With the Matrisome project identifying in excess of 150 ECM proteins, we first began by identifying the proteins that may be expressed in the ventral mesencephalon (VM) during dopaminergic neurogenesis. Using a RNA sequencing database of mouse midbrain development (La Manno et al., 2016), we identified 74 core ECM genes that are expressed in the mouse VM between E10.5 and E14.5. From these, we selected 19 candidate genes for which there is commercially available protein and that are expressed at five reads per kilobase of transcript per million mapped reads (RPKM) or higher (Table 1). Nine of these 19 genes are enriched specifically in the VM compared to the dorsal mesencephalon and more rostral and caudal regions of the developing neural tube. A further nine genes were differentially expressed (>2 fold change) between E10.5 and E14.5, the period of peak dopaminergic neurogenesis in the mouse. Expression of the ECM proteins was also confirmed in transcriptomic data generated from human ventral mesencephalon with a number of cell types found to express the ECM proteins (La Manno et al., 2016; Kee et al., 2017). Both the Sparc and Sparc-L1 genes are expressed by endothelial cells, pericytes and radial glia in both mouse and human datasets; whereas Nell2 is expressed more broadly amongst the progenitors in the ventral mesencephalon.

The complex microenvironment created by up to 74 ECM proteins will generate multiple combinations of cell-matrix interactions that could instruct dopaminergic neurogenesis. Here we examined the role of the 19 selected ECM proteins in driving dopaminergic differentiation in an unbiased and combinatorial manner with the development of a micro-contact printing approach (Tare et al., 2009) (Figure 1A). Both single and pairwise combinations of the 19 identified ECM molecules were formed by generating an array on glass microscope slides with 190 different combinations printed with six replicates for a total of 1,140 features. Deposition of ECM proteins to the glass slide was confirmed via labelling with fluorescent antibodies (Figure 1B). The arrays were produced by transferring the 19 ECM proteins (at a concentration of 1 μg/ml) to a 384 well plate using a robotic liquid handler. The 384 well plates then acted as the source plates for the micro-contact printing of the ECM combinations onto an agarose coated glass slide.

Long-term neuroepithelial stem (Lt-NES) cells (Koch et al., 2009) were seeded onto the arrays in serum free media and agitated every hour for 4 h. Prior to plating Lt-NES cells onto the combinatorial arrays, we confirmed uniformity of cell adhesion across the array using this plating protocol by using a slide printed with just one ECM protein, laminin, used in the standard culturing protocol. The cells on the combinatorial arrays were initially exposed to sonic hedgehog (Shh) and a Wnt agonist for 4 days to specify the Lt-NES cells to a ventral mesencephalic fate (Figure 1C). Thereafter, cultures were exposed to neurotrophic factors (BDNF, GDNF) as well as ascorbic acid and cyclic AMP to induce terminal differentiation. After 7 days of culture, cell were fixed and stained with antibodies against tyrosine hydroxylase (TH) to identify dopaminergic neurons (Figure 1D). The array was imaged using an Operetta high-content microscope and the number of TH + cells on each combination of ECM protein quantified using the software Columbus (Figure 1E). This analysis revealed that the majority of proteins did not increase the number of TH + neurons, while a few combinations reduced them. There were however a small number of combinations that increased the number of TH + cells; and of these, five matricellular proteins (Sparc, Sparc-L1, Nell2, trombospondin-3 (Thbs3) and neurocan (Ncan)) were represented in various combinations more than once (Supplementary Table). We concluded from this unbiased approach examining the pro-differentiation effect of ECM proteins expressed in the VM that these proteins are promising candidates for further examination.

To validate the results seen in the array, we used standard culture platforms. Lt-NES cultures in 24 well plates were treated with the proteins either individually or in all possible combinations of the five proteins. Wells were coated with the appropriate combination of protein at a concentration of 1ug/ml per single protein. Cells were then seeded and cultured in the same manner as the ECM arrays: 4 days with media containing Shh and Wnt agonist followed by BDNF, GDNF, ascorbic acid and cyclic AMP. The validation experiments proved inconclusive for combinations involving Ncan and Thbs3. We were unable to validate the Ncan-mediated increase in differentiation, and combinations of Thbs3 combined with either Sparc or Sparc-L1 displayed inconsistent results depending on the lot of recombinant protein used. However, the experiments using Nell2, a pentameric glycoprotein that is closely related to Thbs3, and Sparc and Sparc-L1 revealed that the combination of Sparc and Sparc-L1 increased the number of TH + cells, although each protein had no effect when used on its own. Nell2 providing a further additive increase to the Sparc/Sparc-L1-driven increase in TH + neurons, resulting in a threefold increase in number as compared with cultures lacking matricellular proteins (Figures 2A,B). There was no effect of Nell2, however, when used on its own or in combination with either Sparc or Sparc-L1 suggesting a complex mechanism of action of these matricellular proteins.

Next we determined whether this increase in TH + neurons in response to Sparc, Sparc-L1 and Nell2 (a combination hereafter referred to as SSN) was a result of an increase in proliferation of the progenitors, their differentiation, or survival of the newly-generated neurons. To do this, we first examined the number of Ki67 cells (Figures 2C,D) in both control and SSN treated conditions and found no significant difference in the number of cells that remained in the cell cycle. We also examined activated caspase 3 (aC3) expression in response to SSN treatment to determine if there was a reduction in the number of apoptotic TH + neurons. Again, no significant difference was detected between control and SSN groups (Figures 2C,E). Finally, we labelled cultures with EdU on day 5 and examined the number of cells at day 8 that were positive for both EdU and TH to identify newly formed TH + neurons (ie those formed by neurogenesis from progenitors dividing at 5DIV). Here, we found a significant increase in both the number of TH + cells (7.1% vs 22,8%) and EdU + TH + double positive cells (1.3 vs 2.8%) following SSN treatment (Figures 2C,F). We conclude that SSN enhances the differentiation of VM progenitors into TH + neurons, a result that could reflect accelerated differentiation, an overall increase in neurogenesis and/or a relative increase in TH + cell differentiation.

Finally, we further examined the potential for matricellular proteins to increase the number of TH + neurons in a human embryonic stem cell-derived model that better recapitulates the human developmental trajectory and also provides direct translational relevance. Human embryonic stem cells were first induced to a neural fate and then patterned using sonic hedgehog (Shh) and a Wnt agonist as previously shown to generate rapidly dividing, FoxA2, Lmx1a and Otx2 triple positive progenitors (Nolbrant et al., 2017; Ahmed et al., 2019) (Figure 3A). Cells were replated onto wells coated with polyornithine, fibronectin and laminin as per the standard culture protocol at day 11. The matricellular proteins at an individual concentration of 1ug/ml were then added in solution as part of the culture media on day 14 and once again at day 21. The ECM molecules were added in solution so as to reduce the need for replating. Having been exposed to the matricellular proteins for a period of 14 days, cells were fixed on day 28 and examined for the number of TH positive neurons (Figure 3B). Similar to the Lt-NES cells, we found an increase (6.8 ± 0.9% on control conditions compared to 17.6 ± 2.3% following SSN treatment) in the number of TH positive cells following SSN treatment (Figure 3C).

Aminoalkyl functionalised glass microscope slides (S5641, Sigma) were dip coated with 2% agarose. The back of the slides were cleaned with a paper tissue and left to air dry coating-side up. Agarose slides were spotted using a Genetix Qarray^mini microcontact printer equipped with 32 solid aQu pins. 190 unique combinations were spotted in replicates of 6, along with alexaflour-488 as a negative control and for image alignment used for analysis. Proteins were spotted at an individual concentration of 1ug/ml. The following recombinant proteins were used: agrin, biglycan, brevican, collagen IV, fibronectin, IGFBP2, laminin-511, laminin-521, matrilin-2, neurocan, Nell2, Nov, postn, prelp, Sparc, Sparc-L1, thrombospondin-3, tsukushin, vitronectin (all purchased from R&D). Slides were stored at 4°C prior to use and were used within 48 h of fabrication.

LT-NES cells were maintained in DMEM/F12 (Invitrogen, 31331–028) maintenance medium with N2 supplement (1:100), B27 supplement (1:1,000), FGF2 (10 ng/ml, R&D), and bFGF (10 ng/ml, R&D).

For screening on ECM arrays: prior to use, ECM array slides were washes with PBS and treated with UV before seeding cells. A suspension of LT-NES cells in 1 ml of media were added to the ECM array slides placed in a 4-well rectangular culture chamber. Cells were allowed to attach for 4 h at 37°C, after which slides were washed and fresh media was added. Arrays were cultured for 4 days with N2 (1:100), B27 (1:1,000), SHH (200 ng/ml, R&D), and CT99021 (1 μM, Tocris Biosciences) added to DMEM/F12 media. After that, cells were kept in media containing N2 (1:100), B27 (1:100), GDNF (20 ng/ml, R&D), BDNF (20 ng/ml, R&D), cyclic AMP (0.5mM, Sigma, D0627) and ascorbic acid (200uM, Sigma A4403) until fixation, at which point slides were washed with PBS and fixed with 3.7% PFA for 15 min.

For validation studies, LT-NES cells were dissociated with TrypLE Select and seeded (100,000 cells per well) in 48‐well plates coated with polyornithine and matricellular proteins (Sparc (941-SP, R&D), Sparc-L1 (2728-SL, R&D) and Nell2 (8946-NL, R&D)) or laminin (), and cultured for 4 days with N2 (1:100), B27 (1:1,000), SHH (200 ng/ml, R&D), and CT99021 (1 μM, Tocris Biosciences). After that, cells were kept on media containing N2 (1:100), B27 (1:100), GDNF (20 ng/ml, R&D), BDNF (20 ng/ml, R&D), cyclic AMP (0.5mM, Sigma, D0627) and ascorbic acid (200uM, Sigma A4403) until fixation where they were treated with 3.7% PFA for 15 min. For EdU labelling, EdU was administered at day 5 for 24 h to capture dividing cells. Cells were then fixed at day 8 for analysis.

Undifferentiated RC17 ESCs (passages 30–58, Roslin Cells, hPSC reg. no. RCe021-A) were maintained in E8 medium (A1517001) on Geltrex (1%, 12760021)-coated plates and passaged weekly with EDTA (0.5 mM). To start differentiation (day 0), human ESC colonies were detached using EDTA (0.5 mM) and placed in non-treated 60-mm culture dishes in differentiation medium consisting of DMEM:F12/Neurobasal (1:1), N2 supplement (1:100), B27 supplement (1:50), SB431542 (10 μM, Tocris Biosciences), rhnoggin (200 ng/ml, R&D), SHH-C24II (200 ng/ml, R&D Systems), and CHIR99021 (0.8 μM, Tocris Biosciences). Medium was changed once on day 2. The resultant embryoid bodies were collected on day 4 and placed on polyornithine (PO)-, fibronectin (Fn)-, and laminin (Lm)-coated plates in reduced N2 (1:200) and B27 (1:100) condition. Growth and patterning factors were removed on day 9 with the cultures kept in DMEM:F12/Neurobasal (1:1), N2 supplement (1:200), and B27 supplement (1:100). On day 11 of differentiation, the cell clusters were dissociated to single cells with accutase and replated onto dry PO/Fn/Lm-coated plates in Neurobasal, B27 (1:50), BDNF (20 ng/ml), GDNF (10 ng/ml), AA (200 μM), and db-cAMP (0.5 mM, Sigma). From day 14, cultures were treated with matricellular proteins Sparc, Sparc-L1 and Nell-2 (1 μg/ml, R&D Systems) up until day 28, when they were fixed in 3.7% PFA for 15 min.

For both the cells on the arrays and the validation experiments in standard culture plates, cells were washed with PBS and blocked in PBTA (5% normal goat/donkey serum, 0.1% Triton X-100, 1% BSA) for 1 h at room temperature. Primary antibodies were incubated in PBTA and staining was carried out overnight at 4°C. Antibodies used were cleaved caspase 3 (1:200, Cell Signaling Technology, 9661S); TH (1:1,000, Millipore, AB152); Ki67 (1:500, abcam, ab15580); Click-iT EdU (Invitrogen, C10337). Corresponding secondary antibodies were Alexa Fluor Dyes (Invitrogen) and used at a dilution of 1:1,000 for 1 h at room temperature. Cells were counterstained with Hoechst 33258 (Sigma). For validation studies where cells are in culture plates, these were washed and kept in PBS. The arrays were washed and mounted with Fluoromount G Mounting Media (Thermo). Images were captured on the Operetta high-content imaging system (Perkin Elmer) or a Lecia SP8 confocal microscope (Leica). For the ECM arrays, the whole slide was scanned and images stitched together using the Columbus software (Perkin Elmer). Image quantification was conducted using an automated image analysis software package (Columbus, Perkin Elmer) using a pre-validated pipeline that quantified nuclei and cytoplasmic TH expression. Image analysis for validation studies was performed on five images per well, three wells per condition across three experimental repeats (n = 3). Cell counts were performed and percent positive cells for each marker was determined and then normalized to the control condition, with data presented as mean fold change ±standard deviation.