Introduction
Alzheimer’s disease (AD) is a progressive, age-related neurodegenerative disorder that is triggered by the appearance and build-up of amyloid-beta (Aβ) plaques in the cortex. Our lab and others have shown that microglia play an integral role in plaque formation and homeostasis, as well as downstream pathogenesis such as loss of synapses, perineuronal nets, and neurons [
9]; Hansen, Hanson, & Sheng [
20]; E. Spangenberg et al [
61,
62]. In addition to initiating the inflammatory response to disease pathology, phenotypically distinct microglia cluster around Aβ plaques and actively regulate plaque morphology (e.g., compaction) [
6]; Condello, Yuan, Schain, & Grutzendler [
8]. Moreover, recent genome-wide association studies associated single nucleotide polymorphisms of genes highly enriched or exclusively expressed in myeloid cells (including
Trem2,
Tyrobp,
Apoe,
Ms4a,
Abca7,
Abi3,
Spi1) with an altered risk of developing AD [
20,
24,
25,
32,
64]. These data indicate microglia as a mediator of AD and a potential therapeutic target.
Microglia surrounding plaques undergo significant physical and chemical changes, including the retraction of their processes and swelling of their cell bodies. These changes are mediated by extensive alterations in gene expression, which reprogram the microglia to mount inflammatory responses and remodel their metabolism and lipid handling [
20]. As a result, they transition to a disease-associated microglia (DAM) phenotype, characterized by specific functional and molecular features. These changes in gene expression have been well studied by single cell RNA sequencing and find heterogenous subsets of microglia in AD, demarcated by the expression of genes such as
Trem2 and
Tyrobp that conventionally differentiate between disease- vs. non-disease-associated microglia, which in AD roughly correspond to plaque- and non-plaque-associated microglia (PAM and NPAM) respectively [
30]. Currently, the role of PAMs in AD is unclear as brain wide microglial gene deletion and overexpression studies have shown contradicting results [
18]; Gratuze, Leyns, & Holtzman [
17,
26,
27,
33,
34,
54,
68]. Thus, there is a critical need to specifically target and modulate PAMs over NPAMs to determine this cell population’s contribution to AD.
Dendranib precision nanomedicine is based on hydroxyl dendrimer (HD) technology. HDs consist of a hydrophobic core, repeating branches that expand outward, and hydrophilic functional groups at the outer surface. Importantly, the high density of surface hydroxyls provides a neutral charge allowing HDs to easily cross the blood brain barrier (BBB) in regions of inflammation and be selectively internalized by activated microglia and macrophages [
44,
45,
47‐
49,
66].
Size and surface chemistry of dendrimers determine their toxicity and biodistribution [
4]. More than 100 dendritic structures have been reported. Some dendrimers have been used clinically for nucleic acid and drug delivery in cancer, including many types of brain tumors [
3,
29,
31,
42,
43,
71] and show potential application in gene therapy [
1,
19,
40]. Notably, polyamidoamine (PAMAM) dendrimers have been shown to cross the BBB during times when pathological insults such as stroke, tumors, or traumatic brain injury compromise the BBB [
56]. Traditional dendrimers alone do not bypass the BBB with high efficiency without resorting to invasive approaches such as carotid artery injections [
63,
74]. Recently, however, systemic administration of HDs have shown promise in bypassing slightly impaired BBB and were shown to be taken up specifically by microglia and macrophages in regions of neuroinflammation in rodent models of cerebral palsy, glioblastoma, Rett syndrome, AD, ALS, and multiple sclerosis (MS) [
44,
45,
47‐
49,
59,
60,
66]. In the AD brain, PAMs are the primary phagocytic macrophages; therefore, HDs theoretically have the capacity to bypass the BBB and become specifically engulfed by PAMs.
Here, we sought to determine whether HDs can successfully bypass the BBB and be specifically phagocytosed by PAMs in the context of AD. To that end, we intraperitoneally injected HDs conjugated to a Cy5 fluorophore into an aggressive mouse model of amyloidosis; 5xFAD mice at 7 months of age. We find that one injection is sufficient for the dendrimers to cross the BBB and leads to brain-wide, PAM-specific engulfment of these HDs into the microglial lysosomal compartment. To therapeutically modulates PAM function, as proof of principle, we used D-45113, a dendranib that inhibits CSF1R tyrosine kinase. Our lab has previously shown that all microglia express CSF1R, and that inhibition of CSF1R leads to the indiscriminate death of the microglia [
12], and that the elimination of microglia in 5xFAD mice can inhibit plaque development early in disease, and rescue synapse and neuronal number associated with late disease [
61,
62]. Additionally, our lab and others have shown that low-dose inhibition of CSF1R can inhibit microglia-plaque association, attenuate neuroinflammation and rescue synaptic integrity and cognition in AD mouse models [
10,
51]. Not to be overlooked, studies inhibiting CSF1R in tauopathy models show reduced levels of microglia which leads to reductions in tau levels, amelioration of inflammation, and synaptic, and neuronal loss [
2,
28,
41]. Altogether, HDs represent a novel and nuanced approach for targeting PAMs and further studies should be undertaken with other microglial modulators to uncover the specific role of PAMs in AD. Establishing the effectiveness of these dendrimers in targeting and treating PAMs will allow us to tailor appropriate therapies towards this subset of microglia and develop therapeutic treatments with greater precision.
Methods
Synthesis of D-45113
D4-alkyne dendrimer (Lot# DP-07-85-3) was dissolved in 20 mL of anhydrous dimethylacetamide (DMA). A CSF1R tyrosine kinase inhibitor with a terminal azide was added to a stirring solution of D4-alkyne. Copper bromide and Pentamethyldiethylenetriamine (PMDTA) were then added to the solution. The stirring solution was placed in a 95ºC oil bath overnight. The reaction mixture was then dialyzed against DMA followed by water (membrane cut-off at 1000 Da). The aqueous solution was then lyophilized to obtain D-45113.
Mice
All animal experiments performed in this study were approved by the UC Irvine Institutional Animal Care and Use Committee (IACUC) and were compliant with ethical regulations for animal research and testing. Mice were mixed sex C57BL/6 (000664) mice. Animals were housed with open access to food and water under 12 h/12 h light-dark cycles. All mice were aged to 5 or 12 months unless otherwise indicated. The 5xFAD mouse expresses five familial AD genes (APP Swedish, Florida, and London; PSEN1 M146L + L286V; [
50] and is characterized by aggressive amyloid pathology throughout the brain and synaptic and neuronal loss in the subiculum. For 5xFAD genotyping, the primer sequences used were PS1 Forward 5′ - AAT AGA GAA CGG CAG GAG CA – 3′ and PS1 Reverse 5′ - GCC ATG AGG GCA CTA ATC AT – 3′.
Animal treatments
All rodent experiments were performed in accordance with animal protocols approved by the Institutional Animal Care and Use Committee (IACUC) at the University of California, Irvine. 7-month-old wild-type (WT) or 5xFAD mice were intraperitoneally (IP) injected with 55 mg/kg G4 PAMAM hydroxyl dendrimers conjugated to a Cy5 fluorophore followed by euthanasia 48 h post injection. For time course D-Cy5 experiments, 7–9-month-old mice were treated as above, but euthanized either 48 h, 15 days, or 21 days post injection. For D-45113 experiments, 4 month and 11-month-old mice were IP injected with 200 mg/kg of D-45113 twice per week for four weeks. At the end of treatments, mice were euthanized via CO2 inhalation and transcardially perfused with 1X phosphate buffered saline (PBS). For all studies, brains were removed, and hemispheres separated along the midline. Brain halves were either flash frozen for subsequent biochemical analysis, or drop-fixed in 4% Paraformaldehyde (PFA; Thermo Fisher Scientific, Waltham, USA) for subsequent immunohistochemical analysis. Half brains collected into 4% PFA for 48 h and then transferred to a 30% sucrose solution with 0.02% sodium azide for another 48–72 h at 4 C. Fixed half brains were sliced at 40 μm using a Leica SM2000 R freezing microtome.
Histology and confocal microscopy
Fluorescent immunolabeling was performed using a standard indirect technique as described previously [
22]. Brain sections were stained with primary antibodies against: ionized calcium binding adaptor molecule 1 (IBA1; 1:1000; 019-19741, Wako and ab5076, Abcam), CD68 (1:200; BioRad) glial fibrillary protein (GFAP; 1:1000; Abcam), NeuN (1:1000; Millipore), OLIG2 (1:200; Abcam), Aβ1–16 (6E10; 1:1000; Biolegend), and anti-lysosomal associated membrane protein 1 (LAMP1; 1:200; Santa Cruz Biotechnologies). For Amylo-Glo staining (TR-300-AG; Biosensis), tissue sections were washed in 70% ethanol 1 × 5 min, followed by a 1 × 2 min wash in distilled water. Sections were then placed in a 1% Amylo-Glo solution for 1 × 10 min then washed with 0.9% saline for 1 × 5 min and distilled water for 1 × 15 s before continuing fluorescent immunolabelling. For Thioflavin-S (Thio-S) staining, tissue sections were placed for 1 × 10 min incubation in 0.5% Thio-S (1892; Sigma-Aldrich) diluted in 50% ethanol. Sections were then washed 2 × 5 min each in 50% ethanol and one 10-min wash in 1xPBS before continuing with fluorescent immunolabelling.
High resolution fluorescent images were obtained using a Leica TCS SPE-II confocal microscope and LAS-X software. For confocal imaging, one field of view (FOV) per brain region was captured per mouse unless otherwise indicated.
Aβ and NfL ELISA
To isolate protein for the ELISA, flash-frozen brain hemispheres were microdissected into cortical, hippocampal, and thalamic regions and grounded to a powder. Hippocampal tissue was then homogenized in Tissue Protein Extraction Reagent (TPER (Life Technologies, Grand Island, NY)) with protease and phosphatase inhibitors present. Samples were centrifuged at 100,000 g for 1 h at 4 °C to generate TPER-soluble fractions. To generate formic acid fractions, protein pellets from the TPER-soluble fraction were then homogenized in 70% formic acid and centrifuged at 100,000 g for 1 h at 4 °C, the formic acid fraction is then neutralized. Quantification of soluble and insoluble fractions of both Aβ and NfL was performed as previously described [
67].
RNA sequencing
Whole transcriptome RNA sequencing (RNA-Seq) libraries were produced from hippocampal tissue of WT/Veh, WT/D-45113, 5xFAD/Veh, and 5xFAD/D-45113 mice sacrificed at 12 months of age. RNA was isolated with an RNA Plus Universal Mini Kit (Qiagen, Valencia, USA) according to the manufacturer’s instructions. Library preparation, RNA-seq, and read mapping analysis were performed by Novogene Co. Gene expression was analyzed using Limma, edgeR, and org.Mm.eg.db packages (Robinson, McCarthy, & Smyth [
55]), with expression values normalized into FPKM (fragments per kilobase of transcript per million mapped reads). Differentially-expressed genes were selected by using false discovery rate (FDR) < 0.05. Heatmaps were created using Morpheus (Morpheus,
https://software.broadinstitute.org/morpheus) and volcano plots were created using VolcaNoseR [
16].
Data analysis and statistics
Both male and female mice were used in all statistical analyses. ThioS, IBA1, NeuN, and OLIG2 counts were measured via the spots function and 6E10, LAMP1, and GFAP volume were measured via the surfaces function on Imaris version 9.6. All analyses were performed on 20x images (550 μm X 550 μm). The number of dendrimer+ cells / FOV in the subiculum and somatosensory cortex were manually counted for 20x images (550 μm X 550 μm) for each mouse via ImageJ. Number of PAMs and NPAMs with dendrimer present in their lysosome were then counted and divided by the total number of PAMs and NPAMs, respectively to get the ratio of PAMs and NPAMs containing Cy5 dendrimer.
Statistical analysis was performed with Prism Graph Pad (v.8.0.1; La Jolla, USA). To compare two groups, the unpaired or paired Student’s t-test was used. Time-course data was analyzed using One-way ANOVA (48 h, 15 days, and 21 days), while D-45113 data with more than two groups used Two-way ANOVA (Treatment: Vehicle vs. D-45113 and Genotype: WT vs. 5xFAD) using GraphPad Prism Version 8. Tukey’s post hoc tests were employed to examine biologically relevant interactions from the two-way ANOVA regardless of statistical significance of the interaction. For all analyses, statistical significance was accepted at p < 0.05. and significance expressed as follows: *p < 0.05, **p < 0.01, ***p < 0.001. n is given as the number of mice within each group. Statistical trends are accepted at p < 0.10 (#). Data are presented as raw means and standard error of the mean (SEM).
Discussion
Dendrimers are dynamic nanomolecules which have been utilized for drug delivery in cancer and in the brain when the BBB has been severely compromised [
56]. Factors determining dendrimer biodistribution and toxicity are chemical composition, architecture, size, and surface properties. Traditional PAMAM dendrimers have trouble passing intact or slightly impaired BBB [
74]. Currently, there exist few ways to deliver therapeutics past the BBB and into the brain. One approach is through cerebrospinal fluid (CSF), intracerebral, and intracerebroventricular injection, however, while this is an efficient means to get therapeutics in the brain, these injections are very invasive procedures [
11,
53]. Other non-invasive delivery methods such as nasal drug administration, exosome delivery, and nanoparticle delivery represent promising avenues for drug delivery to the brain, but these methods have their own caveats as well such as toxicity problems, dosing limitations, and drug conjugation problems [
11,
21,
52,
57]. Previous research has shown that HDs can bypass a partially impaired blood-brain barrier and be phagocytosed by activated microglia and macrophages [
44,
45,
47‐
49,
66]. However, to our knowledge, this is the first study to assess the ability of HDs to target and treat PAMs specifically in the context of AD. With this in mind, we aimed to address two main objectives: first, to investigate whether HDs can selectively target PAMs in the brains of 5xFAD mice; and second, to evaluate the potential of these HDs for biological modulation of PAMs in the brain.
To clarify the precise role of microglia in AD pathogenesis, it is critical to target specific subsets of microglia, particularly PAMs. Among the key regulators of microglial-plaque association in AD, the most extensively studied is triggering receptor expressed on myeloid cells 2 (TREM2). Previous research that involved knocking out, knocking down, or overexpressing TREM2 or its downstream effectors has emphasized the crucial role of TREM2 in promoting microglial association with plaques [
17,
18,
26,
27,
33,
34,
54,
68]. Nevertheless, the impact of this association on the brain remains uncertain. This may be due to the fact that the vast majority of these studies have targeted all cells in the brain and periphery starting
in utero in mouse models. In humans, TREM2 is essential for maintaining normal brain homeostasis [
15,
23,
46,
65], and
TREM2 mutations resulting in loss of function are associated with a distinct neurodegenerative condition, Nasu-Hakola disease [
72]. This represents a confound in the current literature and highlights the importance of developing therapies which target subsets of diverse cells with temporal specificity. HDs, which target only the most phagocytic microglia in AD, appear to be very promising in this regard. HDs have the ability to therapeutically modulate PAM in regions of inflammation while sparing other cell types of potential off-target effects. Furthermore, AD progresses at different rates in different brain regions, resulting in varying microglial responses throughout the disease’s course. These dendrimers may be beneficial in that they may target microglia only when needed in disease.
Here, we show proof of principle that HDs are specifically internalized by PAMs and can have a biological effect when conjugated to a CSF1R inhibitor (D-45113) in a mouse model of AD. While we find that dendrimer is only colocalizes with microglia in the brain, it is possible that other cell types may take up levels of dendrimers that are undetectable via IHC. Also worth noting is the fact that PAMs are more resistant to CSF1R inhibition-mediated depletion compared to NPAMs in 5xFAD mice and a mouse A/T/N model [
39,
62], perhaps indicating that CSF1R may not be the ideal target to robustly modulate PAMs. Regardless, D-45113 administration in 5xFAD mice reduces microglia number, similar to previous studies which pharmacologically inhibit CSF1R [
10,
12,
61]. Interestingly, with reductions in microglia number in 5xFAD mice treated with dendrimer, there is also a reduction in diffuse plaque volume (6E10); however, no difference in dense-core plaque volume (ThioS) is observed. This could in-part be due to the timing of treatment, as dense-core plaques are present in 5xFAD mice as early as 2 months of age [
50]. Additionally, D-45113 treatment results in less microglia-plaque interaction and an increase in plasma NfL levels in 5xFAD mice. This falls in line with previous studies suggesting that microglia-plaque interaction is beneficial in limiting the amount of damage caused by Aβ plaques [
13,
18,
62,
69,
70,
73]. These effects are much more prominent in older mice, as evidenced by behavioral, plaque, microglial, and RNA differences observed in 12-month-old 5xFAD mice treated with D-45113. Previous data from our lab indicate that PAMs show much higher levels of DAM marker CD11c (ITGAX) at 12- versus 4-months of age [
67], perhaps suggesting that microglia in our older cohort of mice may be more prone to dendrimer uptake. Surprisingly, while treatment with D-45113 led to a rescue in EPM behavior and Aβ levels, and a lowering of inflammatory gene expression, there is no rescue in dystrophic neurite or NfL levels. A few reasons we may not see a rescue include: (1) We treated mice that are mid-late stage in disease pathogenesis and perhaps treating earlier and for longer than 28 days will lead to a rescue; (2) The partial reduction of PAMs is not sufficient to rescue the damage apparent in 5xFAD mice; and (3) D-45113 treatment may have independent effects on the brain of these mice. Our findings suggest that while D-45133 treatment may have therapeutic benefits, further investigation is needed to determine the mechanisms behind synaptic damage and rescue with treatment.
Here, we have shown that HDs have the capacity to target and treat PAMs with temporal precision through systemic administration. We were also able to show proof of principle that D-45113, a dendranib that inhibits the CSF1R has biological activity in the AD mouse brain, specifically in PAMs. Previous studies have shown successful conjugation of dendrimers with siRNAs, antisense oligonucleotides, and other commercially available drugs [
7,
11], making these tools essential for delivering therapeutics across the BBB and directly to PAMs. Microglia have been increasingly implicated in tau hyperphosphorylation [
5,
18,
34,
36,
37,
58], and future studies employing dendrimers in plaque + tau mouse models will be crucial to understand the interaction between microglia and the two primary histopathological hallmarks of AD. Overall, however, these results demonstrate that systemically administered HD’s can be conjugated to effector molecules and enact a biological effect on their target microglial population.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.