Reporter cell assay for human CD33 validated by specific antibodies and human iPSC-derived microglia

CD33/Sialic acid-binding Ig-like lectin 3 (SIGLEC3) is an innate immune receptor expressed on myeloid cells and mediates inhibitory signaling via tyrosine phosphatases. Variants of CD33 are associated with Alzheimer’s disease (AD) suggesting that modulation of CD33 signaling might be beneficial in AD. Hence, there is an urgent need for reliable cellular CD33 reporter systems. Therefore, we generated a CD33 reporter cell line expressing a fusion protein consisting of the extracellular domain of either human full-length CD33 (CD33M) or the AD-protective variant CD33ΔE2 (D2-CD33/CD33m) linked to TYRO protein tyrosine kinase binding protein (TYROBP/DAP12) to investigate possible ligands and antibodies for modulation of CD33 signaling. Application of the CD33-specific antibodies P67.6 and 1c7/1 to the CD33M-DAP12 reporter cells resulted in increased phosphorylation of the kinase SYK, which is downstream of DAP12. CD33M-DAP12 but not CD33ΔE2-DAP12 expressing reporter cells showed increased intracellular calcium levels upon treatment with CD33 antibody P67.6 and partially for 1c7/1. Furthermore, stimulation of human induced pluripotent stem cell-derived microglia with the CD33 antibodies P67.6 or 1c7/1 directly counteracted the triggering receptor expressed on myeloid cells 2 (TREM2)-induced phosphorylation of SYK and decreased the phagocytic uptake of bacterial particles. Thus, the developed reporter system confirmed CD33 pathway activation by CD33 antibody clones P67.6 and 1c7/1. In addition, data showed that phosphorylation of SYK by TREM2 activation and phagocytosis of bacterial particles can be directly antagonized by CD33 signaling.

CD33/Sialic acid-binding Ig-like lectin-3 (SIGLEC3) is an innate immune receptor expressed on the cell surface of myeloid cells and is composed of an IgV domain, a C2 domain and a single-pass transmembrane domain followed by an immunoreceptor tyrosine-based inhibitory motif (ITIM) and an ITIM-like domain 1,2 . Recently, a polymorphic allele of CD33 (variant rs3865444(A)) was found to be negatively correlated with the risk to develop Alzheimer's disease (AD), thus, being AD-protective for the carrier 3,4 . This CD33 variant is co-inherited with the CD33 variant rs12459419(T), which modulates the splicing efficiency of exon 2 in CD33 4 . Exon 2 partially encodes for the IgV domain of CD33, which mediates sialic acid binding. Therefore, CD33 lacking exon 2 (D2-CD33/CD33 ΔE2 ) is missing the functional sialic acid binding domain. Additionally, this CD33 ΔE2 variant was found to show a reduction of CD33 surface expression on microglia 3,5 . In line with these observations, functional analyses showed that CD33 expression levels positively correlate with the amount of Aβ and Aβ plaque load in the brains of AD patients, while patients expressing the CD33 ΔE2 variant exhibit decreased amyloid-β deposition in the brain 3 .
Consequently, there appears to be a direct crosstalk between CD33 and TREM2 in AD 12 , but evidence for direct signaling interference between TREM2 and CD33 in a human cellular system is still inconclusive. Studying CD33 www.nature.com/scientificreports/ signaling is hampered by the fact that it has very short and transient kinetics. After interaction of the receptor with an appropriate ligand, the intracellular ITIM domain is first phosphorylated by membrane-associated Src kinases and then recruits phosphatases (SHP1, SHP2 or SHIP1) that lead to dephosphorylation of the ITIM domain itself and theoretically of microcluster-associated ITAM-domains of TYROBP/DAP12, too 7,13,14 . Accordingly, cellular systems to study signaling of the human cell surface receptor CD33 that are also suitable for highthroughput drug screening are rare or unavailable.
Here, we established a human cell-based reporter system for CD33 by fusing the extracellular domain of full-length CD33 (CD33M) or CD33 ΔE2 to TYROBP/DAP12. Using phosphorylation of SYK (pSYK) and calcium imaging as readouts we confirmed that the two putative agonistic CD33 antibodies, clone P67.6 and clone 1c7/1, were able to activate CD33 signaling. Furthermore, CD33 antibody clones P67.6 and 1c7/1 were able to antagonize the TREM2-triggered increase in pSYK and decreased the phagocytic uptake of bacterial particles in human induced pluripotent stem cell-derived microglia.

Results
Generation of human CD33 full-length (CD33M) and variant CD33 ∆E2 reporter cell lines. A chimeric CD33-DAP12 strategy was applied to shift the transient CD33 signaling from inhibitory to activatory (Fig. 1a). The transmembrane and intracellular parts of CD33 were replaced by the human TYROBP/ DAP12 containing a point mutation (p.D50A) to eliminate possible interactions with activatory receptors such as TREM2. Reporter cell lines for both, the full-length CD33M and the variant CD33 ∆E2 lacking the sialic acid binding side were generated (Fig. 1a). Signaling of CD33 was detected as phosphorylation of SYK by the AlphaLISA system and imaging of calcium fluxes by a calcium-sensitive green fluorescent protein (GFP) variant, GCaMP6m (Fig. 1b).
Successful cloning of the two CD33 variants, CD33M and CD33 ∆E2 , fused to DAP12 into the pcDNA5/FRT vector was demonstrated by restriction digestion with EcoRI (Fig. 1c). As predicted, CD33M-DAP12 positive clones showed two bands at 5,227 and 883 bp, which was clearly distinguishable from CD33 ∆E2 -DAP12 positive clones (4,846 and 883 bp) and the empty vector as control (4,195 and 883 bp). To prevent loss of expression of the protein of interest over time, particularly by methylation of the viral CMV promoter 15,16 , the promoter of the pcDNA5/FRT plasmid was successfully exchanged by the human EEF1A1 promoter, in both CD33-DAP12 constructs and the empty vector. Thereby, successful introduction of hEEF1A1 was indicated by an additional band at 1339 bp (CD33M) or 958 bp (CD33 ΔE2 ) after restriction digestion with XhoI (Fig. 1d).
In addition, two more constructs were cloned, in which a calcium-sensitive green fluorescent protein (GFP) variant, GCaMP6m, was introduced into the plasmid for improved live cell calcium imaging (Fig. 1e). The GCaMP6m gene was separated from the CD33-DAP12 construct via an internal ribosomal entry site (IRES), which enables independent translation of the two proteins. Subsequently, all plasmids were validated via Sanger sequencing.
In summary, measurement of endogenous pSYK levels suggested the activation of the full-length CD33M-DAP12 signaling pathway by the CD33 antibody clones P67.6 and 1c7/1, supporting an agonistic mode of action. www.nature.com/scientificreports/ DAP12 [19][20][21][22] . Subsequently, PI3K/PLCγ2 should lead to IP 3 generation and thus increase the intracellular calcium from the endoplasmic reticulum (ER) and other organelles (Fig. 1b). Further, dATP was used as a positive control. Extracellular dATP leads to an increase in intracellular calcium levels via the purinergic P2 receptor family and PI3K/PLCγ2 23,24 . Image acquisition was performed with ƒ = 1 Hz and for 95 s with 5 s of baseline recording (Fig. 4a). For this assay, we adjusted the treatment concentration to 10 µg/ml, which is close to the EC 50 previously determined. The isotype antibody IgG1 did not show a remarkable response in both cell lines. However, addition of 100 µM dATP led to an increase in the relative fluorescence intensity in both cell lines as demonstrated by changes in the area under curve (AUC) and maximum ΔF/F(t) signal. In detail, the maximum ΔF/F(t) signal in CD33M-DAP12-GCaMP6m expressing cells was 139.76% ± 22.04% (p = 0.011) and in CD33 ∆E2 -DAP12-GCaMP6m expressing cells 169.67% ± 23.78% (p = 0.006; Fig. 4b-e). Similarly, the AUC calculated for the treatment with dATP showed 6.22 ± 0.49 (CD33M-DAP12-GCaMP6m, p = 0.002) versus 10.51 ± 1.69-fold change (CD33 ∆E2 -DAP12-GCaMP6m, p = 0.03) compared to the control (Fig. 4d,e). Further, an increase in intracellular calcium transients for the CD33M-DAP12-GCaMP6m expressing cells was observed after addition of the CD33-specific antibody P67.6, whereas the antibodies WM53 and P67.6 F(ab) did not lead to a notable response. In detail, the antibody P67.6 evoked an increase in relative fluorescence intensity with a 3.92 ± 0.64-fold change in AUC (p = 0.04) compared to the control IgG1 and a maximum ΔF/F(t) signal of 62.57% ± 4.91% (p < 0.001). Addition of 10 µg/ml 1c7/1 led to a similar but not significant increase in intracellular calcium levels with an area under the curve of 2.72 ± 0.49-fold change (p = 0.15) compared to IgG1 and a maximum ΔF/F(t) signal of 48.70% ± 7.66% (p = 0.07; Fig. 4b-e). The CD33 ∆E2 -DAP12-GCaMP6m expressing cell line did not show an increase in intracellular calcium levels for any of the tested antibodies ( Fig. 4c-e).

CD33-specific antibodies induced signaling and led to increased intracellular calcium levels in
Thus, the CD33-DAP12 cell lines provide a reliable tool for testing CD33-specific antibodies. Further, the CD33-specific antibody clone P67.6 was confirmed to act agonistic on full-length CD33 fused to DAP12. The CD33-specific antibody clone 1c7/1 tended to have a similar agonistic effect on full-length CD33 fused to DAP12, but without statistical significance.
Thus, the CD33 antibody clones P67.6 and 1c7/1 are able to activate CD33 as well as modulate TREM2 signaling at the level of SYK phosphorylation and decrease the phagocytic uptake of S. aureus BioParticles.

Discussion
Recently, several genome-wide association studies (GWAS) linked polymorphisms in the CD33 gene to Alzheimer's disease (AD) 25,26 . The CD33 SNP rs3865444(C) has been associated with increased risk to develop AD 27 , while the less common allele for the CD33 SNP rs3865444(A) was found to decrease the risk to develop AD 25,26 . Thus, therapeutic targeting of CD33 by antibodies might be beneficial in AD. However, development of CD33interfering drugs was hampered by the lack of appropriate models for studying human CD33 in cellular systems.
In the present study, we developed a human cell-based CD33 reporter system, which was used to confirm two putative agonistic CD33-specific antibodies (clones 1c7/1 and P67.6). Furthermore, the effect of these antibodies constructs. Both, the full CD33 ecto-domain (CD33M) and the ecto-domain lacking the sialic acid binding domain (CD33 ∆E2 ) were fused to TYROBP/DAP12. CD33 ∆E2 can be identified by binding of the antibody clone 1c7/1 (blue) but not WM53 or P67.6 (red), whereas all three antibody clones can bind CD33M. (b) The CD33-DAP12 and CD33-DAP12-GCaMP6m cells were stained for CD33 surface expression with the antibody clones 1c7/1, WM53 and P67.6. A representative flow cytometry histogram plot for the CD33M-DAP12-GCaMP6m cells is shown (left side). All three tested antibodies were able to stain full-length CD33 on the cell surface. Expression of variant 2 CD33 from CD33 ∆E2 -DAP12 and CD33 ∆E2 -DAP12-GCaMP6m cells was only detected by antibody clone 1c7/1. A representative flow cytometry histogram plot for the CD33 ∆E2 -DAP12-GCaMP6m cells is shown (right side). (c) Quantification of CD33 staining showed a high percentage of CD33 expressing cells in the CD33M-DAP12-GCaMP6m line for all three tested antibody clones but only the CD33 antibody clone 1c7/1 was able to detect CD33 in CD33 ∆E2 -DAP12-GCaMP6m expressing cells. The antibody clones WM53 and P67.6 did not show any staining of CD33 ∆E2 -DAP12 expressing cells. (d) Quantification of CD33 staining revealed a high percentage of cells in the CD33M-DAP12 line expressed CD33, and was detected by all three antibody clones. CD33 in CD33 ΔE2 -DAP12 expressing cells was only detected by antibody clone 1c7/1. Data are shown as mean + SEM of three to five independent experiments; *** p ≤ 0.001 compared to Secondary Control determined by Welch ANOVA followed by Games-Howell post hoc test. www.nature.com/scientificreports/ on CD33 signaling was validated in an orthogonal assay using human induced pluripotent stem cell-derived microglia. Here, we showed that the agonistic CD33 antibodies directly counteracted the TREM2-triggered phosphorylation of SYK and decreased the phagocytic uptake of pHrodo-labeled S. aureus BioParticles. Interestingly, the AD-protective CD33 variant with the SNP rs3865444(A) was found to be co-inherited with rs12459419(T), which mediates exon 2 splicing 4 . The exon 2 of CD33 encodes partially for the IgV sialic acid-binding domain. We therefore used CD33 lacking exon 2 (variant CD33 ∆E2 /D2-CD33) as control to evaluate our cellular reporter system. The SIGLEC receptor CD33 shows strong species differences between mouse and human, since SIGLECs evolved very quickly in humans generating several orthologs without direct corresponding homologs in the mouse. In humans, CD33 predominantly signals via intracellular ITIM but not via the ITIM-like domain 6,7,28 , whereas murine CD33 only bears an intracellular ITIM-like, but no ITIM domain. Further, murine CD33 has a positive charged residue in its transmembrane domain, which is known to be critical for interaction with ITAMcontaining proteins such as TYROBP/DAP12 2,22 . A previous study also highlighted the functional differences between human and murine CD33 29 . Deletion of human CD33, but not murine CD33, led to decreased phagocytosis in macrophages and microglia. Further, cell surface expression of murine CD33 was entirely dependent on murine DAP12 expression, which was not described for human CD33 29 . Thus, murine CD33 does not reflect the human situation and requires the development of human-based systems to study CD33 signaling. Of note, human CD33 signaling shows bidirectional kinetics with fast and transient tyrosine phosphorylation of the ITIM by Src family kinases, followed by Src homology-2-containing tyrosine phosphatase 1 (SHP1) and SHP2, which bind to phosphorylated CD33 and dephosphorylate the ITIMs in an autoregulatory manner as well as ITAMassociated signaling molecules 30 . Furthermore, recruitment of SHP1 to the intracellular CD33 domain triggers endocytosis of the CD33 receptor 30 . Interestingly, inhibitory NK cell receptor signaling could be redirected to activatory signaling by replacing the ITIM domain with ITAM-containing signaling chains 31 . This approach was also used in genetic engraftment of a tumor-specific chimeric antigen receptor (CAR) in NK cells and has been tested in vitro to compare the ITAM-containing CD3ζ signaling chain with the ITAM-containing DAP12 chain that even showed increased efficiency 31 .
In the present study, we used a similar approach to overcome the limitations of human CD33 signaling and created a chimeric human CD33-DAP12 reporter system for both CD33M and CD33 ∆E2 . Therefore, we replaced the transmembrane and the intracellular inhibitory domains of human CD33M and CD33 ∆E2 with the corresponding activatory TYROBP/DAP12 domain, in which we mutated the charged residue in the transmembrane domain to avoid association with ITAM-signaling receptors. These constructs were stable transfected into Flp-In-293 cells, an engineered human embryonic kidney cell line. Using this concept, we circumvented measurement of the highly variable and transient ITIM phosphorylation or SHP1/2 recruitment. However, it needs to be considered that this cellular model system is only able to identify ligands or allosteric activators of CD33 binding to the extracellular domain of CD33. Modulators of the intracellular ITIM domain, which also might have the potential to antagonize ITAM signaling cannot be detected. Therefore, this model system is specific for studying extracellular CD33 modulators.   www.nature.com/scientificreports/ Remarkably, both constructs CD33M-DAP12 and CD33 ∆E2 -DAP12 were detected on the plasma membrane using flow cytometry with the CD33 antibody clone 1c7/1, which binds an epitope in the C2 domain of CD33. Thus, CD33 ∆E2 was not translocated into peroxisomes in our CD33 ∆E2 -DAP12 cell lines, as previously described for blood neutrophils and monocytes 5 . Surface expression of CD33 ∆E2 -DAP12 in the reporter cell lines might be attributed to the direct interaction with DAP12, which is already known to stabilize murine CD33 surface expression 29 . Flow cytometric analysis using the CD33 antibody clones WM53 and P67.6 showed CD33 surface expression only in CD33M-DAP12 expressing cells. This is in line with the literature since both antibody clones bind an epitope within the IgV sialic acid-binding domain, which is missing in CD33 ∆E2 due to splicing of exon 2 4 .
On the molecular level, activation of CD33 in the CD33-DAP12 reporter cell lines results in an ITAMmediated cellular response. Phosphorylation of DAP12 by Src kinases leads to recruitment and phosphorylation of SYK, which further results in an increase in intracellular calcium levels in a PI3K-PLCγ2-dependent manner [19][20][21] . Here, we used both measurements, namely intracellular calcium fluxes as well as SYK phosphorylation, as readouts to identify agonistic CD33-specific antibodies. The transient increase in intracellular calcium levels was observed only in CD33M-DAP12, but not in CD33 ∆E2 -DAP12 expressing cells upon stimulation with CD33 antibodies 1c7/1 and P67.6. Interestingly, the F(ab) version of P67.6 was not able to stimulate the CD33M-DAP12 reporter cell line suggesting a need for crosslinking of the receptor to enable downstream signaling. The agonistic effect of the two CD33 antibodies 1c7/1 and P67.6 was also demonstrated by their ability to increase the phosphorylation of SYK in CD33M-DAP12 expressing cells. In contrast, WM53 had no agonistic effect underlining key functional differences between these CD33-specific antibodies.
However, the approach of a chimeric CD33-DAP12 receptor introduced into a non-immune cell has the limitation that it lacks the recruitment of phosphatases during CD33 signaling. To test whether the two identified agonistic CD33 antibody clones (1c7/1 and P67.6) were capable to activate endogenously expressed native CD33, we used human induced pluripotent stem cell-derived microglia as a model system. IPSdMiG expressed typical lineage-specific markers CD11b, CD45, CD64, CD68, CX3CR1, IBA1, P2RY12, PU.1 and TMEM119. Moreover, CD33 surface expression was only detected in CD33M expressing WT but not in CD33 −/− or CD33 ∆E2 iPSdMiG in line with recent findings 5 . In human macrophages and microglia, CD33 theoretically counteracts an immune response originating from ITAM signaling. Phosphorylation of SYK was increased in this microglia model after activation of the ITAM-associated receptor TREM2 by an agonistic antibody. Co-treatment with the agonistic CD33 antibodies 1c7/1 or P67.6 dampened the phosphorylation of SYK. This modulation of SYK phosphorylation was not observed in CD33 −/− and CD33 ∆E2 microglia. Likewise, CD33 antibody clones 1c7/1 and P67.6 were able to decrease the phagocytic uptake of S. aureus bacterial particles in WT but not CD33 −/− and CD33 ∆E2 microglia. Interestingly, antibody concentrations of approximately the IC 50 value determined in the CD33 reporter cell line were only able to attenuate the TREM2-triggered increase of SYK phosphorylation by 20-30%. Thus, it might be possible that the capacity of CD33 to modulate TREM2 signaling is limited due to its autoregulatory nature or that TREM2 is more abundant on the cell surface than CD33, so that higher antibody concentrations would be needed to show a similar activation as in the reporter cell line. In addition, the exact epitope of these antibodies is not known to date. Thus, a small molecule directly targeting the ligand binding domain of CD33 principally might be more effective in attenuating TREM2 signaling than these antibodies.
Taken together, our data show that the novel chimeric CD33-DAP12 reporter cell line we developed can be used to study CD33 activation. We showed that two of the tested CD33-specific antibodies (1c7/1 and P67.6) were capable to activate CD33 in the reporter cell lines as well as in human iPSC-derived microglia. Furthermore, we showed a direct modulation of TREM2 signaling by agonistic activation of CD33 in human iPSC-derived microglia. Thus, the system can be used to identify further agonistic CD33-specific antibodies as well as small molecule modulators binding to CD33.

Culture of Flp-In-293 cells.
Flp-In-293 cells (R75007, Thermo Fisher Scientific) and its derivatives were cultured according to an adapted version of the manufacturer's instructions. Briefly, frozen cells were thawed in pre-warmed 293 medium (DMEM + L-Glutamine and 4.5 g/l D-glucose, 10% fetal bovine serum (FBS), 2 mM L-glutamine, 1 mM sodium pyruvate, 0.1 mM NEAA (Gibco)). Consequently, Flp-In-293 cells were cultivated in 293 medium containing 100 µg/ml Zeocin. Constitutive expression cell lines were cultured in 293 medium + 150 µg/ml Hygromycin B. The cells were passaged when reaching 80-90% confluency by detaching chemically using 0.25% Trypsin/EDTA. Generation of CD33 reporter cell lines. The full mRNA sequence of human CD33/SIGLEC3 and human TYROBP/DAP12 were obtained from NCBI (Gene IDs 945 and 7305, respectively). The CD33-DAP12 fusion protein lacks the intracellular ITIM and the transmembrane domains, which were exchanged by the TYROBP/ DAP12 sequence representing an ITAM domain. Further, a point mutation was introduced into the DAP12 gene (p.D50A), which eliminates possible interactions with TREM2, by using site-directed mutagenesis (#A13282, Thermo Fisher Scientific) following the manufacturer's instructions. Two different CD33-DAP12 constructs were generated: CD33M-DAP12 and CD33 ∆E2 -DAP12, which lacks the sialic binding domain (D2-CD33/ CD33m; Fig. 1a). Moreover, the genetically encoded calcium indicator (GECI) GCaMP6m 32 was introduced into CD33M-and CD33 ∆E2 -DAP12 plasmids separated via an internal ribosomal entry site (IRES) motif. GCaMP6m was purchased as pGP-CMV-GCaMP6m from Addgene (Plasmid #40754). In addition, the CMV promoter of pcDNA5/FRT was exchanged with the human eukaryotic translation elongation factor 1 alpha 1 (EEF1A1) promoter to prevent epigenetic silencing 15,16 . Primers were designed using Geneious v8.1 (Biomatters Ltd) with a 20 bp homologous overhang and matching melting temperatures. Amplification of the target sequences was achieved with AccuPrime Pfx SuperMix (Thermo Fisher Scientific) according to manufacturer's instructions. www.nature.com/scientificreports/ Subsequently, the PCR fragments were cleaned up using agarose gel electrophoresis and extracted using the QIAquick Gel Extraction kit (QIAgen). All cloning steps were performed via In-Fusion cloning (Takara Bio Inc.) following manufacturer's instructions. Briefly, 50 ng of linearized vector was incubated with the inserts at molar ratio 1:2 in presence of the In-Fusion enzyme at 50 °C for 15 min. Then, 2.5 µl of the Infusion reaction was transformed into Stellar Competent Cells (Takara Bio Inc.) heat-shocked and plated onto agar plates containing appropriate antibiotics. Individual clones were picked, inoculated overnight in LB broth followed by plasmid extraction using QIAprep Spin Miniprep kit (Qiagen) according to manufacturer's instructions. Subsequently the plasmids were analyzed for correct inserts using restriction digestion and Sanger sequencing. Stable CD33M-DAP12, CD33 ∆E2 -DAP12, CD33M-DAP12-GCaMP6m and CD33 ∆E2 -DAP12-GCaMP6m reporter cell lines were generated by lipofectamine transfection (Invitrogen) of plasmids into Flp-In-293 cells according to manufacturer's instructions with a 1:9 molar ratio of pcDNA5/FRT to pOG44 plasmid. 48 h post transfection the cells were split to 25-30% confluency. The following day, the culture medium was exchanged to 293 medium containing 150 µg/ml Hygromycin B to select for stable transfected clones. Approximately 20 clones per construct were then picked and expanded. When the clones reached an appropriate number of cells, they were examined for transgene expression by flow cytometry. For positive tested clones, two subclonal dilution steps were performed to ensure monoclonality and isogeneticity. Retested clones for the generated cell lines were then used to perform experiments.  Table 1) were automatically added to the cells after 5 s of baseline imaging. Images were analyzed using FIJI for ImageJ and calcium signal was calculated using the ΔF/F(t) equation 33 . For statistical analysis the area under curve (AUC) and the maximum ΔF/F(t) signal were calculated for each antibody in each individual experiment.

Generation
Generation of iPSC-derived microglia. BIONi induced pluripotent stem cell (iPSC) lines (isogenic control BIONi010-C, CD33 −/− knockout BIONi010-C-9, CD33 ∆E2 variant BIONi010-C-5) were generated and kindly deposited by Janssen Pharmaceutica (commercially available at EBiSC, European Bank for induced pluripotent Stem Cells, https:// cells. ebisc. org) and were cultured on geltrex-coated six-well plates in TeSR-E8 (STEMCELL) medium with a complete medium change every 24 h. For the differentiation into iPSC-derived microglia-like cells (iPSdMiG) BIONi010-C, BIONi010-C-9 and BIONi010-C-5 iPSCs were detached when reaching 70-80% confluency using 1 mg/ml collagenase IV (Thermo Fisher Scientific) for 30 min at 37 °C. The colonies were collected carefully in DMEM/F-12 (Gibco) and pelleted by gravity. The supernatant was aspirated and the colonies were transferred onto non-coated petri culture dishes to allow formation of embryoid bodies (EBs). The differentiation protocol was carried out according to the proprietary protocol at the LIFE & BRAIN GmbH (EP20162230). The iPSdMiG were produced from 4 to 6 weeks of differentiation and harvested from the www.nature.com/scientificreports/ supernatant during the following 7-week peak production phase. Harvested iPSdMiG were plated onto poly-Llysine-coated culture dishes and experiments were performed 24 h after plating.
Semi-quantitative real-time polymerase chain reaction (qPCR) analysis. RNA was isolated from iPSdMiG using the standard chloroform-phenol method. The cells were incubated with QIAzol (Qiagen) and chloroform (Roth) and centrifuged. The RNA-containing upper phase was extracted and incubated with an equal volume of isopropanol (Roth) for at least 2 h at − 20 °C. Finally, the RNA was centrifuged and the pellet was washed three times with 70% ethanol (Roth) and resolved in RNase-free DEPC-treated water. RNA was transcribed with the superscript III reverse transcriptase kit (Invitrogen) following manufacturer's instructions. CD33M and CD33 ∆E2 gene transcript levels were then analyzed by qPCR using isoform-specific primers (CD33M forward: 5′-GCT GTG GGC AGG GGC-3′, CD33M reverse: 5′-CCT TCC CGG AAC CAG TAA CC-3′, CD33 ∆E2 forward: 5′-CCC TGC TGT GGG CAG ACT TG-3′, CD33 ∆E2 reverse: 5′-GCA CCG AGG AGT GAG TAG TCC-3′). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH, forward: 5′-CTG CAC CAC CAA CTG CTT AG-3′, reverse: 5′-TTC AGC TCA GGG ATG ACC TT-3′) was used as house-keeping gene. 200 ng of cDNA was used as input together with SYBR Green PCR Master Mix (Applied Biosystems) and respective gene-specific primers. Amplification and detection were performed on an ABI 5700 Sequence Detection System (PerkinElmer) as follows: 95 °C, 10 s; 40 cycles of 95 °C for 15 s, 60 °C for 30 s and 72 °C for 30 s. Relative gene transcription was quantified using the ΔΔCT method with GAPDH as internal control.
SYK activation in iPSdMiG. SYK activation in iPSdMiG was determined as the ratio of phosphorylated SYK over total SYK. Therefore, 2 × 10 4 iPSdMiG were seeded per well of a PLL-coated 96-well plate (Corning). After 24 h the cells were co-stimulated for 5 min with anti-TREM2 (5 µg/ml, R&D Systems, see Table 2) and respective CD33 antibodies or isotype controls (10 µg/ml, see Table 1). Thereby, a new P67.6 full-length antibody (BioLegend) was used as high endotoxin contamination was measured for the P67.6 Santa-Cruz antibody and possible effects of endotoxins on phosphorylation of SYK in iPSdMiG could not be excluded. pSYK levels were detected by the AlphaLISA SureFire Ultra p-SYK (Tyr525/526) Assay Kit (PerkinElmer) and normalized to the values of total SYK using the AlphaLISA SureFire Ultra Total SYK Assay Kit (PerkinElmer) according to the 2-plate assay protocol for adherent cells. All samples were measured as technical duplicates in 384-well Opti-Plates (PerkinElmer). The plate was measured using the standard AlphaLISA settings on a PerkinElmer EnVision 2104 system. For the analysis, the technical duplicates were averaged and the pSYK signal was normalized to the total SYK signal. Data was displayed as pSYK/tSYK relative to the control anti-TREM2/IgG1 (anti-CD33 Ctrl).

PHrodo S. aureus BioParticle phagocytosis in iPSdMiG.
IPSdMiG were seeded at a density of 2 × 10 4 cells per well of a 96-well µ-plate (ibidi) as described above. Subsequently, iPSdMiG were incubated with 0.25 mg/ ml pHrodo Green Staphylococcus aureus BioParticles (Invitrogen) for 60 min at 37 °C together with 10 µg/ml of the respective CD33 antibodies or the isotype controls (see Table 1). Afterwards, the cells were washed, counterstained with Hoechst 33342 (5 µg/ml, Invitrogen) for 10 min at 37 °C and analyzed by live cell imaging using an IN Cell Analyzer 2200 system (GE Healthcare). The image intensity was measured by Image J version 1.53 h. The background intensity was subtracted from the image intensity and then normalized to the isotype control.
Ethics approval and consents to participate and publish. Ethics approval and consent to participate for generating the iPSC lines and their use for academic research were obtained by EBiSC.
Statistical analysis. Calcium imaging and iPSdMiG results were presented as mean + SEM and analyzed using SPSS v22 (IBM) as indicated. Briefly, the data was checked for normal distribution by Shapiro-Wilk test and for equality of variances by Levene's test prior to analysis. Subsequently, unless otherwise stated Welch-ANOVA with Games-Howell post hoc was used as equality of variances could not be guaranteed. Statistical analysis of pSYK CD33M-DAP12 reporter cell line data was performed using the GraphPad Prism 6.0.0 software