MDMi in 3D show increased survival and more mature microglial features compared to 2D
We previously differentiated monocytes into MDMi in a 2D platform . To develop a more physiologically relevant MDMi model, here we generated MDMi in a 3D platform. Monocytes embedded in Matrigel (Fig. 1A; Fig. S1A) resulted in cultures with 6.2-fold higher cell thickness (i.e., size of the Z-stack that captured the whole cell) compared to 2D (Fig. 1B; Fig. S1B; Movie S1, 2). Remarkably, MDMi survival in 3D was significantly increased by 2.5 fold compared to 2D (Fig. 1C).
We next examined if the 3D culture affected morphology and expression of microglia-enriched markers in MDMi. 3D MDMi showed a highly ramified branched structure compared to 2D (Fig. 1D). Branched structure parameters —branch length and number of branches (endpoints) —, and branch complexity parameters —number of triple and quadruple junctions (points at which branches divide into three or four sub-branches, respectively) — were significantly increased in 3D compared to 2D MDMi (Fig. 1E-G). Ramified microglia have a highly branched morphology with a larger convex area than ameboid cells, thus correlating with a low ramification index . The ramification index of 3D MDMi was lower than 2D MDMi (Fig. 1H), confirming an enhanced ramified morphology of MDMi in 3D likely due to the larger surface area for growth and differentiation provided by the 3D Matrigel scaffold.
Seminal microglial markers, including IBA1, PROS1, GPR34, TMEM119, GAS6 and TREM2, were upregulated and the leukocyte marker CD45 was downregulated in 3D MDMi compared to monocytes (Fig. 1I). Interestingly, we observed a significantly increased expression of PROS1, GPR34 and TMEM119 in 3D MDMi compared to 2D (Fig. 1I). This suggests that the 3D platform promotes selective microglia-enriched markers in MDMi. As immature microglia have been reported to lack TMEM119 expression , the upregulation of TMEM119 in 3D MDMi compared to 2D further indicates an enhanced microglial maturity in the 3D platform. Finally, positive immunostaining of Trem2 and P2ry12 proteins in 3D MDMi confirmed the retention of microglial proteins in MDMi for up to 35 days (Fig. S1C).
MDMi co-cultured in 3D with human neural progenitor cells elicit an inflammatory response to aggregated amyloid-β (Aβ)
To mimic the neuro-glial cues of the brain microenvironment, we used the immortalised human neural progenitor cell (NPC) line ReNcell VM  to establish a co-culture platform with MDMi. Immunofluorescence and qRT-PCR marker characterisation showed that ReNcell VM lead to mixed populations of astrocytes, oligodendrocyte progenitor cells (OPCs) and neurons upon spontaneous differentiation while retaining a radial glial phenotype (Fig. S2A-C). Differentiation of 3D ReNcell VM cultures (Fig. S2D) demonstrated enhanced maturation of astrocytic and neuronal populations compared to 2D (Fig. SE, F), providing a basis for establishing a more mature co-culture platform of MDMi and neuroglia. For the 3D co-cultures, monocytes — pre-mixed with Matrigel — were added to 3D ReNcell VM and differentiated for 35 days, resulting in a mixed 3D co-culture containing MDMi and ReNcell VM-derived neuro-glial cells (Fig. 2G). In 3D co-culture, monocytes readily differentiated into MDMi with a microglia-like morphology (Fig. 2H). In 2D co-culture, contrarily, monocytes retained their round morphology (Fig. S3), indicating that the support provided by the 3D Matrigel scaffold was necessary for co-culturing MDMi and ReNcell VM.
Aβ aggregation is a major histopathological hallmark of AD brains. Hence, we incorporated FITC-Aβ peptides into 3D co-cultures. Remarkably, we observed that FITC-Aβ peptides readily formed observable aggregates within 24 h of incubation in 3D (Fig. 2I) as opposed to 2D, where FITC-Aβ remained unaggregated (Fig. S4). High Aβ plaque load has been shown to induce pro-inflammatory responses in microglia from transgenic AD mouse models [35, 36]. Hence, we investigated whether MDMi are functional in 3D co-culture and respond to FITC-Aβ aggregates. We observed a significantly upregulated secretion of classical pro-inflammatory cytokines IL-1β (2.8 fold) and IL-18 (2.3 fold), and similar trends for IL-6, in 3D co-cultures treated with FITC-Aβ compared to untreated conditions (Fig. 2J). Secretion of other pro-inflammatory cytokines was also stimulated by FITC-Aβ (Fig. S4). Such inflammatory responses to Aβ were largely mediated by MDMi in the 3D co-cultures, as observed by significant cytokine secretion changes following treatment in 3D MDMi mono-cultures and no changes in 3D ReNcell VM mono-cultures (Fig. 2J, S5).
Disease-specific differences in 2D and 3D mono-cultures of AD patient-derived MDMi
We successfully generated 2D and 3D MDMi mono-cultures from healthy control (HC) individuals and AD patients (Fig. 3A) selected based on matched sex, age and APOE genotype (Table 1). Survival of both HC and AD MDMi was significantly increased in 3D compared to 2D by 2.6 and 2.4 fold, respectively (Fig. 3B).
Morphological examination of AD MDMi (Fig. S6) revealed that regardless of a 2D or 3D platform, HC and AD MDMi exhibit similar branch structure (branch length and number) (Fig. 3C, D) and complexity (number of triple and quadruple junctions, and ramification index) (Fig. 3E-G). Comparison of branched structure between 2D and 3D MDMi showed similar trends in HC and AD, including similar branch length and increased number of branches in 3D compared to 2D (Fig. 3C, D). Interestingly, this contrasted with branch length in a young HC cohort (20–40 years of age) of MDMi (Fig. 1E), where MDMi exhibited longer branches in 3D compared to 2D.
Comparison of branch complexity between 2D and 3D MDMi showed an increased number of triple junctions in 3D compared to 2D MDMi in both HC and AD (Fig. 3E). Quadruple junctions increased in 3D compared to 2D MDMi in HC but remained unchanged in AD (Fig. 3F). Ramification index was significantly decreased in 3D compared to 2D MDMi in both HC and AD cohorts, confirming a more ramified morphology in 3D MDMi (Fig. 3G). Our results show that the 3D platform enhances branch number, ramified morphology and complexity parameters in MDMi from HC individuals compared to 2D. However, the 3D platform is limited in enhancing all complexity parameters in MDMi from AD patients compared to 2D.
Next, we examined if AD risk genes are differentially expressed in AD MDMi when cultured in 3D compared to 2D mono-cultures. A panel of AD risk genes, including CLU, TREM2, PLCG2 and PILRB, was examined by qRT-PCR. Fold change of expression in 3D compared to 2D revealed highly heterogeneous distributions within the HC and AD cohorts, with 3D MDMi showing trends of enhanced expression in AD patients compared to HC individuals for most risk genes (CLU, PLCG2 and PILRB) (Fig. 3H, I). Interestingly, PILRB expression was significantly upregulated in AD 3D MDMi compared to HC (Fig. 3J), while no significant differences were observed for the other genes (Fig. S7). These results demonstrate that the 3D platform enhances the expression of microglia-specific AD risk genes in AD MDMi compared to 2D and reflects the heterogeneity of disease phenotypes within AD patients.
Disease-specific differences in 3D co-cultures of AD patient-derived MDMi
We next characterised MDMi from HC individuals and AD patients in the 3D co-culture platform (Fig. 4A). As observed in 3D MDMi mono-cultures, survival of MDMi in 3D co-culture was extended by 2.5 fold compared to 2D MDMi mono-cultures and was similar to 3D MDMi mono-cultures in both HC and AD cohorts (Fig. 4B). The increased survival of MDMi in 3D co-culture indicates that ReNcell VM do not affect MDMi viability. Consistently, similar expression of the apoptosis marker BAX between HC and AD 3D co-cultures (Fig. 4C) suggests that the cell ratio of MDMi and ReNcell VM and the duration of the co-culture were favourable.
To investigate whether AD MDMi reflect disease-specific differences in 3D co-culture, we analysed 1) cell-to-cell interaction with neuro-glial cells, 2) secretion of growth factors and cytokines, and 3) migratory and inflammatory responses to Aβ aggregates. The marked synapse loss in AD is predominantly mediated by microglia through aberrant synapse engulfment . Hence, we first examined if the cell-to-cell interactions between AD MDMi and ReNcell VM show differences compared to HC in the 3D co-cultures. 3D rendering and subsequent surface reconstruction of 3D co-culture images revealed a significantly smaller area of contact and a reduced number of contact points between MDMi and ReNcell VM in AD compared to HC 3D co-cultures (Fig. 4D-F). This suggests a possible impairment in the cell-to-cell interactions between MDMi and ReNcell VM in AD 3D co-cultures.
Microglia can secrete factors that alter neuron and astrocyte homeostasis, thereby contributing to AD pathogenesis . Hence, we compared the secretory profiles of HC and AD MDMi in 3D co-cultures after 35 days of co-culture. AD MDMi in 3D co-culture secreted higher levels of platelet-derived growth factor AA (PDGF-AA) and erythropoietin (EPO), and lower levels of interferon-γ (IFN-γ), compared to HC in 3D co-culture (Fig. 4G-I). This indicates that AD MDMi exhibit an altered secretory activity in the 3D co-culture. When comparing the secretion to 3D MDMi mono-cultures, we observed a significant upregulation of PDGF-AA, EPO (Fig. 4G, H) and other neurotrophic factors such as Angiopoietin 2 and the granulocyte-macrophage colony stimulating factor (GM-CSF) (Fig. 4J, K) in 3D co-cultures from both HC and AD cohorts. Secretion of these factors by 3D ReNcell VM mono-cultures (dotted lines in Fig. 4G-K) changes compared to 3D MDMi mono-cultures, suggesting that the interaction between MDMi and ReNcell VM in the 3D co-cultures has a functional impact on either cell type.
Microglia in the vicinity of Aβ plaques have been shown to exhibit altered proliferation, migration, clustering around Aβ aggregates in a mouse model of AD . Hence, in order to study MDMi behaviours in the presence of Aβ depositions, we added FITC-Aβ into HC and AD 3D co-cultures from and live-imaged for 7 days (Fig. 4L; Movie S3). AD MDMi surveyed longer distances and at a higher velocity around Aβ aggregates compared to HC, with no significant changes in the proportion of MDMi that clustered around the Aβ aggregates (Fig. 4M-O). Interestingly, there were disease-specific differences in inflammatory cytokine secretion between HC and AD MDMi in 3D co-cultures treated with FITC-Aβ. When Aβ was present, IL-6 secretion was significantly decreased in AD compared to HC 3D co-cultures, while increasing trends were observed for IL-1β, IL-18 (Fig. 4P) and other pro-inflammatory cytokines such as TNF- α and IL-10 (Fig. S8). Such differences were not observed under untreated conditions. Together, these results indicate that AD MDMi respond differently to AD-related stressors compared to HC when modelled in the 3D co-culture platform.
Drug treatment induces differential cytokine gene expression in MDMi cultured in 2D and 3D
Cytokines are key secreted molecules used by microglia to execute inflammatory and neuromodulatory functions and may have important roles in AD pathogenesis . To investigate cytokine expression profiles in MDMi we analysed a panel of inflammatory cytokines (IL-6, TNF-α, IL-8, TGF-β, IL-10, IL-1β and IL-18) by qRT-PCR. When comparing cytokine expression levels between platforms, HC MDMi showed no significant differences (Fig. 5A) while in the AD cohort, 3D co-cultures showed a significant downregulation of IL-8 compared to 3D MDMi, and TGF-β and IL-18 compared to 2D MDMi (Fig. 5B). Disease-specific differences were only observed in the 3D MDMi platform, where TNF-α was downregulated in AD compared to HC (Fig. S9).
Based on the differences in baseline cytokine expression between platforms, we next investigated whether drug treatment alters such platform-dependent responses in the HC and AD cohorts. For this, we trialled two FDA-approved compounds — dasatinib and spiperone. These drugs have been shown to mitigate inflammation in in vitro models of microglia [41, 42] and murine models of AD  and have potential as re-purposed drugs for treating neuroinflammation. Significant differences in pro-inflammatory cytokine responses to dasatinib treatment were most notable when comparing 3D to 2D MDMi mono-cultures, with IL-6 being downregulated in HC (Fig. 5C) and IL-8, TGF-β and IL-1β being upregulated in AD 3D MDMi (Fig. 5D). Spiperone treatment only induced significant changes in TNF-α expression in the AD cohort (Fig. S10A, B). Interestingly, dasatinib modified cytokine expression levels compared to untreated conditions in the 2D platform but not in any of the 3D platforms (red triangles in Fig. 5C, D), suggesting potential implications on the capacity of cell model systems to predict drug outcomes in patients.
Interindividual variability in drug responses was displayed using heatmaps, which show cytokine expression levels in 2D and 3D MDMi mono-cultures from each individual of the HC and AD cohorts (Fig. 5E, F; Fig. S10C, D). Cytokine expression in HC and AD MDMi showed highly heterogeneous distributions in both 2D and 3D platforms. Cytokines that highly varied among individuals included IL-10 in 2D and IL-8 in 3D for both HC and AD (Fig. 5E, F). Interestingly, differences between dasatinib-treated 2D and 3D MDMi in specific HC or AD individuals were also evident, namely IL-8, IL-10 and IL-18 (Fig. 5E, F). Similar trends in the heterogeneity of drug responses were observed for spiperone-treated MDMi (Fig. S10C, D).