Replacement of brain-resident myeloid cells does not alter cerebral amyloid-β deposition in mouse models of Alzheimer’s disease

To investigate the effects of replacing brain-resident myeloid cells with peripheral monocytes during cerebral β-amyloidosis, we crossed APPPS1 mice, which develop first amyloid deposits at 6–8 wk of age (Radde et al., 2006), with the CD11b-HSVTK (TK) line (Heppner et al., 2005). TK mice express herpes simplex thymidine kinase in myeloid cells, which allows for myeloid cell ablation through a 2-wk intracerebroventricular (icv) ganciclovir (GCV) administration in wild-type as well as APPPS1 mice (Grathwohl et al., 2009; Varvel et al., 2012). Importantly, we have previously shown that upon discontinuation of GCV administration, peripheral monocytes repopulate the entire brain within 2 wk (either using eGFP bone marrow reconstitution or an irradiation-independent model, i.e., animals expressing red fluorescent protein under the inflammatory monocyte-specific chemokine receptor-2 promoter; Varvel et al., 2012).

3-mo-old depositing APPPS1/TK animals were subjected to 2 wk of icv GCV treatment. Histological examination of brain tissue from APPPS1/TK animals was then performed 2 wk later. As expected, APPPS1/TK⁻ mice displayed congophilic Aβ deposits throughout the neocortex, and Iba1-positive myeloid cells were tightly clustered around these plaques (Fig. 1 a). In contrast, Iba1-positive cells in the repopulated APPPS1/TK⁺ mice did not closely surround congophilic plaques, although some processes extending from the myeloid cells appeared to be in contact with the deposits. In addition, the morphology of the Iba1-positive cells in repopulated APPPS1/TK⁺ mice was remarkably homogeneous throughout the cortex, with cells displaying shorter, asymmetrically oriented processes and enlarged cell bodies compared with Iba1-positive cells in APPPS1/TK⁻ animals, regardless of their position in relation to individual Aβ plaques (Fig. 1 a). Quantitative stereological analysis further demonstrated that ∼2.0 × 10⁶ Iba1-positive cells engrafted during the 2 wk after termination of GCV treatment, approximately double the number of microglia in the APPPS1/TK⁻ controls (Fig. 1 c). Notably, at this early time point after repopulation, no change in congophilic amyloid or total Aβ load was detectable (Fig. 1 b).

To determine whether the continued presence of brain-engrafted monocytes would alter their responses to Aβ plaques and/or modify Aβ deposition, APPPS1/TK mice were analyzed 12 wk after GCV treatment. Again, in APPPS1/TK⁻ animals, Iba1-positive cells clustered tightly around Aβ deposits. Interestingly, some clustering of the engrafted myeloid cells was also observed in the APPPS1/TK⁺ animals at this later time point (Fig. 1 a), and brain sections showed ∼1.8 × 10⁶ Iba1-positive cells in the monocyte-repopulated APPPS1/TK⁺ mice and ∼1.5 × 10⁶ Iba1-positive cells in the aged-matched controls (Fig. 1 c), consistent with our previous work (Varvel et al., 2012). Despite similar numbers of myeloid cells and visible responses of invading monocytes to amyloid deposits, no change in congophilic plaques or total Aβ deposits was observed at this later time point (Fig. 1 b). Furthermore, amyloid-associated dystrophic neuritic structures persisted in the myeloid cell–repopulated APPPS1/TK⁺ mice (Fig. 1 d).

It is possible that amyloid deposition was not altered in the APPPS1 mice because of the early and rapid nature of cerebral β-amyloidosis in this model. Therefore, we crossed TK mice to a second APP transgenic mouse model, APP23, which develops cerebral β-amyloidosis at 6–8 mo of age and shows more slowly progressing Aβ deposition in both dense cored and diffuse plaques (Sturchler-Pierrat et al., 1997).

9-mo-old APP23/TK animals were treated with GCV for 2 wk, and the mice were sacrificed 6 mo later. Histological examination of APP23/TK⁻ control mice revealed the presence of dense core, congophilic amyloid deposits tightly surrounded by Iba1-positive myeloid cells. Strikingly, in repopulated APP23/TK⁺ mice Iba1-positive cells also showed noticeable clustering around congophilic deposits (Fig. 1 e), although their cell bodies remained further detached from plaques than observed in TK⁻ animals, reminiscent of results from the APPPS1 model (Fig. 1 a). Also, infiltrating monocytes remained morphologically distinct from brain-resident myeloid cells, with larger cell bodies and shorter processes (Fig. 1 e). Of note, stereological analysis revealed similar numbers of Iba1-positive cells in APP23/TK⁻ (∼1.3 × 10⁶) and APP23/TK⁺ (∼1.2 × 10⁶) animals (Fig. 1 g).

Despite comparable numbers of Iba1-positive cells and clustering of the invading peripheral myeloid cells around congophilic plaques in APP23/TK⁻ and APP23/TK⁺ animals, no changes in congophilic amyloid or total Aβ load was detectable (Fig. 1 f). Furthermore, dystrophic neuritic structures persisted in the myeloid cell–repopulated mice 6 mo after termination of GCV treatment in the APP23/TK⁺ mice (Fig. 1 j). This suggests that over the 6-mo incubation period infiltrating monocytes do not perform a function that is distinct from microglia during cerebral β-amyloidosis. Rather, with increasing time in the brain the infiltrating cells appear to become more similar to microglia, as indicated by comparable cell numbers being present in the brain (Fig. 1 g), as well as their association with amyloid plaques (Fig. 1 e).

To examine the possibility that myeloid cell replacement might be beneficial before the onset of Aβ deposition rather than during established cerebral β-amyloidosis, we also repopulated predepositing APP23/TK⁺ mice at 5 mo of age. We performed this experiment only in male mice because of a later onset of Aβ deposition compared with females (Sturchler-Pierrat et al., 1997), yielding a larger time window for repopulation in the absence of Aβ deposition. We confirmed that with the same GCV administration protocol, efficient microglial ablation and brain repopulation could also be achieved in APP23 mice (Fig. 1 i). Histological analysis was performed at 10 mo of age, i.e., at early stages of plaque deposition.

Strikingly, in APP23/TK⁺ animals repopulated before the onset of Aβ deposition, Iba1-positive cells clustered strongly around plaques, reminiscent of myeloid cells in APP23/TK⁻ animals. Also, in plaque-free areas Iba1-positive cells in APP23/TK⁺ animals appeared morphologically more similar to microglia, showing smaller cell bodies and a more ramified appearance than in the other repopulated groups (Fig. 1 i). Despite these similarities, no alteration of congophilic or total Aβ plaque load was observed (Fig. 1 j). Of note, the number of Iba1-positive cells was not significantly different in APP23/TK⁻ versus APP23/TK⁺ animals (Fig. 1 k).

It has recently been reported that reduced numbers of plaque-associated microglia are encountered in APP transgenic mouse models either haploinsufficient or deficient in Trem2 (Ulrich et al., 2014; Jay et al., 2015; Wang et al., 2015). Therefore, we analyzed whether the differences in plaque association observed in our repopulated versus nonrepopulated groups could be related to Trem2 expression. In TK⁻ animals, plaque-associated Iba1-positive cells were Trem2 immunoreactive (Fig. 2 a), in line with the findings by Jay et al. (2015). Surprisingly, early after repopulation, when infiltrating monocytes did not cluster around plaques, these cells showed no detectable Trem2 expression in APPPS1 mice (Fig. 2 b; APPPS1/TK⁺ group 2 wk after GCV treatment discontinuation is shown). However, in long-term repopulated animals where plaque association was evident, Iba1-positive cells clustering around amyloid deposits also showed Trem2 expression (Fig. 2 c; the APP23/TK⁺ group repopulated at the age of 5 mo, which displayed the most pronounced plaque association, is shown). Our current data do not allow for a conclusion regarding the origin of myeloid cells (brain resident or peripheral) in control TK⁻ mice, wherein the brain-resident myeloid cells have not been ablated; nevertheless, they indicate that Trem2 expression is induced in myeloid cells in response to the local plaque environment rather than as a result of macrophage origin.

In summary, our data suggest that in our experimental models monocytes do not perform a long-term function in the brain that is sufficiently distinct from the role of microglia to modify cerebral β-amyloidosis. These data are in line with our previous work (Varvel et al., 2012), where we reported nearly complete functional overlap between resident microglia and long-term engrafted peripheral monocytes in the measures examined, including parenchymal coverage and responses to purinergic agonists. Furthermore, monocytes invading the brain quickly down-regulated monocyte-specific markers such as, e.g., CCR2 (Varvel et al., 2012). Accordingly, it was recently shown that upon transplantation into different tissues (such as lung, spleen, and liver), bone marrow–derived macrophages show molecular reprogramming to match the resident macrophage population (Lavin et al., 2014). Thus, our work may indicate that also in the brain, the tissue environment dominates over myeloid cell origin, although more work will be needed to definitively prove this point.

A recent study used a CSF1-receptor inhibitor to deplete resident microglia and found that, in contrast to the HSVTK model, myeloid repopulation occurred from a brain-resident microglial progenitor (Elmore et al., 2014). This raises the possibility that in our long-term repopulation experiments, some peripherally derived myeloid cells may, over time, be replaced by endogenous microglial precursors. Additionally, it is possible that some of the repopulating myeloid cells may originate from the proliferation of small numbers of surviving microglia. However, the engrafted myeloid cells in our model showed clear morphological differences throughout the entire period of our experiments, whereas in the study by Elmore et al. (2014), the repopulating cells showed a ramified morphology indistinguishable from microglia. Therefore, the majority of brain-resident myeloid cells after GCV-induced microglial ablation are likely to be derived from peripheral monocytes. Strong support for this interpretation also comes from the study by Prokop et al. in this issue, where heterochronic parabiosis experiments clearly demonstrate that the repopulating cells are peripherally derived.

Our data presented here show that even during brain disease, i.e., cerebral β-amyloidosis, engrafted monocytes adopt features (such as cellular ramification, plaque association, and Trem2 expression) of brain-resident myeloid cells (be it microglia or long-term engrafted peripheral monocytes) and do not modify Aβ deposition. Therefore, myeloid replacement by itself is not a likely therapeutic option for AD or other chronic neurodegenerative diseases that are not a direct result of genetic variants in myeloid genes.

Female hemizygous TK mice (Heppner et al., 2005) were crossed to male hemizygous APPPS1 (Radde et al., 2006) or hemizygous APP23 mice (Sturchler-Pierrat et al., 1997). APP23 mice express a transgene consisting of human APP with the KM670/671NL mutation under a Thy-1 promoter element, and APPPS1 mice carry an additional mutated human presenilin 1 (PSEN1) transgene. APPPS1 mice were generated on a C57BL/6 (B6) background, whereas APP23 and TK mice were originally derived on a B6D2 background and were backcrossed to B6 mice for >10 generations. Mice were kept under pathogen-free conditions, and experiments were in compliance with protocols approved by the local animal use and care authorities (Regierungspräsidium Tübingen).

Approximately 40 h before surgery, osmotic pumps (model 2004, 0.25 µl/h; Alzet) were filled with a 50 mg/ml solution of valganciclovir (F. Hoffmann-La Roche AG) diluted in PBS and primed at 37°C. Mice were anesthetized using ketamine and xylazine (100 mg/kg ketamine, 10 mg/kg xylazine), placed on a warming pad, and secured on a modified stereotactic apparatus. The skin and the periosteum were removed and the pump reservoir was placed into a subcutaneous pocket, formed on the back of the animal. The coordinates for the cannula from bregma were +0.1 mm anteroposterior, 1.0 mm lateral, and 2.5 mm dorsoventral. The cannula (Brain Infusion kit III 1–3 mm; Alzet) was held in place by dental cement (Heraeus). After surgery, mice were treated with paracetamol (i.p. 5 mg/kg daily; Essex Pharma GmbH) for 3 d.

2 wk after osmotic pump surgery, the animals were anesthetized using ketamine and xylazine (100 mg/kg ketamine, 10 mg/kg xylazine). A small (1 cm) incision was made in the skin between the scapulae to gain access to the subcutaneous pump reservoir and tube connecting the pump reservoir to the head cannula. The connecting tube was cut, and the pump reservoir was removed. The cannula and the attached tube were left in place, and the free end of the tube was sealed with dental cement. The incision was closed with sutures and the animal was returned to its cage. After surgery, mice were treated with acetaminophen (i.p. 5 mg/kg daily; Essex Pharma GmbH) for 3 d.

After transcardial perfusion, brains were removed and the hemisphere contralateral to the implanted cannula was fresh frozen for biochemical analysis. The hemisphere that received cannula insertion was fixed in 4% PFA for 24 h, followed by cryoprotection in 30% sucrose in PBS. Brains were subsequently frozen in 2-methylbutane and coronally sectioned at 40 µm using a freezing-sliding microtome. Immunohistochemical stainings were performed using the VECTASTAIN Elite ABC kits (Vector Laboratories). The following primary antibodies were used: rabbit polyclonal antibody to ionized calcium binding adapter molecule 1 (Iba1; Wako Pure Chemical Industries), rabbit polyclonal antibody to human Aβ (CN3), and rabbit polyclonal antibody 5313 to the N terminus of APP (as a marker of dystrophic neurites; provided by C. Haass, Ludwig Maximilians University Munich, Munich, Germany). All antibodies were used at a dilution of 1:2,000. Congo red staining was performed according to standard laboratory procedures. Images were acquired on an Axioplan 2 microscope with AxioCam MRm, using a 20×/0.5 air objective and AxioVision 4.7 software (all from Carl Zeiss).

For immunofluorescence staining, sections were prepared as in the previous paragraph, blocked for 1 h with 5% goat serum, and incubated overnight, at 4°C, with antibodies against rabbit anti-Iba1 (1:2,000; Wako Pure Chemical Industries) and sheep anti-TREM2 (1:100; R&D Systems). Please note that this antibody was established by Jay et al. (2015) to be specific for murine Trem2 (as indicated by lack of staining in Trem2^−/− animals). Primary antibody incubation was followed by extensive washing and incubation with anti–sheep–Alexa Fluor 546 and anti–rabbit–Alexa Fluor 647 secondary antibodies (Jackson ImmunoResearch Laboratories, Inc.). Plaques were then counterstained with Methoxy-X04 (4% vol of 10 mg/ml Methoxy-X04 in dimethyl sulfoxide and 7.7% vol Cremophor EL in PBS) and sections washed extensively. Images were acquired with an LSM 510 META (Axiovert 200M) confocal microscope using a 63× oil objective (1.4 NA) and LSM software 4.2 (all from Carl Zeiss). Laser lines 405, 543, and 633 were used to sequentially excite the fluorophores, and z-stacks of 17-µm thickness were acquired (1-µm intervals). Maximum-intensity projections were created using IMARIS 7.7.2 software (Bitmap).

Numbers of Iba1-positive cells and Aβ load were assessed on random sets of every 12th systematically sampled 40-µm-thick sections through the neocortex typically yielding 8–10 sections/mouse. Analysis was performed with the aid of the Stereologer software (Stereo Investigator 6; MBF Bioscience) and a motorized x-y-z stage coupled to a video microscopy system (Microfire color microscope camera; Optronics). For Iba1 quantification, the optical fractionator technique was used with three-dimensional dissectors (area 495 × 495 µm², height 15 µm, guard height 2 µm, counting frame 50 × 50 µm²). Iba1-positive cells with complete soma within the dissector volume were counted (Grathwohl et al., 2009; Varvel et al., 2012). Congophilic amyloid and Aβ load was analyzed using the area fraction technique (Bondolfi et al., 2002).

Statistical analysis was performed using JMP 7.1 software. The results are expressed as mean values ± SEM. Student’s t test was used for comparison of two groups, and two-way ANOVA was used for comparison of multiple groups followed by Tukey’s HSD post hoc analysis.

We thank Dr. Adriano Aguzzi (University of Zurich, Zurich, Switzerland) for kindly providing the HSVTK mouse line and J. Odenthal for experimental help and advice.

This work was supported by the Alzheimer’s Association (N.H. Varvel), The Alexander von Humboldt Foundation (N.H. Varvel), and a Roman Herzog Postdoctoral Fellowship of the Hertie Foundation (J.J. Neher).

The authors declare no competing financial interests.