Axon-seq Nijssen et al, 2018

SUPPLEMENTAL METHODS, FIGURES AND FIGURE LEGENDS

Supplementary experimental methods

Mouse embryonic stem cell lines

Mouse embryonic stem cell (mESC) lines expressing GFP under the Hb9 promoter were generously provided by Sebastian Thams (Tung et al., 2015). We used a control parental line (passage 14) and a cell line overexpressing human SOD1G93A (passage 12).

Differentiation of mouse embryonic stem cells into motor neurons mESCs were expanded on tissue culture dishes coated with 0.1% gelatin (Sigma-Aldrich) in ES media for two days: Knock-Out DMEM with 15% Knock-Out serum, 2mM L-glutamine, 100nM non-essential amino acids, 100 U/ml each of penicillin and streptomycin, 80nM β- mercaptoethanol (all Invitrogen) and 106 U/ml leukaemia inhibitory factor (Merck-Millipore). Then, ES were dissociated and purified for 30 minutes on gelatin-coated plates. Embryoid body formation was conducted for 2 days by culturing mESCs in suspension (4x105 cells/ml media) in neural differentiation media: 1:1 mix of Neurobasal and DMEM/F12 (Invitrogen) supplemented with 1X B27 (Invitrogen) 2mM L-glutamine, 100nM β-mercaptoethanol and 100 U/mL each of penicillin and streptomycin. During the initial day of embryoid body formation the media was kept in constant swirling motion at 30 rpm to ensure that uniformly sized spheres were formed. Embryoid bodies were subsequently patterned by the addition of 500nM of smoothened agonist (SAG, R&D) and 100nM retinoic acid (RA, Sigma-Aldrich) to the neural differentiation media for four days (Nichterwitz et al., 2016). Embryoid bodies containing motor neurons were dissociated for 20 minutes using TrypLE express (Invitrogen), filtered through a 40 µm cell strainer (Fisher Scientific) and spun down for 5 minutes at 1200 rpm for cell plating into the microfluidic devices.

Culturing of motor neurons in microfluidic devices

Microfluidic devices (SND150, Xona microfluidics) were attached to sterilized round cover glasses (28mm diameter, Menzel-Gläser) by drying these extensively and applying gentle force to seal them on the glass coverslips. Both compartments were then coated overnight with 15 µg/ml poly-L-ornithine (Sigma-Aldrich). The secondary coating consisted of a mix of 2 µg/ml of fibronectin (Sigma-Aldrich) and 10 µg/ml of laminin (Sigma-Aldrich).

Motor neurons were resuspended at 100,000 cells/µl and loaded into the microchannel in a 5µl droplet. At this point, neural differentiation media was supplemented with 10 ng/ml of GDNF, BDNF, CNTF and NT3, 200µM of ascorbic acid (Sigma-Aldrich), 100nM of RA and 2µM of 5-Fluoro-2′-deoxyuridine (5FdU, Sigma-Aldrich). Cells were allowed at least 30 minutes to attach in their compartment, after which media was added into the adjacent wells. The axonal compartment was filled with similar media, but instead containing 50ng/ml of GDNF and BDNF to facilitate axonal recruitment. Volumes in the wells were adapted to Axon-seq Nijssen et al, 2018 ensure a flow across each chamber to gradually provide media, and to establish a flow and thus trophic factor gradient from the axonal compartment to the somatic compartment. Media in the devices was changed daily.

Axonal recruitment was stopped after 2 days. The concentrations of GDNF and BDNF in the axonal compartment were reduced to 5 ng/ml. In the motor neuron compartment, trophic factor concentrations were doubled from this point onwards, to 20 ng/ml. Additionally, RA was no longer included in the motor neuron media.

Harvesting for RNA sequencing and immunocytochemistry

The cultures were harvested after one week in the microfluidic devices using a lysis solution composed of nuclease free H2O (Gibco) with 2% triton X-100 (Sigma-Aldrich) and RNAse inhibitor (1.5U, Takara). Lysis solution (75 µl) was flushed through each compartment separately (one after the other) and lysis was allowed to progress for two minutes, after which all the lysis buffer was collected and snap-frozen on dry ice until further processed. For immunocytochemistry cells were fixed in the devices with 4% PFA for 20 min. All subsequent immunocytochemistry procedures and imaging were performed on intact devices.

Immunocytochemistry

After fixation, cells were permeabilized with 0.1% triton X-100 and blocked with 10% donkey serum. Primary antibody incubation was performed overnight at 4°C, and secondary antibody incubation for 1h at RT. A full list of primary antibodies is provided in Supplemental Table S1. Secondary antibodies were conjugated with Alexa Fluor 488, 568 and 647. Nuclei were counterstained with Hoechst 33258. cDNA library preparation

To prevent RNA degradation all procedures that allowed for it were carried out rapidly, and work surfaces and equipment were cleaned with RNaseZAP (AM9788M, Thermofisher) or DNAoff (9036, Takara). For all experiments, reagents were of molecular biology/PCR grade if available. Only tubes that were certified nuclease-free were used. Cell preparations harvested from the microfluidic devices were quickly thawed, vortexed and spun down. The volumes of axonal samples were always measured and 1µl of RNAse inhibitor was added before reducing the volumes from approximately 75µl to 12-15µl using a concentrator (5301, Eppendorf) set to aqueous solution mode at room temperature. For reverse transcription (RT) of axonal samples, 10µl of the lysates were used after concentration and the volumes of the reactions scaled-up accordingly (2X) to a final volume of approximately 20µl. Only 5µl of the lysates from somatodendritic preparations were directly used for the RT reactions, which were carried out in a final volume of 10 µl. Further library preparation of cell culture samples as well as human motor neurons for Illumina sequencing was carried out using the Smart- Axon-seq Nijssen et al, 2018

Seq2 protocol (Picelli et al., 2013) with some previously described modifications (Nichterwitz et al., 2016).

RNA-seq data processing

Samples were sequenced on the Illumina HiSeq2000 or HiSeq2500 platforms, generating reads of 43 and 51bp, respectively. For mouse cell culture samples, reads were mapped to the mm10 mouse reference genome and the human SOD1 (hg38/GRCh38) with HISAT version 2.0.3 (Kim et al., 2015), using publicly available infrastructure from the main Galaxy server, available at www.usegalaxy.org. Aligned reads were then extracted and assigned using the GenomicAlignments package (version 1.8.4, (Lawrence et al., 2013)) in R, with the function summarizeOverlaps, mode set to ‘union’. Normalized RPKM values were calculated using gene exon lengths from mm10 Ensembl version GRCm38.87 for mouse. Quality control was performed and the exclusion criteria were: < 300,000 (axons) or < 500,000 (somas) uniquely mapped reads, Spearman correlation to other samples < 0.4 or < 2,500 genes expressed at RPKM > 0.1. If one or more of these criteria were not met, samples were excluded. To ensure that axonal samples to be used for analysis were not cross- contaminated with cell somas, during either the cell seeding process or the lysis, we conducted Spearman correlation, unsupervised hierarchical clustering and principal component analysis of all samples that passed the quality control. Axonal samples that clustered together with soma samples in both analyses were removed from further analyses. These samples typically displayed numbers of detected in the range for soma samples rather than other axonal samples.

RNA-seq data analysis

Differential analysis was performed in R using DESeq2 version 1.12.4 (Love et al., 2014). Only genes with expression in at least four samples (irrespective of sample type) were considered for analysis. P-values were adjusted for multiple testing using the default Benjamini & Hochberg correction with FDR set to 10%, after which an adjusted p-value of < 0.05 was considered significant. All heatmaps were generated in R using euclidian distance as clustering method where applicable. analyses were conducted using PantherDB (Mi et al., 2017) with the full Mus musculus background gene list. STRING analyses were conducted using the public infrastructure available online (Szklarczyk et al., 2015). Confidence was increased to > 0.7 and text-mining was deselected to ensure empirically validated hits. Gene set enrichment analyses (GSEA) were performed in R using the fgsea-package (version 1.2.1, available on: https://github.com/ctlab/fgsea). Gene sets were obtained from the Broad Institute Molecular Signatures Database (MSigDB v6.1, available on: https://software.broadinstitute.org/gsea/msigdb/). The used gene sets for testing were the hallmark (H), curated (C2) and GO-term-derived (C5) gene sets, as well as two custom gene sets for nuclear- and mitochondrial-encoded subunits for the respiratory electron chain (Fig. 2G, H). Axon-seq Nijssen et al, 2018

Use of published datasets

All data used in cross-comparisons of datasets were obtained from the NCBI Gene Expression Omnibus except data pertaining to the study by Rotem et al. (2017), as no GEO submission was available. Here, a table of raw counts was used that was in a supplemental table of the article. The GEO accession numbers and their respective study references for the axonal transcriptomics studies are as follows: GSE22638 (Gumy et al., 2011), GSE51572 (Minis et al., 2014), GSE66230 (Briese et al., 2016), GSE59506 (Saal et al., 2014). Data from single mouse in vitro motor neurons was obtained from GSE76514 (Nichterwitz et al., 2016). Data from purified astrocytes, microglia and OPCs was obtained from GSE52564 (Zhang et al., 2014). In all cases for RNAseq, data was remapped and RPKM values calculates using the pipeline described above, to avoid differential mapping bias. As this was not possible for the data from Rotem et al. the raw counts were used and converted to RPKMs.

Axon-seq Nijssen et al, 2018

Supplementary figure legends

Supplemental Figure 1. Characterization of motor neuron cultures and quality control of axonal samples. (A) Expression of all Hox transcription factors, as well as Isl1, Foxp1 and Phox2b in the somatic samples and in single picked Hb9::GFP positive motor neurons. (B) Expression of marker genes indicates little glial contamination in the axon samples and enrichment for cytoskeletal transcripts encoding neurofilaments and peripherin. (C) cDNA yield of axonal samples as compared to somatodendritic and single cell samples reveals that axonal samples have a similar amount of cDNA after library preparation as single cell sequencing libraries. (D) Analysis of the number of detected genes in axon samples, bulk motor neuron cultures and single motor neurons reveals that axons have significantly fewer detected genes than soma samples. (E) Heatmap of 200 differentially expressed genes, the top 100 enriched in soma and the top 100 enriched in axon by p-value. (F) Heatmap of all transcripts that were not detected in axons, but present at an average of > 1 RPKMs in somas (n = 2,903 genes).

Supplemental Figure 2. STRING analysis on motor axon-specific genes. STRING analysis for -protein interactions based on the 396 genes specific for motor axons reveals several clusters of interactions relating to ribosome biogenesis, vesicle transport (ER- golgi and synaptic), mRNA splicing and distinct mitoribosome subunits.

Supplemental Figure 3. STRING analysis on DRG axon-specific genes. STRING analysis for protein-protein interactions based on the 427 genes specific for DRG axons reveals among others interaction clusters for mRNA splicing, rRNA processing and distinct collagen subtypes.

A B Isl1 Axon Foxp1 Foxp1 Soma Soma Phox2b Nefl Hoxa1 Single MN Hoxb1 Single MN Hoxd1 Nefm Astrocyte Hoxa2 8 log2(RPKM) Nefh Hoxb2 Microglia Hoxa3 6 Map2 OPC Hoxb3 4 Mapt Hoxd3 2

Hoxa4 log2(RPKM) Hoxb4 0 Tubb3 10 Hoxc4 Gap43 Hoxd4 −2 Hoxa5 Prph 5 Hoxb5 Hoxc5 Chat Hoxa6 0 Hoxb6 Vacht Hoxc6 Hoxa7 Gfap Hoxb7 Eaat1 Hoxb8 Hoxc8 Aqp4 Hoxd8 Hoxa9 Iba1 Hoxb9 Hoxc9 Ccl3 Hoxd9 Hoxa10 Pdgfra Hoxc10 Sox10 Hoxa11 Hoxc11 Mog C D 104 25,000 RPKM > 0.1 RPKM > 1 103 20,000

15,000 yiel d 102

cDN A 10,000 Detected gene s 101 5,000

100 0 Axon Soma Single MN Axon Soma Single MN E F

Axon Axon Soma Soma log2(Fold Change)

8 log2( 2 6 1 4 RPKM) 0 2 −1 0 −2 −2

Supplemental gure 1 A STRING analysis - Motor axon specific (n = 396 genes)

Vps37d Orc5 Gpc4 Chmp1b Orc6 Cspg5 Katnb1 Katna1 Krt1 Sdsl Gpn3 Krt10 Bckdha Gpn1 Fdx1 Grp Itga2b Hscb Rars2 Cav2 Cacna2d3 Tpd52l1 Yars2 Agpat4 Cav1 Rasgrp2 Sh3gl2 Mrps31 Mrps27 Edn1 Lama5 Mcat Mrps22 Mrpl47 Fbxl4 Mboat2

Ctgf Sv2a Mrps18c Bet1l Cbr4 Vamp1 Fbxo2 Atg7 Pnpla6 Cog8 Fbxo44 Gng7 Bet1 Atg4a

Stx18 Aldh5a1 Dolk Dpm1 Rgs19 Vps45 Tada2a Gnaz Adcyap1 Taf5 Bnip1 Mrto4 Cln5 Cln3 Fbll1 Pcsk4 Sst Pomc Fbxo8 Grwd1 Ell2 Pak1ip1 Ramp3 Pou2f1

Noc4l Gtf2h2 Adora1 Znrd1 Tcea1 Rrs1 Cdkn2b Exosc8 Rbm34 Hdac11 Optn Anapc10

Sfi1 Haus1 U2af1l4 Nme6 Hdac1

Nudt21 Coq5 Haus8 Iqcb1 Aaas Gemin7

Med30 Snrnp40 Snrnp48 Cenpw Ruvbl1

Coq2 Coq4 Nup37

Med4 Med31

Actr6 Lsm10 Dsn1 Pmf1

Lrrc51 A STRING analysis - DRG axon specific (n = 427 genes) Taf2 Brd8 Ino80 Baz1b

Kat2a Nfatc3 Ep400 Myo1c

Mef2c

E2f1 Cabin1

Kifc3 Kif3b Cog1 Rb1 Copb1

Nbas

Kdm4b Copg2 Surf6 Nktr

Kif15 Rnf168 Orc2 Kif23 Cdca3 Nusap1 Ddx6

Utp14a Rbm19 Mdc1 Pcnt Gnl3l

Mcm5 Pcm1 Vps26b Sorcs2 Prpf4 Nol10 Fbxw8 Ftsj3 Rbm22 Chd9 Plk1 Utp6 Tgs1 Nde1 Nasp Prcc Ptpdc1 Ddx46 Ednrb Huwe1 Rbbp6 Isy1 Lpar1 Taok1 Rrp12 Ppp2r5a Lsg1 Rbm5 Aqr Rrm1 Taok2 Sympk Ogt Map3k3 Amotl1 Dhx38 Dhx16 Wdr82 Shoc2 Lats2 Ppp1r12a Map2k3 Skiv2l Hcfc1 Ppp1r10 Las1l Decr1 Hcls1 Smarcd1 Cybb Chd8 Ss18 Col3a1 Coro1a

Iars2 Plxna3 Pkn2 Prpf18 Inppl1 Col1a2 Col5a3 Dpysl4 Gart Kdm3b

Umps Pikfyve Pkn1 Rac2 Arhgap28 Mpp2 Adss Pik3c2a Arhgef2 Akap13 Ulk2

Eaf1 Stat2 Hspa4l Aff1 Pi4kb Rhoj Arap1 Prkce Nfkb1 Arhgap11a Lamb1 Rb1cc1 Atg13 Nid1 Stat3 Arhgap32 Ocrl Dnajc6 Foxo4 Prkcd Cdk13 Erc1 Med24 Nr2f2 Snx9 Rab8b Gtf2h1 Med13l Il6st Plcb1 Hip1 Med23 Ncor1 Gfap Ext2 Gpc1 Eif2ak4 Med15 Med12 Epn2 Cbl Ngfr Ercc5

Itpr3 Ntrk1 Nedd4l Stam2 Rara Eif2a Phc3 Suz12 Mbp Arrdc4 Scmh1

Phf8 Gas7 Pknox1 Dvl3

Mafb Sorbs1 Eif5b Prickle1 Suv39h1 Pxn

Plp1 Kdm5b Mta2 Daam2

Pbx2 Git2 Gatad2b Eif4g1 Lims1 Gab1 Plod1 Kdm2a

Upf2

Axl

Kdm5a