Regionalized neural differentiation of hESCs
RC17 hESCs from Roslin Cells (Edinburgh, UK), normally karyotyped and mycoplasma-free, were maintained on Laminin 521 (Biolamina) coated culture dishes (Sarstedt) in StemMACS iPS Brew XF medium (Miltenyi Biotec) and passaged with EDTA (0.5 mM) once weekly. The RC17 cell line used for this work is deposited in the UK Stem Cell Bank (https://nibsc.org/ukstemcellbank), and is registered in the online registry for human pluripotent stem cells hPSCreg (https://hpscreg.eu/, number RCe021-A).
The cells were differentiated towards progenitors of dorsal forebrain (dFB), ventral forebrain (vFB), dorsal midbrain (dMB), rostral ventral midbrain (rVM), caudal ventral midbrain (cVM), dorsal hindbrain (dHB), and ventral hindbrain (vHB) fates. For all conditions, media composition, coating, seeding densities and replating steps were followed until day 16 as previously described [11, 12]. All conditions received dual SMAD inhibition (SB431542 10 µM and Noggin 100 ng/ml) from day 0–9 of differentiation. Patterning into each of the different regions was obtained by differential addition of patterning factors CHIR99021 (referred to as CHIR), SHH-C24II (referred to as SHH) and FGF8b, all from Miltenyi Biotec, as follows: dFB (no additional factors added), vFB (SHH 300 ng/ml day 0–9), dMB (CHIR 0.7 µM day 0–9 + FGF8b 100 ng/ml day 4–16), rVM (CHIR 0.7 µM day 0–9 + SHH 300 ng/ml day 0–9), cVM (CHIR 0.7 µM day 0–9 + SHH 300 ng/ml day 0–9 + FGF8b 100 ng/ml day 9–16), dHB (CHIR 2 µM day 0–9) and vHB (CHIR 2 µM day 0–9 + SHH 300 ng/ml day 0–9). The basal medium used during differentiation of all regional fates consisted of DMEM/F12 (Invitrogen) mixed 1:1 with NeuroMedium, (Miltenyi), supplemented with 1% N2 supplement from day 0–11. From day 11–16, cells were kept in NeuroMedium (Miltenyi) supplemented with 2% NeuroBrew-21 (Miltenyi) as well as BDNF (20 ng/ml) and Ascorbic acid (0.2 mM). The cell culture medium was harvested from the cells on day 11, 14 and 16, and medium from all these three timepoints was pooled for vesicle preparation by centrifugation (Experiment 3b). For global secretome analysis (Experiments 1, 2 and 3a), bovine serum albumin originating from the B27 medium was first removed from the cultures by washing the cells three times in PBS on day 16. Subsequently, the cells were cultured in NeuroMedium with 0.2% N2 supplement for 24 hours until medium harvest for MS analysis on day 17. This procedure allowed to remove BSA from the input medium for MS, thereby significantly lowering the background signal on the global secretome MS analysis.
mRNA extraction and qRT-PCR
Samples were homogenized using a QiaShredder column and RNA was isolated using RNeasy Micro kit (both from Qiagen), running on a QiaCube instrument, according to the manufacturer’s procedures. Reverse transcription was performed with random hexamer primers and Maxima First Strand cDNA Synthesis Kit (Thermo Scientific) using up to 1 µg of RNA from each sample. The complementary DNA was pipetted onto a 384-well plate, together with SYBR green Mastermix (Roche Life Sciences) and primers using an automated liquid handler (I.DOT One, Dispendix). Samples were analyzed by real-time quantitative PCR on a LightCycler 480 instrument (Roche Life Sciences) using a two-step protocol with a 60°C annealing/elongation step, for 40 cycles (Ct calculations capped at 35). All qRT-PCR samples were run in technical duplicates, and the averaged Ct values were used for calculations. Data are represented using the ΔΔCt method. For each gene and samples, the fold change was calculated as the average fold change relative to undifferentiated hESCs, based on two different housekeeping genes (ACTB and GAPDH). List of primers used, and respective sequence is provided in Table 1.
Sample preparation for whole supernatant (Global Secretome) for MS
Media samples from VM cultures harvested at day 17 (Fig. 1a, Experiment 1, n = 3 biological replicates, Experiment 2, n = 5 biological replicates, Experiment 3a, n = 6 biological replicates) were prepared for mass spectrometry using in-solution digestion. Proteins were denatured with 8M Urea (50mM Ambic) and reduced with 10 mM (50mM AmBic) Dithiothreitol (DTT) at 56oC for 1h with 900 rpm shaking. Subsequently, samples were alkylated with 20mM (50mM AmBic) Iodoacetamide (IAA) in darkness for 30 min at room temperature. Ethanol was added to all samples with a ratio 1:9 (v/v, sample:ethanol) for protein precipitation and incubated over night at -20oC. After precipitation, samples were centrifuged at 12000 rpm x 15 minutes at 4oC and ethanol was removed with a pipette. Protein pellets were dried in a concentrator to remove any remaining trace of ethanol, followed by pellet dissolution in 100 µl 50 mM AmBic. For protein digestion, 2 µg Trypsin with a ratio 1:50 (w/w, Trypsin:sample) was added to each sample followed by incubation at 37oC for 17h with shaking (350 rpm). Protein digestion was stopped by reducing pH to 4 with Formic acid (v/v 10% in AmBic). iRT peptides (Biognosys AG) were added to each sample in a ratio 1:10 (v/v iRT:sample). Samples were then dried in a concentrator and stored at -80oC.
Table 1
List of human primers, by gene, full name, and forward and reverse sequence
Primer
|
Gene (full name)
|
Forward Primer
|
Reverse Primer
|
ACTB
|
actin beta
|
CCTTGCACATGCCGGAG
|
GCACAGAGCCTCGCCTT
|
CNTN2
|
contactin 2
|
GTCACGGGAGTACCAGAACG
|
TGTAGACAAAGTACTGGGCATCG
|
CORIN
|
corin, serine peptidase
|
CATATCTCCATCGCCTCAGTTG
|
GGCAGGAGTCCATGACTGT
|
CPE
|
carboxypeptidase E
|
CTCTGAAGACCTACTGGGAGGA
|
GCATTCGCAATTGGGTTACCTT
|
EN1
|
engrailed homeobox 1
|
CGTGGCTTACTCCCCATTTA
|
TCTCGCTGTCTCTCCCTCTC
|
EN2
|
engrailed homeobox 2
|
CCTCCTGCTCCTCCTTTCTT
|
GACGCAGACGATGTATGCAC
|
FEZF1
|
FEZ family zinc finger 1
|
GGTACATTCCACATTCGTGAGC
|
TCACGTGCAATAATCAAAACCA
|
FGF8
|
fibroblast growth factor 8
|
ACAGCGCTGCAGAATGCCAAGT
|
GAAGTGGACCTCACGCTGGTGC
|
FOXA1
|
forkhead box A1
|
GGGCAGGGTGGCTCCAGGAT
|
TGCTGACCGGGACGGAGGAG
|
FOXA2
|
forkhead box A2
|
CCGTTCTCCATCAACAACCT
|
GGGGTAGTGCATCACCTGTT
|
FOXG1
|
forkhead box G1
|
TGGCCCATGTCGCCCTTCCT
|
GCCGACGTGGTGCCGTTGTA
|
FST
|
follistatin
|
GATGGGAAAACCTACCGCAATG
|
CATCTGCCTTGGTACTGGACTT
|
GAPDH
|
glyceraldehyde-3-phosphate dehydrogenase
|
TTGAGGTCAATGAAGGGGTC
|
GAAGGTGAAGGTCGGAGTCA
|
GBX2
|
gastrulation brain homeobox 2
|
GTTCCCGCCGTCGCTGATGAT
|
GCCGGTGTAGACGAAATGGCCG
|
GDF7
|
growth differentiation factor 7
|
GACGCTGCTCAACTCCATGGCA
|
TTGGCGGCGTCGATGTAGAGGA
|
HOXA1
|
homeobox A1
|
GTACGGCTACCTGGGTCAAC
|
ACTTGGGTCTCGTTGAGCTG
|
HOXA2
|
homeobox A2
|
CGTCGCTCGCTGAGTGCCTG
|
TGTCGAGTGTGAAAGCGTCGAGG
|
HOXA3
|
homeobox A3
|
GGCCAATCTGCTGAACCTCA
|
GAGTTCAGATAGCCACCGGC
|
HOXB1
|
homeobox B1
|
GGCCTTCTCAGTACTACCCTCT
|
CCGTAGCTCGAGGGATGAAAAT
|
IRX3
|
iroquois homeobox 3
|
GGCTTGCGCCCCGTAGAAATGT
|
AGGAGCCAGGTCAGGTCCGAAC
|
LGI1
|
leucine rich glioma inactivated 1
|
CAACAATCTCCAGACACTCCCA
|
CCCCTCAGGTCCACATTTGTTA
|
LHX2
|
LIM homeobox 2
|
GGGCGACCACTTCGGCATGAA
|
CGTCGGCATGGTTGAAGTGTGC
|
LMX1A
|
LIM homeobox transcription factor1a
|
CGCATCGTTTCTTCTCCTCT
|
CAGACAGACTTGGGGCTCAC
|
NKX2-1
|
NK2 homeobox 1
|
AGGGCGGGGCACAGATTGGA
|
GCTGGCAGAGTGTGCCCAGA
|
NKX6-1
|
NK6 homeobox 1
|
GGATCCCAACTCGGACGACGAGA
|
AGGATGAGCTCTCCGGCTCGG
|
OTX2
|
orthodenticle homeobox 2
|
ACAAGTGGCCAATTCACTCC
|
GAGGTGGACAAGGGATCTGA
|
PAX5
|
paired box 5
|
CCCCATTGTGACAGGCCGTGAC
|
TCAGCGTCGGTGCTGAGTAGCT
|
PAX6
|
paired box 6
|
TGGTATTCTCTCCCCCTCCT
|
TAAGGATGTTGAACGGGCAG
|
PAX7
|
paired box7
|
CTTCAGTGGGAGGTCAGGTT
|
CAAACACAGCATCGACGG
|
PAX8
|
paired box 8
|
ATAGCTGCCGACTAAGCATTGA
|
ATCCGTGCGAAGGTGCTTT
|
PDGFC
|
platelet derived growth factor C
|
ACAAGGAACAGAACGGAGTACA
|
GTATGAGGAAACCTTGGGCTGT
|
SERPINF1
|
serpin family F member 1
|
TCGGACCCTAAGGCTGTTTTAC
|
CTTTCAGGGGCAGGAAGAAGAT
|
SHH
|
sonic hedgehog
|
CCAATTACAACCCCGACATC
|
AGTTTCACTCCTGGCCACTG
|
SIX3
|
SIX homeobox 3
|
ACCGGCCTCACTCCCACACA
|
CGCTCGGTCCAATGGCCTGG
|
SIX6
|
SIX homeobox 6
|
CTCAACAAGAATGAGTCGGTGC
|
ACTCCTTGGTGAACTTGTGGTT
|
SOX10
|
SRY-box 10
|
CTTTCTTGTGCTGCATACGG
|
AGCTCAGCAAGACGCTGG
|
TBR1
|
T-box, brain 1
|
TCGTCCCCGCTCAAGAGCGA
|
CCTTGGCGCAGTTCTTCTCGCA
|
TFF3
|
trefoil factor 3
|
TCTGGAGCCTGATGTCTTAACG
|
GACGCAGCAGAAATAAAGCACA
|
WNT1
|
Wnt family member 1
|
GAGCCACGAGTTTGGATGTT
|
TGCAGGGAGAAAGGAGAGAA
|
WNT3A
|
Wnt family member 3A
|
GCGATGGCCCCACTCGGATACT
|
TAGCTGCCCAGAGCCTGCTTCA
|
Preparation of vesicle-enriched samples for MS
To enrich for secreted vesicles, media samples harvested at day 11, day 14 and day 16 (see Fig. 1a, Experiment 3b, n = 6 biological replicates) were run in a differential centrifugation protocol in the following order: 300 g x 10 min at 4oC, 2000 g x 10 min at 4oC and 10 000 g x 30 min at 4oC. In between each centrifugation step, the supernatant was transferred to new tubes. Media samples from the same cultures were pooled and transferred to ultracentrifugation tubes. Samples were ultra-centrifuged at 100 000 g x 70 minutes at 4oC. The supernatant was discarded and 12 ml 50 mM AmBic was added to the top of each tube to wash the pellet, followed by another ultra-centrifugation step at 100 000 g x 70 minutes at 4oC. After centrifugation, the top 11 ml of media was discarded while the remaining 1 ml volume was mixed with a pipette to dissolve the vesicle pellet. The 1 ml sample was then transferred to new tubes for MS sample preparation. Sample volumes were reduced to 100 µl using a concentrator, followed by the addition of 50 µl RIPA buffer for vesicle lysis and protein denaturation. To further improve lysis, samples were placed in a Bioruptor 300 sonication system (Diagenode) and run for 50 cycles (High Power 15s/ OFF 15 s) at 4oC. After lysis, proteins in the samples were reduced, alkylated and precipitated according to the method for the whole supernatant samples as described above. After precipitation, samples were centrifuged at 14 000 rpm x 15 min at 4oC and the supernatant was discarded. Samples where further dried in a concentrator to remove any trace of ethanol. To dissolve the pellet, 50 µl AmBic (100 mM) was added to each sample. In order to remove glycosylations on Asparagine residues, 1.5 µl PNGase F (Promega) was added to each sample and incubated for 18h with little shaking. For protein digestion 1.4 µg Trypsin was added to each sample with a ratio 1:50 (w/w, Trypsin:sample) and incubated at 37oC for 22h with shaking (350 rpm). Protein digestion was stopped with 10 µl Formic acid (v/v 10% in AmBic). Samples were dried in a concentrator and stored in -80oC.
Data-dependent acquisition MS runs (DDA)
Supernatant samples from cVM and rVM (Experiment 1) were run in DDA mode on a Q Exactive Plus (Thermo Fisher Scientific) to be used for subsequent global DDA analysis. An EASY-nLC 1000 ultra-high-performance liquid chromatography system (Thermo Fisher Scientific) was connected to the MS instrument. Peptide separation was performed on an EASY-Spray column (ES802, Thermo Fisher Scientific) by running a linear acetonitrile gradient going from 5–30% solvent B (0.1% formic acid in acetonitrile) for 90 minutes. As solvent A, 0.1% formic acid was used. MS1 spectra were acquired in profile mode with a resolution of 70 000. In each cycle, the top 15 most intense precursor were selected in MS1 for fragmentation, but with a dynamic exclusion time of 20 s. Acquired MS2 spectra were centroided, with a resolution of 17 500. Normalized collision energy for fragmentation (NCE) was set to 30. Scan range in MS1 and MS2 was set to 400–1600 m/z and 200–2000 m/z respectively. Automatic gain control (AGC) target was set to 1e6 in both MS1 and MS2. Maximum ion injection time (IT) was set to 100 ms in MS1, and 60 ms in MS2.
In order to build sample-specific spectral libraries for later DIA analyses (Experiment 3), supernatant samples from cVM and rVM (global DIA and vesicles DIA dataset), were run on a Q Exactive HF-X (Thermo Fisher Scientific) in DDA mode. Connected to the MS instrument was an EASY-nLC 1200 ultrahigh-performance liquid chromatography system (Thermo Fisher Scientific). An EASY-Spray column (ES803, Thermo Fisher Scientific) separated peptides in a non-linear acetonitrile gradient for 2h (solvent B | 1–7%:8 min, 7–12%:15 min, 12–27%:65 minutes, 27–32%:15 min, 32–37%:9 min, 37–52%:8 min, 52–90%: 2 min). MS1 spectra recorded in profile mode had a resolution of 120 000. The top 20 most abundant precursors were chosen for fragmentation in each cycle, and the dynamic exclusion time was set to 15 s. Centroided MS2 spectra were acquired at a resolution of 15 000, with NCE = 27. Scan ranges were set to 350–1650 m/z in MS1, and 200–2000 m/z in MS2 respectively. The AGC target was set to 3e6 in MS1, and 1e5 in MS2. The maximum IT was set to 20 ms in MS1, while it was set to 20 ms in MS2.
Data-independent MS acquisition (DIA)
Samples for all DIA analyses were acquired on a Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific), using the same liquid chromatography (LC) system and gradient settings as for the global DDA runs to build spectral libraries. For data-independent acquisition (DIA), the instrument method was set to acquire a full MS1 scan (resolution 120 000, scan range: 350–1650 m/z) in profile mode, followed by 44 variable MS2 windows (resolution 30 000) with the following ranges: 350–371, 370–387, 386–403, 402–416, 415–427, 426–439, 438–451, 450–462, 461–472, 471–483, 482–494, 493–505, 504–515, 514–525, 524–537, 536–548, 547–557, 556–568, 567–580, 579–591, 590–603, 602–614, 613–626, 625–638, 637–651, 650–664, 663–677, 676–690, 689–704, 703–719, 718–735, 734–753, 752–771, 770–790, 789–811, 810–832, 831–857, 856–884, 883–916, 915–955, 954–997, 996–1057, 1056–1135 and 1134–1650 m/z. A stepped NCE was used for fragmentation (NCE = 25.5, 27, 30). AGC targets were set to 3e6 in both MS1 and MS2. Maximum IT was set to 60 ms in MS1 and ‘auto’ in MS2.
For later spectral library building, pooled supernatant samples (Global) and vesicle samples respectively, were run with gas-phase fractionated (GPF) DIA methods. For the pooled supernatant samples, there were 6 methods with DIA windows covering different MS1 ranges (400–500 m/z, 500–600 m/z, 600–700 m/z, 700–800 m/z, 800–900 m/z, 900–1000 m/z). Centroided MS1 and MS2 spectra were recorded with a resolution of 30 000. For the pooled vesicles samples, there were 10 GPF-DIA methods with DIA windows covering 10 different MS1 ranges respectively (300–400 m/z, 400–500 m/z, 500–600 m/z, 600–700 m/z, 700–800 m/z, 800–900 m/z, 900–1000 m/z, 1000–1100 m/z, 1100–1200 m/z, 1200–1650 m/z).
For each GPF-DIA method, a set of 51 overlapping DIA windows with a fixed window size of 4 m/z were acquired to cover the full MS1 ranges. The only exception was the GPF-DIA method for the 1200–1650 m/z range, having a fixed window size of 18 m/z. The AGC target was set to 3e6 in MS1, and 1e6 in MS2.
DDA-based spectral library generation
DDA MS raw files belonging to Experiment 3 (Global and Vesicles) were imported into Fragpipe v.16.1-build5 (https://github.com/Nesvilab/FragPipe). As database, the human proteome FASTA file was used (UP000005640, Uniprot/Swissprot release 21_03) with decoys appended (reversed target sequences). To build the spectral library, the default ‘SpecLib’ workflow was loaded and the default settings for all tools were used. In this workflow, the database search engine MSFragger v3.3 [13] was employed to identify MS/MS spectra, followed by Percolator [14] for confidence estimation. Protein grouping and post processing was performed using ProteinProphet [15] and Philosopher [16] followed by spectral library building with EasyPQP (https://github.com/grosenberger/easypqp).
DIA-based spectral library generation
DIA raw files were loaded into DIA-NN v.1.8 [17] to build a wide-window DIA spectral library for the global dataset and the vesicle dataset respectively. Confidently identified spectra (q-value < = 0.01) were extracted from each DIA file to be included in the final library. Narrow-window libraries were also built in DIA-NN for both datasets, using acquired GPF-DIA runs. Similarly, wide-window DIA spectral libraries were built for both datasets in Fragpipe v.16-build5 using the existing workflow ‘MSFragger-DIA-wide-window-SpecLib’. Also, narrow-window spectral libraries were built with the workflow option: ‘MSFragger-DIA-narrow-window-SpecLib’ using default settings. For all spectral libraries the canonical human proteome FASTA database was used (UP000005640, Uniprot/Swissprot release 21_03).
Super spectral library generation
In total, ten different spectral libraries were built for Experiment 3, five for each of the analyses, Global and Vesicles. As different library building strategies resulted in slightly different targets, the libraries were imported into R (v.4.2.1) and combined into non-redundant super spectral libraries, one for each dataset, using a custom R script.
Data analysis of global DDA runs
Raw DDA files acquired by DDA on the Q Exactive Plus were loaded into MaxQuant v.1.6.1.0 [18–20] for label-free quantification of proteins. DDA MS files were put in different parameter groups based on their Experiment (1 or 2) to ensure batch-specific normalization and quantification with the MaxLFQ algorithm [21]. Identification settings used the default false-discovery rate of 1% on protein, peptide and peptide-spectral-match level. As FASTA database, the human canonical proteome was used (UP000005640, Uniprot/Swissprot release 21_03). Match-between-runs to transfer identifications between runs was enabled. Carbamidomethylation on Cystein (UniMod:4) was set as fixed modification and variable modifications were oxidation on Methionine (Unimod:35) and acetylation on protein N-terminal (UniMod:1). For label-free quantification, it was required that at least one peptide was identified from MS/MS for pairwise comparisons. The minimum LFQ peptide ratio was set to 1, in order to allow more low-abundant proteins to be quantified.
Data analysis of DIA runs
Acquired DIA raw files acquired on the Q Exactive HF-X were searched against their respective super spectral library in DIA-NN v.1.8 [17]. The quantification strategy was set to ‘Robust LC (high accuracy)’ while cross run normalization was set to RT-dependent (default). Based on the median recommended MS1 accuracies reported by DIA-NN for each run, the MS1 accuracy was set to 7.96 ppm for the Global DIA dataset (Experiment 3a) while being set to 8.48 ppm for the Vesicles dataset (Experiment 3b). MS2 accuracies were automatically set by DIA-NN to 20 ppm for both analyses. Relaxed protein inference was enabled in DIA-NN to avoid the assignment of the same protein to more than one group during protein inference. The human proteome FASTA file (UP000005640, Uniprot/Swissprot release 21_03) was used for annotations in DIA-NN.
ELISA
Supernatant samples were collected from the differentiating cells at the day 11 and 16 and immediately frozen. ELISA kits for the targets proteins were used in according to the manufacturer’s instructions: CNTN2, CORIN, FST, PDGFC, SERPINF1, TFF3 (all from R&D Systems), CPE (Nordic Biosite), LGI1 (Cusabio) (see Table 2). Before analysis, each supernatants sample was centrifuged at > 10.000 rpm for 10 minutes to remove cell debris. Initial tests were performed to ascertain dilution factors for the various proteins and samples, although some measurements were above or below the detection limit. Sample measurements above detection limit were excluded. Samples assayed at 1:1 dilution and with measurements below the detection limit were attributed the Minimum Detectable Dose according to the manufacturer’s information, or, in the absence, the minimum calculatable value using the respective dilution curve and 4-PL curve fit. The measured protein concentration values were then normalized to the cell count in the respective well, yielding pg.ml− 1.10–6 cells.
Table 2
Protein
|
Kit Manufacturer
|
Kit name
|
Kit target
|
Catalog #
|
CNTN2
|
R & D Systems
|
DuoSet
|
human Contactin-2/TAG1
|
DY1714-05
|
CORIN
|
R & D Systems
|
Quantikine
|
human Corin
|
DCRN00
|
CPE
|
Nordic Biosite
|
n.a.
|
human CPE
|
KBB-B314QW-96
|
FST
|
R & D Systems
|
Quantikine
|
human Follistatin
|
DFN00
|
LGI1
|
Cusabio
|
n.a.
|
human LGI1
|
CSB-EL012898HU
|
PDGFC
|
R & D Systems
|
Quantikine
|
human PDGF-CC
|
DCC00
|
DuoSet
|
DY1687-05
|
SERPINF1
|
R & D Systems
|
DuoSet
|
human SerpinF1/PEDF
|
DY1177-05
|
TFF3
|
R & D Systems
|
Quantikine
|
human TFF3
|
DTFF30
|
Table 2: List of ELISA kits, by target protein, manufacturer and kits respective name, described target and catalog number.
Statistical Analysis of ELISA and qRT-PCR data
All ELISA and qRT-PCR data was managed in Excel and statistically analyzed using GraphPad Prism 9 software, P < 0.05 was considered significant. For multi-regional comparisons, one-way analysis of variance (ANOVA) was performed followed by a Sidak multiple comparison test between the rVM and cVM and remaining regions. All datasets were tested for their normal and Log-Normal distribution (Shapiro– Wilk and Kolmogorov-Smirnov) and homoscedasticity (Brown–Forsythe) before ANOVA. Alternatively, a non-parametric Kruskal–Wallis analysis was conducted instead, followed by a Dunn’s multiple comparison test. All multiple comparison tests were corrected using statistical hypothesis testing.
For pairwise comparison between rVM and cVM, a two-tailed unpaired t-Test was performed, or in case the datasets and the Log-transformed datasets lacked a Gaussian distribution or showed significantly different variances, a Mann-Whitney test was performed instead.
For calculating the correlation between the EN1 mRNA expression and the ELISA-assayed Protein levels, a two-tailed Spearman correlation was performed on the Log-Log data. A straight, non-linear, least squares regression was fitted to the Log-Log data, computing the 95% confidence interval.
Statistical Analysis of DDA and DIA analyses
Result files from the Global analysis and the Vesicles analysis were imported into R for processing and differential expression analysis. The protein groups table (proteinGroups.txt) from the MaxQuant search was filtered to not contain decoys nor entries only identified by site. A quantitative matrix was extracted by selecting the ‘LFQ intensity’-columns from the table, and the quantitative values were subsequently log2-transformed. Imputation was applied to the matrices using the R package imputeLCMD [22] v.2.0, where the K-nearest neighbors algorithm impute values missing at random, while the ‘MinProb’-algorithm was used to impute values missing not at random. Differential expression analysis was performed by running a moderated t-test using the R package DEqMS [23] v.1.8.0 to compare samples belonging to cVM with those in rVM.
Output reports from DIA-NN, for the Global analysis and v´Vesicles analysis, were imported into R for downstream processing. Reports were filtered to only contain confidently identified entries (Global precursor q-value < = 0.01, Global protein group q-value < = 0.01). Quantitative protein groups matrices were computed with the MaxLFQ [21] algorithm, implemented in the R package ‘diann’ v.1.0.1 (https://github.com/vdemichev/diann-rpackage). Following log2-transformation, the matrices were filtered to only contain protein groups having at least 60% quantitative values evenly distributed among samples in both conditions (cVM or rVM), or at least 50% quantitative values given that all were present in one group only. Retained protein groups were then imputed using the ‘MinProb’ algorithm described above (see global DDA analysis). Similarly to the global DDA analysis, DEqMS [23] v.1.8.0 was used to perform differential expression analysis between samples in the cVM condition and the rVM condition.
GO-term enrichment analysis
A GO-term enrichment analysis for cellular components between the Global DIA dataset and the Vesicles DIA dataset was performed in R (v.4.2.1) with the package Clusterprofiler [24] v.3.18.1. To find enriched GO-terms for cellular components in the Global DIA dataset, the enrichGo function was used to query gene names for identified proteins in the Global DIA dataset against all identified gene names (Global DIA + Vesicles DIA). Inversely, all gene names in the Vesicles DIA dataset were queried against all identified gene names to find enriched cellular component GO-terms in the Vesicles dataset. Only significant results were considered (q-value < = 0.05, Benjamini-Hochberg FDR estimation [25]).