an eIF2alphaS51A allele on neurological impairments in the 5xFAD mouse model of Alzheimer’s disease. Neural Plast 2015, 825157.

23.

Boeddrich

, Babila

, Wiglenda

, Diez

, Jacob

, Nietfeld

, Huska

, Haenig

, Groenke

, Buntru

, Blanc

, Meier

, Vannoni

, Erck

, Friedrich

, Martens

, Neuendorf

, Schnoegl

, Wolfer

, Loos

, Beule

, Andrade-Navarro

, Wanker

(2019) The anti-amyloid compound DO1 decreases plaque pathology and neuroinflammation-related expression changes in 5xFAD transgenic mice. Cell Chem Biol 26, 109–120e107.

24.

Neuner

, Heuer

, Huentelman

, O’Connell

KMS

, Kaczorowski

(2019) Harnessing genetic complexity to enhance translatability of Alzheimer’s disease mouse models: a path toward precision medicine. Neuron 101, 399–411e395.

25.

Marsh

, Abud

, Lakatos

, Karimzadeh

, Yeung

, Davtyan

, Fote

, Lau

, Weinger

, Lane

, Inlay

, Poon

, Blurton-Jones

(2016) The adaptive immune system restrains Alzheimer’s disease pathogenesis by modulating microglial function. Proc Natl Acad Sci U S A 113, E1316–1325.

26.

Landel

, Baranger

, Virard

, Loriod

, Khrestchatisky

, Rivera

, Benech

, Feron

(2014) Temporal gene profiling of the 5XFAD transgenic mouse model highlights the importance of microglial activation in Alzheimer’s disease. Mol Neurodegener 9, 33.

27.

Siwek

, Muller

, Henseler

, Trog

, Lundt

, Wormuth

, Broich

, Ehninger

, Weiergraber

, Papazoglou

(2015) Altered theta oscillations and aberrant cortical excitatory activity in the 5XFAD model of Alzheimer’s disease. Neural Plast 2015, 781731.

28.

Heneka

, Carson

, El Khoury

, Landreth

, Brosseron

, Feinstein

, Jacobs

, Wyss-Coray

, Vitorica

, Ransohoff

, Herrup

, Frautschy

, Finsen

, Brown

, Verkhratsky

, Yamanaka

, Koistinaho

, Latz

, Halle

, Petzold

, Town

, Morgan

, Shinohara

, Perry

, Holmes

, Bazan

, Brooks

, Hunot

, Joseph

, Deigendesch

, Garaschuk

, Boddeke

, Dinarello

, Breitner

, Cole

, Golenbock

, Kummer

(2015) Neuroinflammation in Alzheimer’s disease. Lancet Neurol 14, 388–405.

29.

Calsolaro

, Edison

(2016) Neuroinflammation in Alzheimer’s disease: Current evidence and future directions. Alzheimers Dement 12, 719–732.

30.

Moreira

, Carvalho

, Zhu

, Smith

, Perry

(2010) Mitochondrial dysfunction is a trigger of Alzheimer’s disease pathophysiology. Biochim Biophys Acta 1802, 2–10.

31.

Nixon

(2007) Autophagy, amyloidogenesis and Alzheimer disease. J Cell Sci 120, 4081–4091.

32.

Braak

, Alafuzoff

, Arzberger

, Kretzschmar

, Del Tredici

(2006) Staging of Alzheimer disease-associated neurofibrillary pathology using paraffin sections and immunocytochemistry. Acta Neuropathol 112, 389–404.

33.

Chu

, Cummins

, Sparling

, Tsutsui

, Brideau

, Nilsson

KPR

, Joseph

, Stys

(2017) Axonal and myelinic pathology in 5xFAD Alzheimer’s mouse spinal cord. PLoS One 12, e0188218.

34.

Crevecoeur

, Kaminski

, Rogister

, Foerch

, Vandenplas

, Neveux

, Mazzuferi

, Kroonen

, Poulet

, Martin

, Sadzot

, Rikir

, Klitgaard

, Moonen

, Deprez

(2014) Expression pattern of synaptic vesicle protein 2 (SV2) isoforms in patients with temporal lobe epilepsy and hippocampal sclerosis. Neuropathol Appl Neurobiol 40, 191–204.

35.

Yang

, Maqsodi

, Ma

, Bui

, Crawford

, McMaster

, Witney

, Luo

(2006) Direct quantification of gene expression in homogenates of formalin-fixed, paraffin-embedded tissues. Biotechniques 40, 481–486.

36.

Zheng

, Luo

, McMaster

(2005) Sensitive and quantitative measurement of gene expression directly from peripheral whole blood, without RNA isolation and target amplification. Blood 106, 208b–209b.

37.

Barrett

, Troup

, Wilhite

, Ledoux

, Rudnev

, Evangelista

, Kim

, Soboleva

, Tomashevsky

, Edgar

(2007) NCBI GEO: mining tens of millions of expression profiles–database and tools update. Nucleic Acids Res 35, D760–765.

38.

Williams

(2013) SPIEDw: a searchable platform-independent expression database web tool. BMC Genomics 14, 765.

39.

Dupont

, Plummer

Jr. (1998) Power and sample size calculations for studies involving linear regression. Control Clin Trials 19, 589–601.

40.

Zhang

, Chen

, Sloan

, Bennett

, Scholze

, O’Keeffe

, Phatnani

, Guarnieri

, Caneda

, Ruderisch

, Deng

, Liddelow

, Zhang

, Daneman

, Maniatis

, Barres

, Wu

(2014) An RNA-sequencing transcriptome and splicing database of glia, neurons, and vascular cells of the cerebral cortex. J Neurosci 34, 11929–11947.

41.

Ashburner

, Ball

, Blake

, Botstein

, Butler

, Cherry

, Davis

, Dolinski

, Dwight

, Eppig

, Harris

, Hill

, Issel-Tarver

, Kasarskis

, Lewis

, Matese

, Richardson

, Ringwald

, Rubin

, Sherlock

(2000) Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 25, 25–29.

42.

Fabregat

, Jupe

, Matthews

, Sidiropoulos

, Gillespie

, Garapati

, Haw

, Jassal

, Korninger

, May

, Milacic

, Roca

, Rothfels

, Sevilla

, Shamovsky

, Shorser

, Varusai

, Viteri

, Weiser

, Wu

, Stein

, Hermjakob

, D’Eustachio

(2018) The Reactome pathway knowledgebase. Nucleic Acids Res 46, D649–D655.

43.

Pangman

, Sloan

, Guse

(2000) An examination of psychometric properties of the mini-mental state examination and the standardized mini-mental state examination: implications for clinical practice. Appl Nurs Res 13, 209–213.

44.

Zhao

, Lejnine

, Spond

, Zhang

, Ramaraj

, Holder

, Dai

, Weiner

, Laterza

(2015) A candidate plasma protein classifier to identify Alzheimer’s disease. J Alzheimers Dis 43, 549–563.

45.

Lue

, Walker

, Rogers

(2001) Modeling microglial activation in Alzheimer’s disease with human postmortem microglial cultures. Neurobiol Aging 22, 945–956.

46.

Ofengeim

, Mazzitelli

, Ito

, DeWitt

, Mifflin

, Zou

, Das

, Adiconis

, Chen

, Zhu

, Kelliher

, Levin

, Yuan

(2017) RIPK1 mediates a disease-associated microglial response in Alzheimer’s disease. Proc Natl Acad Sci U S A 114, E8788–E8797.

47.

Chiu

, Morimoto

, Goodarzi

, Liao

, O’Keeffe

, Phatnani

, Muratet

, Carroll

, Levy

, Tavazoie

, Myers

, Maniatis

(2013) A neurodegeneration-specific gene-expression signature of acutely isolated microglia from an amyotrophic lateral sclerosis mouse model. Cell Rep 4, 385–401.

48.

Hickman

, Kingery

, Ohsumi

, Borowsky

, Wang

, Means

, El Khoury

(2013) The microglial sensome revealed by direct RNA sequencing. Nat Neurosci 16, 1896–1905.

49.

Holtman

, Raj

, Miller

, Schaafsma

, Yin

, Brouwer

, Wes

, Moller

, Orre

, Kamphuis

, Hol

, Boddeke

, Eggen

(2015) Induction of a common microglia gene expression signature by aging and neurodegenerative conditions: a co-expression meta-analysis. Acta Neuropathol Commun 3, 31.

50.

Kamphuis

, Kooijman

, Schetters

, Orre

, Hol

(2016) Transcriptional profiling of CD11c-positive microglia accumulating around amyloid plaques in a mouse model for Alzheimer’s disease. Biochim Biophys Acta 1862, 1847–1860.

51.

Wang

, Cella

, Mallinson

, Ulrich

, Young

, Robinette

, Gilfillan

, Krishnan

, Sudhakar

, Zinselmeyer

, Holtzman

, Cirrito

, Colonna

(2015) TREM2 lipid sensing sustains the microglial response in an Alzheimer’s disease model. Cell 160, 1061–1071.

52.

Han

, Tang

, Yin

, Maalouf

, Beach

, Reiman

, Shi

(2014) Pituitary adenylate cyclase-activating polypeptide protects against beta-amyloid toxicity. Neurobiol Aging 35, 2064–2071.

53.

Hansen

, Hanson

, Sheng

(2018) Microglia in Alzheimer’s disease. J Cell Biol 217, 459–472.

54.

Pluvinage

, Haney

, Smith

BAH

, Sun

, Iram

, Bonanno

, Li

, Lee

, Morgens

, Yang

, Shuken

, Gate

, Scott

, Khatri

, Luo

, Bertozzi

, Bassik

, Wyss-Coray

(2019) CD22 blockade restores homeostatic microglial phagocytosis in ageing brains. Nature 568, 187–192.

55.

Keren-Shaul

, Spinrad

, Weiner

, Matcovitch-Natan

, Dvir-Szternfeld

, Ulland

, David

, Baruch

, Lara-Astaiso

, Toth

, Itzkovitz

, Colonna

, Schwartz

, Amit

(2017) A unique microglia type associated with restricting development of Alzheimer’s disease. Cell 169, 1276–1290e1217.

56.

Fan

, Brooks

, Okello

, Edison

(2017) An early and late peak in microglial activation in Alzheimer’s disease trajectory. Brain 140, 792–803.

57.

Bundy

, Vied

, Badger

, Nowakowski

(2019) Sex-biased hippocampal pathology in the 5XFAD mouse model of Alzheimer’s disease: A multi-omic analysis. J Comp Neurol 527, 462–475.

58.

Mathys

, Davila-Velderrain

, Peng

, Gao

, Mohammadi

, Young

, Menon

, He

, Abdurrob

, Jiang

, Martorell

, Ransohoff

, Hafler

, Bennett

, Kellis

, Tsai

(2019) Single-cell transcriptomic analysis of Alzheimer’s disease. Nature 571, 332–337. Author Correction: Nature 571, E1.