Introduction
Accumulation of Aβ aggregates in the brain parenchyma is hypothesized to trigger a complex neurodegenerative cascade that ultimately results in Alzheimer’s disease (AD). Based on this hypothesis there has been intense interest in therapeutic targeting of Aβ and Aβ aggregates [
1‐
3]. Numerous immunotherapeutic approaches to targeting Aβ have been evaluated in preclinical rodent models of Aβ accumulation and multiple antibodies and active immunotherapies have advanced to clinical trials [
4‐
10]. Indeed, preclinical studies have repeatedly established the disease-modifying potential of anti-Aβ immunotherapy. However, the results to date from many human anti-Aβ immunotherapy trials have been disappointing [
11‐
16].
Data from the aducanumab phase 1b study suggested that reductions of CNS amyloid, or at least the amyloid ligand PET signal, can be achieved in relatively short time-periods [
8,
17,
18]. Though initial phase 1b trial data suggested that this amyloid reduction might be associated with functional and cognitive benefits [
8], the phase 3 trial targeting treatment in mild AD was initially halted due to lack of clinical efficacy [
19]. However, more recent reanalysis of the trial with some additional data suggests clinical efficacy associated with the high dose treatment. These reanalyzed data, whose interpretation is controversial [
20,
21], support a new biologic drug application that is currently being reviewed by the FDA. Similar reports of reduced amyloid PET ligand binding following immunotherapy have been reported with the antibody BAN2401 [
19,
22], and phase 3 studies of this antibody in symptomatic AD are ongoing.
Even though amyloid ligand reduction binding has been observed, it is not yet clear how well this will correlate with alterations in amyloid levels in the postmortem brain. Further, it did not appear that everyone treated with these antibodies showed large reductions in the PET amyloid signal. Despite the lack of evidence for truly robust and universal efficacy in terms of slowing functional and cognitive decline, there is hope that Aβ immunotherapies, with evidence for target engagement in humans, especially if used in the preclinical stages of AD or in primary prevention, could still show meaningful clinical efficacy [
23].
However, there are still significant gaps in our understanding regarding the mechanism of anti-Aβ immunotherapies to reduce Aβ deposition [
24‐
29]. One of the original hypotheses regarding a peripheral sink induced by the high concentration of free antibody in the periphery has largely been ruled out [
27,
30]. Indeed, even in humans, robust peripheral target engagement of soluble Aβ with the central domain monomer selective antibody, solanezumab, has failed to reduce amyloid plaques and was not associated with significant functional or cognitive benefit in mild AD [
15,
31]. Both Fc-dependent and Fc-independent mechanisms have been proposed to underlie efficacy, and there is robust data generated using different anti-Aβ antibodies and different preclinical models to either support or refute either mechanism [
32‐
35]. Fc-dependent mechanism are purported to result in microglial activation and subsequent clearance of deposited Aβ. Fc-independent mechanisms likely work by binding aggregates and possibly interfering with subsequent aggregation, or enhancing efflux of the bound Aβ engaged by the antibody in the brain to the periphery. One possible factor that may help to explain the different preclinical observations regarding Fc-activation is that diffuse Aβ deposits seem to be reduced more with antibodies that activate Fc receptors than do the more densely cored plaques [
26,
35‐
37]. Notably, Fc-dependent microglial activation following plaque engagement is the mechanisms of action proposed for aducanumab and supported by preclinical data with the murine version of that human antibody [
8].
We and others have previously shown that altering innate immune activation states in the mouse brain via expression of cytokines, exposure to LPS or genetic manipulations can alter the time course of amyloid deposition in APP transgenic mice. Our internal data is consistent, showing that immune activating anti-inflammatory cytokines decrease amyloid loads and immune inhibitory, anti-inflammatory cytokines increase amyloid [
38‐
43]. Other data in the field creates a more confusing picture (reviewed in [
44,
45]). Many published studies show similar data with pro-inflammatory manipulations, whereas other data show that activation of an anti-inflammatory state can reduce Aβ accumulation or knockout of immune activating protein results in more amyloid deposition [
38,
46‐
50]. Fewer immune manipulations have been reported in tau mice [
51‐
53]. The handful of published studies suggest that there may be opposite effects of immune manipulations on Aβ and tau pathology [
54‐
56]. For example, manipulation of CX3CR1 and CX3Cl1 seem to have opposite effects on Aβ and tau pathologies [
57].
One concern with all of the studies we collectively do in our mouse models of AD pathology is that they are kept in relatively sterile conditions (e.g., specific pathogen free housing), meaning that they are subject to limited immune priming [
58,
59]. Herein, we explored how immune priming via constitutive expression of Il6 or Il10 influenced subsequent passive Aβ immunotherapy. We also evaluated how IL6 an IL10 and their receptors are altered at a transcriptomic level in (i) AD temporal cortex and cerebellar cortex in a large series of AD and Control brains and (ii) in longitudinal cohorts of APP TgCRND8 mice. We find that mAb5 passive immunotherapy alone and expression of rAAV-Il6 significantly attenuated Aβ accumulation, whereas expression of rAAV-Il10 significantly increased Aβ accumulation. rAAV-Il6 in combination with mAb5 resulted in a significant decrease in Thioflavin S positive plaque counts compared to either intervention alone, but the effect was only slightly additive. In contrast, rAAV-Il10 preconditioning completely abrogated the beneficial effect of mAb5 immunotherapy on amyloid deposition. Large-scale transcriptomic data reveal that Il10 and Il6 and their receptors show quite variable expression in the brain, in both humans and mouse models. These results have important implications for ongoing human AD immunotherapy trials, as they indicate that underlying immune activation within the brain may influence the ability of passive immunotherapy to alter Aβ deposition.
Discussion
We have found that preconditioning with Il6 or Il10 dramatically alters the effects of subsequent passive Aβ immunotherapy with the anti-Aβ antibody mAb5. Although a modest additive effect on one measure of Aβ deposition, (decreased Thioflavin S plaque count) was observed with Il6 preconditioning and subsequent mAb5 immunotherapy, Il10 preconditioning blocked the subsequent impact of mAb5 immunotherapy. The data has a high degree of validity as each of the manipulations by themselves replicated findings from our previous work [
30,
38,
41]. Il10 expression increased Aβ deposition and astrocytosis. Il6 expression decreased Aβ deposition and produced both an astrocytosis and microgliosis. mAb5 immunotherapy alone decreased Aβ deposition to some degree without appreciable effect on astrocytes or microglial cells. In our previous studies, we had not performed a simultaneous comparison of the effect of Il6 and mAb5 immunotherapy, though our impression from those studies was that Il6 reduced Aβ almost as well as a passive Aβ immunotherapy [
30,
41]. Here, we confirm that impression, and note that both treatments reduce Aβ deposition nearly to the same degree. The combination of Il6 and mAb5 shows no evidence for being synergistic and only appears to be partially additive, as the only significant impact of both Il6 and mAb5 together is on the number of Thioflavin S positive cored plaques.
There are a few caveats and limitations to this study. First, we have only evaluated the effects of preconditioning on the subsequent efficacy of a single anti-Aβ mAb. It is possible that other anti-Aβ mAbs, may be influenced to a greater or lesser degree by the immune preconditioning. Second, we have not looked at multiple time points. As we have previously shown that increased plaque load at time of immunization reduces efficacy of that therapy by itself, we would not expect that longer studies would appreciably alter the findings with Il10 preconditioning. It is, however, possible that longer treatment with both Il6 and mAb5 might show more robust effects of the combination. Third, we powered these studies to be able to detect effects on Aβ deposition, and the group sizes were not sufficient to evaluate behavioral impacts.
We have previously shown that suppression of an inducible mutant APP transgene in combination with passive anti-Aβ immunotherapy results in true clearance, not just suppression of ongoing deposition, of both the more diffuse Aβ deposits surrounding cored plaques as well as smaller non-dense core plaques. A similar study in the same inducible APP model, again shows that Il6 had a similar effect [
61]. These data along with our data on Il6 preconditioning with subsequent mAb5 immunotherapy suggest that at least partially additive effects on Aβ accumulation and possibly clearance can be achieved through combinations of blocking Aβ production, aggregation and clearance.
Il10 preconditioning may abrogate the effect of subsequent Aβ immunotherapy through several non-exclusive mechanisms. Given our previous work and that of others showing the reduced efficacy of immunotherapy in mice with higher amyloid loads, it may simply be that Il10 increases Aβ deposition to the point that subsequent mAb5 administration is no longer effective [
24,
58,
62‐
64]. We have also catalogued a large number of transcriptomic changes and functional alterations in microglial cells and microglial phagocytosis attributable to Il10 brain overexpression. Such complex changes induced by Il10 may also contribute to the lack of efficacy of subsequent passive immunotherapy.
The efficacy of immune checkpoint inhibitors in cancer is clearly influenced by the local tumor immune microenvironment [
65] Herein, we have utilized publicly available transcriptomic data that we have generated to explore the notion that changes and variability in the immune system in individual human or mice brains might be an extrinsic factor that could alter responses to an anti-Aβ targeted immunotherapy. Taking a simplistic approach to this issue, we focused on how the cytokines studied herein and their receptors are altered both temporally in a mouse model of amyloid Aβ deposition and by the AD state. First, these data show that there are significant increases in the cytokine receptors mRNA levels in the TgCRND8 mouse model and in the temporal cortex of humans with AD. However, these receptor RNAs are not as robustly upregulated as much as some other microglial genes (e.g. TREM2, CST7). Second, these data show that in the mouse brain mRNAs for Il10, which is almost undetectable, and Il6 are expressed at much lower levels on average than their receptors. In human temporal cortex the difference between level of cytokine and level of receptor is not as large, but still appreciable. The relatively low levels of these cytokines mRNAs compared to their receptor mRNAs, prompt a number of intriguing questions that will need to be pursued in future studies. These include: Are these differences reflected at the protein level? How can such typically low levels of cytokines robustly engage the receptor? Is the periphery, typically, a primary or at least significant source of Il6 and Il10 in the brain? Third, and perhaps most striking, is that there is a great deal of variance in individual mouse brains and even more so in the human temporal cortex with respect to the relationship between the expression of IL6 and IL10 and the high affinity receptors IL6R and Il10RA. We believe this later feature is highly relevant to our experimental studies. Such data shows that in the individual human control or AD brain, there is a high degree of variability in the activation state of the immune system. Given our current data that Il10 and Il6 differentially alter the subsequent effects of anti-Aβ immunotherapy, we hypothesize that variation in innate immune activation states within the human brain, may contribute to the variability in response to an Aβ targeting immunotherapy, at least with respect to the effects on clearance of Aβ.
Of course, by themselves the variable levels of these select cytokines and receptors do not necessarily inform on the functional status of the immune system within the brain. Additional systems level multiomic studies including single cell studies and more comprehensive algorithms to predict immune status may help illuminate a set of biomarkers that better define, at an individual level, the brain’s immune status [
66‐
68]. Ex vivo analysis of the immune microenvironment within a tumor has demonstrated utility in understanding how well immune checkpoint inhibitors may work. The lack of direct access to brain tissue in the AD field makes it much more challenging to assess immune status in an individual AD brain. Additional CSF and imaging biomarkers that better track innate immune status will be needed to better understand the influence of innate immune status on outcomes of anti-Aβ immune therapy [
69].
In summary, our experimental data show that altering the brain’s immune activation state by priming with cytokines that have different effects by themselves on Aβ deposition can markedly impact the efficacy of subsequent passive anti-Aβ immunotherapy. These results have important implications for ongoing human AD immunotherapy trials, as they indicate that underlying immune activation states within the brain, which at least in the postmortem brain appear to be highly variable, may influence the ability of passive immunotherapy to alter Aβ deposition.
Methods
Animal models and AAV injection
Mice. All animal husbandry procedures performed were approved by the Institutional Animal Care and Use Committee. TgCRND8 were maintained as described before [
70], transgenic males were crossed with B6C3F1 ntg females.
rAAV2/1 viruses for ICV injections expressing Il6 and Il10 under the control of the cytomegalovirus enhancer/chicken β actin promoter were generated as described previously [
41]. Briefly, AAV vectors expressing the cytokines under the control of the cytomegalovirus enhancer/chicken beta actin (CBA) promoter, a WPRE, and the bovine growth hormone polyA were generated by plasmid transfection with helper plasmids in HEK293T cells. 48 h after transfection cells were harvested and lysed in the presence of 0.5 % Sodium Deoxycholate and 50U/ml Benzonase (Sigma) by freeze thawing, and the virus isolated using a discontinuous Iodixanol gradient, and affinity purified on a HiTrap HQ column (Amersham). The genomic titer of each virus was determined by quantitative PCR.
Neonatal rAAV injections and antibody treatment. TgCRND8 mice were injected with 2 µl of rAAV ICV into the both hemispheres using a 10 µl Hamilton syringe with a 30 g needle on day P0 (Il10) or P1 (Il6) as described before [
38,
41] and aged till 2 months. They were then divided into two gender-matched cohorts, and immunized bi-weekly i.p. with mAb5 (IgG2b) or mouse IgG (0.5 mg/per mouse) diluted in 0.9 % saline, for 4 months, a regimen that was established in [
30] .
Measurement of Il6 and Il10 in the brain and plasma. Brains from mice injected with rAAV Il6 and Il10 were sagitally dissected and the left hemisphere was snap-frozen in isopentane. They were then homogenized at a concentration of 150 mg/ml and sequentially extracted with protease inhibitor cocktail (Roche) in RIPA buffer, 2 % SDS buffer, and 70 % formic acid (FA) as described previously.
Sandwich capture Il6 ELISA assays using RIPA soluble lysates were done with mouse specific reagents (BD Biosciences). The same procedure was performed on plasma from injected mice.
Aβ levels from the 2 % SDS– and 70 % FA–extracted samples were quantified using end-specific sandwich ELISA as previously described [
71]. Aβ40 was captured with mAb 13.1.1 (human Aβ35–40 specific; T.E. Golde) and detected by HRP-conjugated mAb 33.1.1 (human Aβ1–16; T.E. Golde). Aβ42 was captured with mAb 2.1.3 (human Aβ35–42 specific; T.E. Golde) and detected by HRP-conjugated mAb 33.1.1 (human Aβ1–16; T.E. Golde). ELISA results were analyzed using SoftMax Pro software.
Immunohistochemical imaging and image processing. Right hemibrain was fixed in 4 % paraformaldehyde. Immunohistochemical staining was done using pan Aβ antibody 33.1.1 (1:1500, T. Golde), Iba-1 (1:1000; Wako), GFAP (1:500; Chemicon). 1 % Thioflavin S (Sigma) staining was done on paraffin embedded brain sections using established protocols. Immunohistochemically and fluorescent stained sections were captured using the Aperio Scanscope XT or FL image scanner and analyzed using either Aperio positive pixel count or ImageJ program. Brightness and contrast alterations were applied identically on captured images using Adobe Photoshop CS3.
Quantification of Aβ deposition and gliosis. Immunohistochemically and fluorescent stained sections were captured using the Scanscope XT or FL image scanner (Aperio) and analyzed using ImageScope program. Aβ plaque burden and intensity of astrogliosis staining was calculated using the Positive Pixel Count program (Aperio). At least three sections per sample, 30 μm apart, were averaged by a blinded observer to calculate plaque burden. For Thioflavin S quantitation, one section per sample was used by a blinded observer to manually count the plaques using Adobe Photoshop CS5.
Statistical Analysis. Data were analyzed using Prism 6 (GraphPad) and presented as mean ± SEM. Overall data were tested for normality and, after being deemed to have a normal distribution, were analyzed via one-way ANOVA followed by Dunnett’s multiple comparison test. All comparisons were done between various groups and control. Sex differences in TgCRND8 mice were assessed by a post hoc analysis of the cohorts. Final images were created using Photoshop CS5 (Adobe).
RNaseq data. RNA sequencing data for TgCRND8 transgenic mice was downloaded from Synapse (doi:
https://doi.org/10.7303/syn3157182). The gene count matrix was normalized for sex, and sequencing batch using robust linear regression (lqs method, MASS package in R) after filtering genes with less than 1 CPM in at least 50 % of the samples, zero imputation and standardizing the covariates at their median value. Further details using this approach are described by Glusman et al. [
72]. The normalized count matrix was used as input for analysis using DESeq2 [
73]. Human RNA sequencing data was downloaded from Synapse (doi:
https://doi.org/10.7303/syn5550404). Data for the moue studies is reported as FPKM with and for the human studies as cqn.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.