The derivation and characterization of the iPSC lines used in this study is described elsewhere: SFC840-03-03 (32), SFC841-03-01 (33) SFC856-03-04 (34)). All lines were derived from dermal fibroblasts from disease-free donors recruited through StemBANCC (Morrison et al., 2015) and the Oxford Parkinson’s Disease Centre: participants were recruited to this study having given signed informed consent, which included derivation of hiPSC lines from skin biopsies (Ethics Committee: National Health Service, Health Research Authority, NRES Committee South Central, Berkshire, UK, who specifically approved this part of the study (REC 10/H0505/71)). The iPSC lines were all derived using non-integrating Sendai reprogramming vectors (Cytotune, Life Technologies), cultured in mTeSR™1 (StemCell Technologies, #85850) on hESC-qualified Matrigel-coated plates (BD, #356234), passaging as clumps using 0.5 mM EDTA in PBS (Beers et al., 2012). Large-scale, low-passage frozen SNP-QCed batches were used for experiments to ensure consistency.

All cells were incubated at 37°C, 5% CO₂. Standard culture volumes for different plate formats were as follows: 10cm dish; 10 mL, 6-well; 2 mL, 12-well; 1 mL, 24 well; 0.5 mL, 96-well; 0.1 mL. iPSC were cultured in an in house alternative to commercial Essential 8 medium called OXE8 medium (35) on Geltrex™ (Gibco, #A1413201)-coated tissue culture dishes and passaged either as clumps using 0.5 mM EDTA in PBS, or using TrypLE™ Express (Gibco, #12604013). For 24 hours after TrypLE™ passaging, media was supplemented with 10 µM Y-27632 (Abcam, ab120129). Differentiation of iPSC into iPSC-Macs was carried out as previously described (35, 36) culturing in OXM medium and with the final 7 day differentiation occurring with cells plated at stated densities. iPSC-derived microglia were cultured based on a previously publication (37) by differentiation of iPSC-Mac precursor cells (PreMacs) for 14 days on fibronectin (Sigma, #F4759-1MG) coated tissue culture plates in RPMI (ATCC formulation, ThermoFisher, #A10491-01) supplemented with 100 ng/ml rhIL34 (Peprotech, #200-34-100uG), 25 ng/ml rhM-CSF (Life technologies, #PHC9501), 50 ng/ml rhTGF-β1 (Biolegend, #781804) and 1% Pen/Strep, with feeding every 3 days.

Human subject monocytes from healthy donors were provided by NHS Blood and Transplant services, under contract 17/WM/0333. Written informed consents were obtained from the donors before sample collection. Peripheral blood mononuclear cells (PBMC) were prepared by Ficoll gradient centrifugation followed by red blood cell lysis using ACK Lysis buffer (Gibco, #A1049201). After lysis, cells were divided for monocyte isolation either via adherence selection, or negative selection. To isolate by adherence, PBMCs were resuspended in RPMI 1640 (no serum) and ~1x10⁷ monocytes (assumption of 10% PBMCs are monocytes) plated in a 10 cm tissue culture dish and incubated for 2 hours at 37°C. Non-adherent cells were then removed by washing once with pre-warmed PBS. Remaining adherent monocytes were cultured for 2 weeks, with 100% media change every 3-4 days, in MDM media; RPMI medium (Sigma, #R8758) made up to 10% FBS (Sigma, #F9665), 1% GlutaMAX (100x, Gibco, #12634010), 1% HEPES (Gibco, #15630080), 1% Pen/Strep (Gibco, #15140122), and supplemented with 10 ng/mL recombinant human M-CSF. After 2 weeks, MDM were lifted by 15-minute incubation with TrypLE™ at 37°C and re-plated at desired densities and cultured for 2 more days. Alternatively, monocytes were isolated from PBMCs by magnetic bead negative selection using the human Pan Monocyte Isolation kit (Miltenyi Biotec, #130-096-537) according to manufacturer’s instructions. Monocytes isolated this way were then plated at desired densities and differentiated in MDM medium for 7 days with a 50% media change on day 4.

THP-1 monocytic cells (ATCC, #TIB-202) were cultured in RPMI (Sigma, #R8758) made up to 10% FBS, 1% GlutaMAX, 50 µM 2-mercaptoethanol (Gibco, #31350-010), and 1% Pen/Strep, with passaging every 3 days. For terminal differentiation into macrophages, THP-1s were treated with 5 ng/mL phorbol-12-myristate-13-acetate in culture medium for 48 hours. RAW264.7 (ATCC, #TIB-71) were maintained in DMEM (Sigma, #D6429) made to 10% FBS and 1% Pen/Strep and were passaged at 80-90% confluency (approximately every 3 days). Before plating for density assays, RAW cells were passaged twice into DMEM, 1% FBS, and 1% Pen/Strep to slow down growth.

PreMacs were plated in a 96-well plate (Greiner, #655180) at the stated densities between 1x10⁴ and 1x10⁵ cells/well in 0.15 mL media and terminally differentiated for 7 days as described above, with the volume reduced to 0.1 mL on day 4. On day 7, 50% of media was removed and replaced with fresh medium containing 2x concentrated TLR agonist (LPS; InvivoGen, #tlrl-eklps. Pam3CSK4; InvivoGen, #tlrl-pms. TLR3; InvivoGen, #tlrl-pic) and cells were returned to 37C, 5% CO2 for 24 hours. Supernatant was then collected, spun at 400g for 5 minutes to remove dead/floating cells, and stored at -80°C until cytokine quantification. Assay was repeated as above with MDM, THP-1, and RAW264.7 cells with the following modifications. “Re-plate” MDM; stimulation 2 days after plating. THP-1; stimulation immediately after 2-day differentiation by PMA. RAW264.7; stimulation 2 days after plating.

iPSC-Macs or MDM were cultured in 12-well plate format at stated densities. On the final day of differentiation, 100x concentration LPS and IFNγ in PBS were added directly to the well (10 µL, final concentration 100 ng/mL LPS and 20 ng/mL IFNγ (Gibco, # PHC4031)) and cells incubated overnight (16 hours). Supernatant was collected and spun 400g for 5 minutes to remove dead/floating cells and stored at -80°C until cytokine quantification. Cells were then washed with PBS and either lysed directly for RNA preparation (see below) or lifted by 10-15 minute incubation with Accutase™ Cell Dissociation Reagent (Stemcell technologies, #A1110501) (iPSC-Macs) or TrypLE™ (MDM) for immuno-staining and flow cytometry using antibodies and protocol as described previously (35).

PreMacs were plated for terminal differentiation at stated densities as described for cell titration. On day 5 or day 6 of terminal differentiation (48 hours or 24 hours prior to stimulation), media was directly exchanged between densities, acting quickly to prevent prolonged exposure of iPSC-Macs to air which can encourage lifting. On day 7, cells were stimulated with 1 ng/mL LPS and supernatant collected for analysis as described for the cell titration assay.

Supernatants were thawed and diluted in water either 1:10 - 1:100 for quantification of TNFα (Invitrogen, #88-7346-88), 1:20 - 1:200 human IL-6 (Invitrogen, #88-7066-88), or 1:10 - 1:100 for mouse IL-6 (Invitrogen, #88-7064-88). Assay was run according to manufacturer’s instructions.

Supernatants were diluted 1:2 in advanced DMEM/F12 and type I interferon was assayed using a HEK-293 ISRE luciferase reporter cell line as previously described (38).

Measurement of a panel of 36 cytokines, chemokines, and acute phase proteins was done using the Proteome Profiler Human Cytokine Array Kit (R&D Systems, #ARY005B). 1 mL of undiluted supernatant was used. The assay was run according to manufacturer’s instructions as instructed for use of Proteome Profiler Arrays with LI-COR Detection, as can be accessed at https://www.rndsystems.com/resources/technical/use-proteome-profiler-arrays-li-cor-detection. Arrays were visualized on the LI-COR Odyssey 9260 and quantified using Image Studio Lite version 5.2. A threshold of average signal (pixel density) normalized to the positive control ≥0.001 was chosen to distinguish unexpressed cytokines and chemokines which were not displayed if values below this threshold were observed across all 3 densities in both treatment groups.

Viability was determined in unstimulated, 1 ng/mL LPS, 1 µg/mL Pam3CSK4, or 3 µM Staurosporine (MP Biomedicals, #0219140080) stimulated cells. Cells were cultured as per cell titration assay at desired densities and stimulated by 50% media change with 2x agonist containing fresh media. After 24 hours, cells were stained with ReadyProbes™ Cell Viability Imaging Kit (Invitrogen, #R37609) diluted in Live Cell Imaging Solution (Invitrogen, #A14291DJ) for 15 minutes at room temperature. This was then replaced with fresh Live Cell Imaging Solution and cells were immediately visualised on the EVOS^® Fl Auto to quantify stained nuclei.

Low magnification, brightfield microscopy images were taken using the EVOS^® XL Core Imaging system. For confocal microscopy, PreMacs were plated for differentiation on 22mm glass slides. After 7 days, cells were fixed with 4% paraformaldehyde (Alfa Aesar, #J61899.AK) for 10 minutes followed by a PBS wash. Cells were then permeabilized for 1 hour at room temperature in PBS, 5% bovine serum albumin (Sigma, #A7906-100G), and 0.3% Triton X-100. They were then stained with Alexa Fluor™ 488 Phalloidin (Invitrogen, #A12379) diluted 1:400, and DAPI 1:1000, for 30 minutes in permeabilization buffer, followed by 3 washes in permeabilization buffer. Glass cover slips were then mounted onto slides with Dako fluorescence mounting medium (Agilent, #S3023), and imaged on the Olympus SoRa spinning disc microscope at 40x magnification. For scanning EM (SEM), cells were cultured on 15 mm glass cover slides. Cells were fixed first for 1 hour at room temperature in 2.5% glutaraldehyde in 0.1 M PIPES buffer at pH 7.4 followed by a wash in phosphate buffer and second fixation for 1 hour at 4°C with 1% OsO₄ in 0.1 M PIPES Buffer. Samples were then washed 3 times with Milli-Q water for 5 minutes before dehydration. To dehydrate, samples were incubated for 5 minutes each in 50%, 70%, 90% and 95% ethanol (EtOH), and 3x10 minutes with 100% EtOH. Samples were then dried by incubating for 3 minutes in 1:1 EtOH: hexamethyldisilane (HMDS) followed by incubating twice for 2 minutes in pure HMDS. HMDS was then removed and fumes allowed to evaporate before mounting samples on carbon adhesive tape on an SEM stub and sputter coating with 15 nm gold. Samples were imaged using a Zeiss Sigma 300 SEM operated at 2 kV.

Macrophage adhesion across our 3 densities after transient exposure to Accutase™ was determined as previously reported (35).

Cells were lysed over ice in RLT buffer from RNeasy Mini Kit (QIAGEN, #74104) supplemented with 10 µL/mL β-mercaptoethanol. Lysates were homogenized using the QIAshredder kit (QIAGEN, #79656) and stored at -80°C until extraction. All RNA extractions were done using the RNeasy Mini Kit (QIAGEN, #74104) including the optional DNase treatment step (RNAse-Free DNase Set, QIAGEN, #79254) and eluted in 35-50 µL RNAse-free water. RNA concentration was quantified with the Nanodrop 2000c. cDNA was made using the High-Capacity RNA-to-cDNA™ Kit (Applied Biosystems, #10704217). For each experiment, equal quantity of RNA was added across samples, using up to 9µL RNA. qPCR reactions comprised of 1 volume sample cDNA to 3 volumes master mix (Power SYBR™Green PCR Master Mix (Applied Biosystems, #4367659), forward and reverse primer mix, and nuclease-free water at a ratio of 5 µL: 1 µL: 2 µL respectively). Primers used are listed in supplementary table 5 . Reactions were in run in triplicate in 384-well format, 6-8 µL/sample in the Applied Bioscience QuantStudio™ 5. Default cycling conditions were used ( supplementary table 6 ). Analysis of gene expression was performed by normalization of Ct values to the average Ct value of two housekeeper genes, VIPAR and UBE4A, and displayed as fold difference as calculated by 2^-ΔCt.

RNA-Seq data was generated by Novogene. RNA was extracted in triplicate from a total of 1.7x10⁶ iPSC-Macs (SFC840-03-03 cell line) across each density cultured in 6-well plate format, using the RNeasy Mini Kit, including optional DNase treatment step. Sample handling, library preparation, and sequencing were carried out as previously reported (35).

Pre-processing, mapping, and quantification was carried out as previously reported (35). All downstream analyses were done using the R 4.0 programming language (R Core Team, 2020). Transcript abundance estimates in TPMs were summarized to the gene level using the tximport 1.16.0 package to correct for sample specific transcript length biases (39). Lowly expressed genes with length-scaled abundance estimates less than 10 in 3 samples were excluded by the default filtering function in the edgeR package (40). Differential expression was tested using limma-voom (41). Time at harvesting macrophages from a factory was defined as a batch variable in the design. GO enrichment analysis was conducted using the EGSEA package by leveraging 12 prominent gene set testing algorithms to obtain a consensus rank for each GO term (42). GO terms with less than 10 genes expressed in our dataset were excluded for a more stringent enrichment analysis. The background gene set used for the analysis was the total number of the unique genes observed in the experiment. KEGG analysis was done on differentially expressed genes, with all expressed genes as background, using the kegga function from the limma package. A publicly available dataset of human MDM, iPSC-Mac, iPSC-microglia, and M1 or M2 polarized iPSC-Mac and MDM was pre-processed as above and used as a reference (43). The gene expression values from both studies were corrected for library size using the fpm function from the DESeq2 package. Batch effects were estimated using surrogate variable analysis and subsequently removed with the limma removeBatchEffect function (44). The samples from this study were then projected onto a principal component analysis (PCA) space generated by the samples from the reference study to visualize the level of transcriptomic similarity. 3D volcano and radial plots were produced with the volcano3D package (45).

Macrophages were cultured at desired densities in a 24-well plate. 2.5, 5, and 10 μg Zymosan A (S. cerevisiae) BioParticles™ Alexa Fluor-488 (ThermoFisher, #Z23373) were added to low, mid, or high density cells respectively by 50% media change and incubated for 1 hour or 4 hours 37°C, 5% CO2. Particles not taken up were quenched with 0.025% (v/v) Trypan Blue (Sigma, #T8154) in PBS before lifting the cells by 15-minute incubation with Accutase™. Cells were fixed with 4% paraformaldehyde and fluorescence measured by flow cytometry with the BD LSRFortessa™ X-20.

Anion-exchange chromatography-mass spectrometry (IC-MS) was used for analysis of polar and ionic metabolites in the media samples as described in detail previously (46). We used compound discover software (Thermofisher Scientific) for data processing and statistical analysis. Metabolites were identified using predicted composition and searches against four global database MzCloud, Metabolika, ChemSpider, and an in-house database of authentic standards that included retention times. Matches to the in-house database were: retention time: true, RT (retention time) tolerance 2, and mass tolerance 5 ppm. Peak areas for the extracted ion chromatogram (EIC) for each identified metabolite are presented as a relative measure of abundance. To estimate the levels of itaconate and taurine, a standard curve was prepared. Standard samples were prepared in the medium used for growing the cells and all samples were measured using IC-MS. Analysis of data and calculations of EIC peak areas were performed using MestReNova software.

The AccQ∙Tag ultra derivatization kit (Waters) was used according to the manufacturers protocol to chemically derivatize amino acids present in the media samples. Samples were analyzed using Thermo Utimate 3000 UHPLC system coupled directly to a Q-Exactive Hybrid Quadrupole-Orbitrap mass spectrometer as described in detail previously (47). Data collection was performed in positive ion mode. Progenesis QI (Non-Linear Dynamics, Elstree, UK) software was used to analyze data with thresholds settings of p < 0.05 and FC > 1.2. Area under the curve for extracted ion chromatograms was used as a measure of relative metabolite abundance between samples.

Macrophages were cultured as described above for 6 days. On day 6, 25% of the media was replaced with fresh media containing IL-10 neutralizing antibody (R&D, #MAB217-100), an isotype control (R&D, # MAB002), or recombinant human IL-10 (Peprotech, #200-10) so that the final stated concentrations were achieved, Macrophage were then cultured for a further 24 hours before then being stimulated with Pam3CSK4 and TNFα secretion into the supernatant analysed as described above.

Specific statistical analysis of data is described in the figure legends. In general, normality of data was tested by Shapiro-Wilk test. Normally distributed data group comparisons were made using either two-tailed t-test, one-way or two-way ANOVA, or mixed effects analysis when repeated measure values were missing, as appropriate. Where multiple variables were measured from the same sample, Greenhouse-Geisser correction was applied. When not normally distributed, a two-sided Wilcoxon test was applied. Results with P < 0.05 were considered significant, and significance was defined as * <0.05, ** <0.01, ***<0.001, ****<0.0001. Outside of transcriptomic and metabolomic analysis, data analyses were performed using GraphPad Prism v9 and Microsoft Excel.

Plating densities stated in publications can vary from as low as ~9x10³ cells/cm² to as high as 4x10⁵ cells/cm², if they are even reported ( Figures S1A, B , Supplementary Table 1 ), and changes in density have been reported to alter human macrophage responses to Toll-like receptor (TLR) stimulation (29, 31, 48). We first determined whether similar changes occur to cytokine secretion by iPSC-Macs. iPSC-Macs were plated at increasing densities and stimulated with lipopolysaccharide (LPS, TLR4 agonist), Pam3CSK4 (TLR1/2), or Polyinosinic:polycytidylic acid (Poly(I:C), TLR3) for 24 hours. The amount of tumor necrosis factor-alpha (TNFα) secreted per LPS stimulated cell was inversely related to the cell density ( Figure 1A bottom), resulting in a highly variable but fairly level overall concentration in the culture supernatant ( Figure 1A , top). Likely causes of this variability include small variations in true cell number at the time of stimulation, as well as intrinsic variables in the iPSC-Mac differentiation protocol such as culture age and iPSC donor. This density effect was clearer after Pam3CSK4 stimulation, where total TNFα and TNFα/cell clearly decreases as density increases >2x10⁴ cells/well ( Figure 1B ). This decline was not unique to TNFα, as a change in IL-6 secretion per cell was also observed ( Figures S1C, D ). Type I interferon secretion post TLR3 stimulation was also significantly reduced ( Figure 1C ). Inflammatory stimulation can induce cell death via inflammasome activation resulting in apoptosis or necroptosis (49). To test whether decreased cytokine production by high density cells is due to increased cell death, viability was compared between unstimulated, TLR stimulated, or staurosporine (an apoptosis inducer) treated iPSC-Mac. For this we selected three densities corresponding to 1.5, 3, and 6x10⁴ cells/well in a 96 well plate (0.44, 0.88, and 1.76x10⁵ cells/cm² respectively). We will hereon refer to these as “low”, “mid” and “high” density cells ( Figure S1A ). While total percentage of viable iPSC-Macs was reduced by staurosporine, TLR stimulation did not cause noticeable changes ( Figure 1D , left) and no change in relative total cell number between densities was observed as determined by the number of nuclei present ( Figure 1D , right). Total number of nuclei following Pam3CSK4 stimulation was significantly increased at mid density versus untreated control, but not at low densities, likely reflecting variability within the assay (e.g. cells lost during wash steps) rather than functional differences.

Typically, inflammatory macrophages are described as classically activated or “M1 macrophages”, which can be generated by in vitro polarization using a combination of LPS and IFNγ (50). We assessed whether TNFα secretion after M1 polarization is also density dependent. While low and mid density “M1” cells secreted similar concentrations of TNFα/cell post-stimulation, high density cells secreted significantly less ( Figure 1E ). We next sought to determine whether other cytokines or chemokines are influenced by cell density. We have previously reported that our iPSC-Macs produce several cytokines constitutively under resting conditions (35). Therefore, we also checked whether this resting state is influenced or explained by plating density. Indeed, unstimulated iPSC-Macs of all densities constitutively secreted CCL2, CCL3/4, CXCL1, ICAM-1, IL-1RA, IL-6, IL-8, MIF, and Serpin E1 ( Figure 1F , left). However, low density cells secrete significantly more CCL3/4 (low: -0.271 ± 0.071 arbitrary units (AU), Log₁₀ transformed data, high: -1.714 ± 0.242 AU) and CCL5 (low: -1.254 ± 0.098 AU, high: -3.364 ± 0.098 AU) than high density cells. High density cells only secreted more CCL2 (low: -1.165 ± 0.130 AU, high: -1.019 ± 0.141 AU) and CXCL12 (low: -2.935 ± 0.187 AU, high: -2.509 ± 0.249 AU). However, per cell, low density cells still produce more CCL2 than high (low: 1.622 ± 0.038 AU, values normalized against the group mean, high: 0.561 ± 0.053 AU), and only CXCL12 is not significantly differentially secreted between densities (low: 1.345 ± 0.191 AU, high: 0.883 ± 0.130 AU, p = 0.206) ( Figure 1G , right). After stimulation with LPS/IFNγ, a majority of cytokines and chemokines measured were upregulated across densities ( Figures 1F , left, S1E ). Most notable of these were CCL5, CXCL10, CXCL11, IL-1β, and TNFα ( Figure S1E ). IL-1RA and IL-8 meanwhile were down regulated. With CCL3/4, low density cells displayed no change in secretion, whereas secretion increased approximately 5-fold from high density cells ( Figure S1E ). Again, low density cells still secreted more cytokines and chemokines per cell than high ( Figure 1G , right) with the exception of CXCL12 (low: 1.017 ± 0.311 AU, high: 0.933 ± 0.097 AU), CXCL11 (low: 0.887 ± 0.378 AU, high: 0.843 ± 0.418 AU), and IL-10 (low: 1.032 ± 0.247 AU, high: 0.760 ± 0.204 AU). Overall, we find that iPSC-Macs plated at low densities have an increased inflammatory cytokine and chemokine profile under resting and M1-polarized conditions.

We next investigated whether there were other density dependent phenotypic differences between our iPSC-Macs. Differentiated macrophages showed obvious morphological differences by brightfield microscopy. Low density cells had an elongated bipolar morphology ( Figure 2A ). As density increases, these protrusions shrank, and more rounded cells were apparent ( Figures 2B, C ). This was more clearly observed with confocal and electron microscopy. Confocal microscopy revealed that differences in cell structure corresponded to differences in actin cytoskeleton. Electron microscopy showed that at high density, morphology was heterogeneous, with both flattened and rounded, non-polar cells visible ( Figure 2C ). Despite these structural differences, cells were equally adherent across densities as determined by sensitivity to enzymatic lifting ( Figure S2 ). This would suggest that the rounded cells seen at high density are not undifferentiated, non-adherent cells.

After LPS/IFNγ polarization, low density cells did not always show obvious changes in morphology with brightfield microscopy ( Figure 2D ), but confocal and electron microscopy revealed that they appear to become more flattened and multipolar. Mid and high-density cells displayed clearer changes, projecting multiple processes and forming clusters of cells ( Figures 2E, F ). These clustered cells were not fused, as has been reported with M2-like cells (51, 52). In addition, we observed by confocal and electron microscopy thick tube-like structures reminiscent of tunnelling nanotubes in the high density environment (53) ( Figure 2F ). In summary, low density cells appeared to have a larger surface area and were typically more bipolar. High density cells responded more noticeably to inflammatory stimulation, with considerable direct cell-cell interactions occurring.

A previous study of murine bone marrow derived macrophage differentiation reported that density influences surface expression of markers such as CD11b (54). To test if density influences the differentiation state of iPSC-Macs, we measured expression of several typical macrophage surface markers; CD11b, CD14, CD16, and CD45. We also compared two markers associated with M1 and M2 polarization, the co-stimulatory molecule CD86 and the mannose receptor CD206, respectively (50, 55). Flow cytometry analysis identified two observable populations by scattering (Pop 1 and Pop 2, Figure 3A ). While low and mid density cells were evenly distributed between these populations, high density cells were smaller, belonging predominantly to Pop 1 (FSC-A^low SSC-A^high, Figure 3B ) (56). Surface marker immuno-staining showed clear expression of all markers above isotype control levels ( Figure 3C left, S3A ). When normalized against the pooled mean of all three densities, expression of CD14 and CD206 were both significantly higher on high density cells, while CD11b and CD45 showed trends towards higher expression in low and mid density cells ( Figure 3C , right). However, when sub-gated by population, this difference in CD11b and CD45 was only significant in Pop 2, suggesting a relation to cell size ( Figures 3D, E ). CD14 and CD206 on the other hand were significantly higher regardless of population and are therefore independent of cell morphology. Once stimulated with LPS/IFNγ, scattering of iPSC-Mac of all densities became more uniform ( Figure 3F ). CD11b, CD16, and CD206 expression all declined while CD14 and CD45 expression only dropped in low and mid density cells, but not high density cells ( Figure 3G , left, S3B, C ). Overall, high density cells expressed significantly higher CD14, CD16, CD45, and CD206 ( Figure 3G , right). CD86 surface expression increased for all iPSC-Mac, indicating successful M1 polarization, but no clear trend between densities was observed ( Figure 3G , right, S3 ). These results indicate that iPSC-Mac surface antigen composition and cell morphology are independently influenced by plating density.

We next sought to determine whether the density-dependent phenotypic differences could be explained by changes in gene expression. Using RNA-seq of resting iPSC-Macs at our three densities, we first compared the transcriptional profile against our prior report of iPSC-Macs cultured in OXM versus XVIVO-15™ based medium, as well as iPSC-Mac differentiation cultures reported by Gutbier et al. (35, 43). Principal component analysis showed that, irrespective of density, our iPSC-Macs cluster closely with unstimulated iPSC-Mac and iPSC-microglia from both published studies ( Figure 4A ). Despite the results so far suggesting M1 and M2-like polarization of low and high density cells respectively, they did not cluster with M1 and M2 polarized macrophages ( Figure 4A , right). Analysis of gene expression found 4599/15148 genes were differentially expressed between densities. Of these, 318 were upregulated only in low density cells, and 1495 upregulated only in high ( Figures 4B, C ). Mid density cells displayed an intermediate transcriptional phenotype. 2090 genes were upregulated in both low and mid compared to high, and 623 were upregulated in both mid and high compared to low. These differentially expressed genes include cytokines such as IL1B, CCL3, CCL4, CCL5 (higher in low density), and CXCL12 (higher in high density), the expression of which was consistent with secretion analysis ( Figures 1F, G , 4B, C . Supplementary File ). Gene Ontology (GO) term analysis was also consistent with our findings so far, with “chemokine activity”, “Ccr chemokine receptor binding” and “Cytokine activity” the most highly enriched terms in low density versus high density cells ( Figure 4D , S4A ). “Structural constituent of the cytoskeleton” was also enriched, consistent with figure 2 . High density cells were enriched for receptor binding and signaling-related terms including “Peptide antigen binding” and “Transmembrane Signaling Receptor Activity” as well specific receptor binding such as insulin and mannose, consistent with increased CD206 expression ( Figure 3 ). Of note, the “Major histocompatibility (MHC) class II Protein Complex” was significantly enriched ( Figure S4B ), with all constituent genes of this term robustly upregulated in high density cells ( Figure S4C ). Curiously many TLRs are more highly expressed in high density cells, ruling out density regulation of the inflammatory response being caused by increased surface sensitivity to stimulation ( Figure S4D ). Closer analysis of known TLR intracellular signaling genes indicates a more nuanced control, though. Intermediates MyD88, IRAK1/2 and TRAF6 are more highly expressed in low density macrophages. Conversely, known TLR/NF-κB inhibitory genes such as SOCS3, PTPN6, TNFAIP3 and NFKBID are also more highly expressed at low density, as well as anti-inflammatory transcriptional regulators like ATF3 and NR4A2 (57) ( Figure S4E ). Why, then, with so many negative regulators of TLR signaling upregulated in low density macrophages do we see a stronger pro-inflammatory response from these cells is unclear based on transcriptomic data alone.

We confirmed our RNA-seq results by qPCR for a selection of inflammation regulating secretory and surface proteins among the top 20 differentially expressed genes across densities and identified in Figure 1 , using macrophages from an independent genetic background ( Figure 4E , Supplementary Table 2 ). Many of these differences remained even after LPS/IFNγ stimulation ( Figure 4F ).

Considering the higher expression of scavenger receptors and enrichment for endocytic GO terms ( Figure S4B ), we considered whether high density cells may have increased capacity for phagocytosis. Uptake of fluorophore conjugated zymosan appeared to be higher in high density cells versus low within 1 hour of exposure, but after 4 hours was not significantly different across densities ( Figure S4F ). Therefore, phagocytic rate, but not total capacity, appears to differ between densities. However, caution must be taken as we cannot rule out the possibility that this is due to physical limitations of the assay (e.g. distance between particle and cell) rather than phenotype.

Macrophage inflammatory state is known to influence and to be influenced by their metabolic state, and secretion of metabolic intermediates can regulate the inflammatory response (58, 59). We investigated whether differences in macrophage density were linked to changes in iPSC-Mac metabolism. Kyoto Encyclopedia of Genes and Genomes (KEGG) term analysis of our RNA-Seq data showed enrichment of metabolic pathways, with over half of genes expressed in macrophages classified under this term differentially expressed between densities (627/1129 genes) ( Figure 5A ). Of the metabolism related KEGG terms, the most significantly enriched terms were “biosynthesis of amino acids”, “biosynthesis of cofactors”, and “Glycolysis/Gluconeogenesis”. In all 3 of these terms, more genes were upregulated in low density iPSC-Mac than high. To see how these may change after LPS/IFNγ stimulation, we selected common metabolic hallmark genes related to glycolysis (HK1 and GAPDH), fatty acid synthesis (FASN), glutamine synthesis (GLUL), and autophagy (LC3A) to quantify by qPCR ( Figures 5B, C ). Under resting conditions, low density cells have higher expression of all genes except GLUL, consistent with reported metabolism of inflammatory macrophages (60, 61) ( Figure 5B ). Upon stimulation with LPS/IFNγ, HK1 and GAPDH expression were no longer significantly different across densities, while LC3A expression became significantly higher in high density cells ( Figure 5C ). Our differentiation medium does not contain all amino acids used in protein synthesis so iPSC-Mac must synthesize their own. Also, autophagy is used during nutrient deprivation as a method of recycling intracellular components (62). Considering these facts, we asked whether the increased expression of metabolic genes, particularly those involved in amino acid biosynthesis, were due to nutrient deprivation. We used liquid chromatography–mass spectrometry (LC-MS) to determine abundance of extracellular metabolites at day 7 of iPSC-Mac terminal differentiation. The majority of extracellular amino acids showed no clear differences in abundance between different plating densities ( Supplementary Table 3 ). Glutamine, alanine, and glycine abundance had increased relative to the base medium, while serine and glutamic acid were depleted across all densities ( Figure 5D ). With the possible exception of serine, it appears that amino acid availability will not explain differences in iPSC-Mac response. Furthermore, culturing macrophages in an excess volume of media did not reverse the inflammatory phenotypes observed, either at resting state, determined by qPCR of differentially expressed genes FCGR2B, CXCL12, and IL-1β, or after Pam3CSK4 stimulation, as shown by TNFα secretion ( Figures S5A, B ). It is, therefore, likely changes in metabolic gene expression play a more specific role in regulating these macrophages than simply as part of a starvation response. We also measured altered metabolite abundances and metabolic intermediates ( Supplementary Table 4 ). Among these were several known markers and modulators of inflammation. Neopterin (63), was more abundant at low densities, and the anti-inflammatory molecules glutathione, itaconate, and the amino acid taurine (64–67), were more abundant in high density conditioned medium. Quantification of itaconate and taurine indeed demonstrated nanogram concentrations of these metabolites, 67.7 – 231.5 ng/mL and 6.3 – 35.5 ng/mL respectively ( Figure 5E ).

This would suggest that iPSC-Macs cultured at different densities may occupy different metabolite environments at the time of stimulation, which in turn may feedback on their response. To test this, we exchanged media between iPSC-Macs cultured at our different densities either 24 or 48 hours prior to stimulation with LPS ( Figures 5F, G ). When exchanged 24 hours prior to stimulation, density dependent suppression of TNFα secretion was no longer observed in either cells cultured in low or high density conditioned medium ( Figure 5F ). In low density conditioned medium, high density cells secreted more TNFα per cell, whereas low and mid density cells in high density conditioned medium produced less. Interestingly, this same effect was only seen in high density conditioned medium when media was exchanged 48 hours prior to stimulation ( Figure 5G ). This would suggest that high density cells are either actively depleting a pro-inflammatory molecule, or producing a stable anti-inflammatory molecule. While metabolites are one possible mediator of these effects, the most common and potent anti-inflammatory agent produced by macrophages is the cytokine IL-10 (68). As we have already observed that secretion of IL-10 post-stimulation appears to be unaffected by plating density ( Figure 1G ), we tested whether neutralization of extracellular IL-10 would nullify the density effect. Addition of recombinant IL-10 clearly suppressed the immune response, and this was reversed by addition of an anti-IL-10 neutralizing antibody. However, antibody treatment alone did not specifically increase TNFα secretion by high density cells compared to low density ( Figure S5C ). We therefore conclude that IL-10 is not responsible for the suppressed inflammatory response of high density macrophages. Alternatively, recent publications have identified nitrous oxide (NO) as an intermediate released into the supernatant that suppresses the macrophage inflammatory response in a density-dependent manner much like bacterial quorum sensing (26). However, inhibition of NO synthesis by L-NIL, an inhibitor of iNOS, had no effect on our iPSC-Macs ( Figure S5D ). Overall, we show here a link between density and macrophage metabolic activity. The culture supernatant, which contains varying, density dependent, concentrations of extracellular metabolites, actively enhanced or suppressed their ability to secrete inflammatory cytokines. However, as we did not remove protein components from the supernatant, we cannot rule out a potential dominant role of extracellular proteins other than IL-10 at this time.

We have so far demonstrated a significant effect of density on iPSC-Macs but no other macrophage models. We next tested whether primary MDMs are also influenced. Isolation methods can influence MDMs phenotype and response (69), so we isolated monocytes via two different approaches ( Figure 6A ). The first was by magnetic bead negative selection of monocytes followed by culturing for 7 days before stimulation, similar to our iPSC-Mac differentiation method (referred to as “Sort and plate”). Alternatively, MDM were selected by adherence to plastic. As this method gives non-exact plating densities, these cells were allowed to conclusively differentiate for 2 weeks. Subsequently, cells were lifted and re-plated at desired densities for another 48 hours before stimulation (here referred to as “Re-plate”). Morphological differences like those seen with iPSC-Macs were not always observed in MDM due to donor variability ( Figure S6A ). However, surface marker expression did display density dependent changes ( Figures 6B, C , S6B, C ). In “Sort & plate” MDM, lower expression of CD14 was observed in low density cells while “Re-plate” MDM showed significant differences in CD11b (low > high), CD86 and CD206 (high > low), but not CD14. After stimulation with LPS/IFNγ, “Sort & plate” MDM show very similar trends as iPSC-Macs, with reduced CD14, CD16, CD45, and CD206 in low density cells relative to high, and increased CD86 ( Figures 6D , S6D, E ). “Re-plate” MDM did not mirror iPSC-Macs, but none-the-less had the same pattern of CD16, CD86, and CD206 expression ( Figures 6E , S6F, G ). Transcriptional analysis by qPCR for the secretory, surface, and metabolic genes measured in iPSC-Macs ( Figures 4E , 5B ), again showed differences between densities ( Figures 6F, G ). Unstimulated “Sort & plate” and “Re-plate” MDM had higher expression of ADA2, FCGR2B, HLA-DMB, and CXCL12 in high density cells. In addition, “Re-plate” low density cells had higher expression of CCL3, IL1B, and FASN ( Figure 6F ). After stimulation with LPS/IFNγ, gene expression patterns in MDM were largely similar to those in iPSC-Macs, with higher expression of cytokines and chemokines at low density, and higher expression of surface markers as well GLUL and LC3A at high ( Figure 6G ). However, some features were unique to MDM, namely higher expression of the anti-inflammatory molecule TGFB, and of the glycolysis related HK1 in “Sort & plate” MDM. Finally, secretion of TNFα and IL-6 by MDM after TLR2 stimulation reflected iPSC-Macs, declining significantly as cell density increased regardless of isolation method ( Figures 6H, I , 1B , S1D ). Therefore, MDM also display density dependent regulation of their inflammatory state consistent with that observed in iPSC-mac, with some variation depending on isolation method used.

We next examined whether the human monocytic THP-1 immortalized cell line showed density dependent inflammatory features. Curiously, while TNFα secretion per cell declined as density increased after TLR2 stimulation, IL-6 secretion did not ( Figure 6J ). This IL-6 pattern was also seen with the mouse peritoneal macrophage-like RAW264.7 cell line, both cultured in standard (10%) serum conditions or low (1%) serum conditions to slow down cell replication ( Figure 6K ). We conclude that this density dependent homeostatic function displayed by authentic macrophages is not-well modelled by immortalized cell lines.

Finally, KEGG analysis for pathways associated with human disease were, unsurprisingly, enriched for inflammatory diseases. However, strikingly, half of the top 10 enriched terms were related to neurodegenerative diseases, including Amyotrophic lateral sclerosis (ALS), Parkinson’s, Alzheimer’s, and Huntington’s disease ( Figure 7A ). iPSC derived microglia (iMG) have emerged as a powerful tool for studying neurodegeneration, with numerous protocols described for their derivation in vitro (70, 71). Given that these protocols use a wide range of plating densities for generating microglia, (~2x10⁴ – 1x10⁵/cm²), it is important to determine whether density dependent changes in phenotype are also apparent in iMG. Low density iMG had improved morphology compared to high ( Figure S7 ) and fewer cells were lost during differentiation. The typical genetic markers of microglia CX3CR1, TMEM119, CR1 and GPR34 were not significantly affected by density (Figure 7B). However, 2 genes identified by genome wide association studies (GWAS) for Alzheimer’s disease, APOE and TREM2 (72), were upregulated at high density. Consistent with our iPSC-Mac results, IL-1β transcription was increased in low density microglia. In conclusion, macrophage density influences microglia differentiation and may be relevant to the breakdown of homeostasis in neurodegenerative disease.