Animal ethics

All of the animal studies were approved by the Animal Care Committee at the University of Manitoba (protocol AUP-22-005).

Cell culture

The HDMB03 and MB3W1 group 3 MB cell lines were kindly provided by Milde et al.53 and Wölfl and colleagues54, respectively. Short tandem repeat profiling (American Type Culture Collection) was used to authenticate cell lines in 2021. For tumoursphere formation, both cell lines were dissociated into single-cell suspensions. HDMB03 cells were then resuspended in StemPro NSC Serum-Free Medium (Life Technologies) and MB3W1 cells were resuspended in stem cell media (DMEM/F-12, 2% B27, 1% MEM Vitamin Solution, 20 ng ml−1 basic fibroblast growth factor, 20 ng ml−1 epithelial growth factor, 40 U ml−1 penicillin and 40 μg ml−1 streptomycin). Suspensions were plated at clonal density (for HDMB03, three cells per μl; for MB3W1, five cells per μl) in either 6- or 24-well ultra-low-attachment plates. Tumourspheres were incubated undisturbed at 37 °C under 5% CO2 for either 3 or 5 d, counted and then measured. Tumoursphere number (5 d), tumoursphere size (3 or 5 d), cell viability (3 d) and total cell number (5 d) were evaluated. ImageJ (Fiji) was used to evaluate individual tumoursphere size. The results were displayed as the cumulative frequency distribution of the tumoursphere area. Only tumourspheres greater than 25 μm in diameter were included in these analyses55. For morpholino treatment, a custom splice-blocking morpholino to induce exon skipping was designed for PPHLN1 (ENSG00000134283.13, exon 6; 5′-AGAGTCTAGCTCAAAACTCACCTCT-3′) and MADD (ENSG00000110514.14, exon 26; 5′-GTCCCTTCCTGCCAATTTGAGAGCA-3′). Concentrations of 1, 5 or 10 μM in double-distilled water were added to single-cell suspensions and then cultured in ultra-low-attachment plates. An Annexin V Apoptosis Detection Kit (Annexin V-PE; BD Biosciences) was used to evaluate cell death, as previously described15,56.

OTX2 siRNA silencing and RNA processing

Silencer Select siRNAs 9931 and 9932 (Life Technologies) were used at 30 nM to knockdown OTX2 expression in group 3 cells, as previously described13,15. A scramble (non-silencing) sequence was used as a negative control. For bulk RNA-seq, OTX2 was silenced in tumourspheres from three independent biological replicates of HDMB03 and MB3W1 cells and silencing was confirmed by western blot 3 d following transfection, as described13,15. Total RNA was previously extracted using a Norgen RNA extraction kit (Norgen Biotek) followed by bulk RNA-seq (paired end; 150 base pairs) targeting 100 million total reads for adequate depth of coverage for splicing analyses. Sequencing was performed at the Ottawa Hospital Research Institute/StemCore laboratories as described3. The raw data generated were previously deposited in the Gene Expression Omnibus (GEO) database under the access code GSE189238.

RNA-seq, processing and analysis

Raw reads were mapped to the human genome using the hg19 reference with the STAR aligner (version 2.5.2b)57. Picard tools and SAMtools were used to remove PCR-duplicated reads and low-mapping-quality reads (mapping quality < 20), respectively. rMATS (version 3.0.9)58—an event-based tool—was used to identify differentially spliced events. We analysed five distinct alternative splicing events using rMATS: mutually exclusive exons, retained introns, skipped exons, alternative 3′ splice sites and alternative 5′ splice sites. Briefly, rMATS features a count-based model that calculates PSI scores among replicates using spliced reads and reads that map to the exon body. To call statistically significant splicing events, each splicing event had to: (1) have a ΔPSI of >10%; (2) be supported by at least 15 unique reads; and (3) have an FDR of <0.05. For gene expression analysis, aligned reads were counted using htseq-count version 0.6.1p1 (ref. 59) and normalized according to the DESeq size factors method60. To call DEGs, we used a fold change of ≥2 and an FDR of <0.05. Statistical analysis was performed and plots were generated using R environment version 3.6.0. Sashimi plots were generated using ggsashimi.py script61 with the parameter --min-coverage 5. We performed motif analyses to evaluate the enrichment of RBP motifs enriched near OTX2-regulated ASEs using a compilation of 110 known RBP motifs from the literature27,28. Briefly, we used rMAPS62 to identify known RBP motifs that were significantly enriched in OTX2-regulated ASEs compared with control exons (background). rMAPS examines adjacent intronic and exonic sequences and counts the number of times the motif matches each sequence to compute an enrichment score.

To explore the splicing landscape of group 3 and group 4 MB, we examined previously published bulk RNA-seq results for group 3 and group 4 tumours3. Briefly, we excluded exons that were not expressed in at least 25% of samples and excluded 5% of the total number of samples that had the least number of detected exons. We performed correlation-based analysis between OTX2 expression and PSI values and applied FDR correction. We considered statistically significant exons that had an adjusted P value of <0.01.

To determine whether OTX2-regulated exons recapitulate the developmental origins of group 3 and group 4 tumours, we used previously published cerebellar bulk RNA-seq data6,31 for the rhombic lip (RLVZ and RLSVZ at post-conception weeks (PCWs) 9 and 10; n = 4) and EGL (at PCW 15; n = 4). We also compared the splicing profile of each group 3 and 4 MB tumour with that of the human rhombic lip (pooled RLVZ and RLSVZ at PCWs 9 and 10; n = 4). We applied the same filters as described above. Because the depth of coverage for each tumour can impact the number of significant ASEs, only exons that were expressed in all tumours were retained (an event needs to be supported by at least 15 unique reads) for subsequent analyses. Briefly, the list of exons expressed in each tumour was extracted. Then, we removed samples that had too few or too many expressed exons and considered them as outliers (samples that were outside box plot whiskers: n < Q1 – 1.5 × interquartile range (IQR) or n > Q3 + 1.5 × IQR). Next, we compared the lists of expressed exons for each tumour individually to determine the minimal number of exons expressed across all tumours (n = 15,342). This list of minimal expressed exons was then compared with the list of significant ASEs for each group 3 and group 4 tumour, resulting in the number of events that were significantly spliced relative to the human rhombic lip. For the TCGA and GTEx datasets, PSI values were extracted from the TCGA SpliceSeq database63 and SnapMine64, respectively. Bulk RNA-seq data from retinoblastoma and normal retina samples were accessed from GEO under the accession codes GSE196420 (ref. 32) and GSE99248 (ref. 33), respectively.

For RNA-seq on PPHN1-Mo-treated cells, HDMB03 cells were treated with 1 μM Ctrl-Mo or PPHLN1-Mo for 5 d. Cells were collected and RNA was extracted using a Total RNA purification kit (Norgen Biotek). RNA-seq was performed as described above using the same alignment parameters and analysis tools.

We performed functional and pathway enrichment analysis on genes exhibiting significant changes in expression following OTX2 silencing; this was done separately for genes up- and downregulated following knockdown. Enrichment analysis was performed using the ClusterProfiler tool65 to look at the enrichment of Gene Ontology molecular function and biological process annotations and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Within each group of annotations, enrichment was calculated by comparison with the hypergeometric distribution (equivalent to a one-tailed Fisher’s exact test). Multiple testing correction was performed using the Benjamini–Hochberg approach. The KEGG pathway data were obtained from the KEGG website on 11 February 2022.

RT-PCR validations of select DSGs

The DSGs PPHLN1, PAPOLA, MADD and MBD1 were selected for further validation by RT-PCR. Total RNA was extracted from both HDMB03 and MB3W1 tumourspheres using a Total RNA Purification Kit (Norgen Biotek). First-strand complementary DNA was generated using the SuperScript III First-Strand Synthesis kit (Life Technologies). PCR was performed using the MiniAmp Plus Thermal Cycler (Applied Biosystems, Thermo Fisher Scientific) with the following parameters: 98 °C for 10 s, 55 °C for 5 s and 72 °C for 5 s over 30–35 cycles. The primer sequences for each gene evaluated are listed in Supplementary Table 16. GAPDH was used as a loading control for morpholino validation. Samples were run on 2% agarose gels and imaged on a Fusion FX Vilber Lourmat chemiluminescence imaging system.

siRNA silencing of LASR complex members and select DSGs

Gene silencing of LASR complex members and select DSGs was performed using Silencer Select siRNAs (Life Technologies) at 30 nM (HDMB03) or 15 nM (MB3W1) concentrations. A non-silencing scramble siRNA was utilized as a negative control. siRNA sequences for all genes are provided in Supplementary Table 17. Gene silencing was validated by immunoblotting. A list of the antibodies used for validation studies is provided in Supplementary Table 18.

ChIP-qPCR

HDMB03 cells were seeded at 2 × 105 cells per well under stem cell conditions and cultured for 72 h. Cells were then fixed and quenched using 1% paraformaldehyde and 0.125 M glycine, respectively. Lysates were next sonicated to 100–500 base pair fragments followed by incubation overnight with primary antibodies (Supplementary Table 18) at 4 °C. Protein A Dynabeads (Invitrogen) were used to capture antibody/chromatin complexes for 4 h followed by washing, elution and reverse crosslinking. Chromatin was purified using the Qiagen PCR Purification Kit. Chromatin was amplified using primers designed to flank known OTX2 binding sequences (Supplementary Table 16) and GoTaq qPCR master mix (Promega). Fold enrichment of OTX2 binding was calculated and graphed with Prism.

Overexpression of wild-type OTX2 and OTX2 lacking its homeodomain

Wild-type and homeodomain-deleted human OTX2 sequences cloned into the plasmid pcDNA3.1+ were purchased from GenScript. The coding sequences were amplified by PCR, adding overlap sequences to the 5′ ends of primers to allow NEBuilder HiFi assembly (New England Biolabs) with NheI/ClaI-linearized pCAG-H2B-mAG-P2A-3XHA-TurboID3, yielding pCAG-H2B-mAG-P2A-3XHA-hOTX2 and pCAG-H2B-mAG-P2A-3XHA-hOTX2ΔHD. All cloning reagents were purchased from New England Biolabs. The sequences of the resultant plasmids were verified by whole-plasmid sequencing (Eurofins Genomics).

Pseudotime analysis of cells from the developing human cerebellum

Single-nucleus RNA-seq data from the developing human cerebellum were obtained and processed as previously described31. Briefly, sample fastqs were aligned to the human reference genome hg19 and the resultant count data were interpreted in Seurat (version 4.1.0)66 using the R environment (version 4.1.3) in order to perform quality control, normalization by SCTransform67, dimensionality reduction (uniform manifold approximation and projection), clustering (shared nearest neighbour and Louvain modularity optimization) and cell type identification. Trajectory inference within the developing human glutamatergic cells was also performed as previously described3. Briefly, Slingshot (version 1.6.1)68 was used to find the expected granule neuron and UBC lineages and TradeSeq (version 1.2.01)69 was used to statistically correlate gene expression with lineage pseudotime values. The resulting significant lineage-associated genes were then used in downstream analyses.

Protein–protein interaction networks

Protein–protein interactions (PPIs) within both the OTX2-regulated DSGs list and the OTX2 TurboID hits list were determined by querying the STRING database (version 11.5)70. Briefly, gene lists were inputted and the resulting networks of PPIs were further filtered for only known physical interactions (‘Experiments = T; Databases = T’) with confidence scores of at least 0.150 (low confidence) or 0.400 (medium confidence), where indicated. This list of known PPIs was visualized using Cytoscape (version 3.8.0)71. Nodes were coloured by relevant enriched pathways in the queried lists using g:Profiler72 with the Gene Ontology Biological Processes (GO:BP), KEGG, Reactome and WikiPathways databases and an FDR threshold of 0.1. A full list of enriched pathways is provided in Supplementary Table 2. Nodes were manually rearranged to prevent node overlap; thus, the edge length is not representative of any value.

Immunoblotting

Protein quantities (10–25 μg) were loaded onto 10–12% Tris-glycine gels and resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis followed by transferral to nitrocellulose membranes, blocking with 5% milk and incubation overnight with select primary antibodies (details provided in Supplementary Table 18). Secondary antibodies conjugated to horseradish peroxidase were added for 1 h at room temperature followed by detection using SuperSignal West Pico. A Fusion FX Vilber Lourmat chemiluminescence imaging system was used to capture the images.

Immunofluorescence

HDMB03 cells were plated at 5 × 104 cells per well in 24-well plates on glass coverslips that had been coated in poly-d-lysine (Thermo Fisher Scientific) for 10 min before being washed with molecular-grade H2O. For OTX2 silencing, HDMB03 cells were treated with scrambled RNA (negative control) or Silencer Select siRNA 9931 (Supplementary Table 17). HDMB03 cells were transfected with OTX2WT or OTX2ΔHD plasmid (the latter resulting in deletion of the DNA-binding homeodomain) 24 h after plating, as described. At 72 h after HDMB03 cells were plated, they were washed three times with phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde at room temperature for 30 min. Cells were then washed three times with PBS before permeabilization with 0.1% Triton X-100 in PBS for 20 min at room temperature. After washing three times with PBS, cells were blocked with 3% lamb serum/PBS for 1 h at room temperature, followed by replacement of the blocking buffer with 1:100 primary antibody (RBFOX2, HNRNPC or HNRNPM) (Supplementary Table 18) in 1% lamb serum/PBS. After 1 h incubation at room temperature, the cells were washed three times with PBS before incubating with either anti-rabbit (RBFOX2 and HNRNPM) or anti-mouse (HNRNPC) secondary antibody. For the OTX2 silencing and OTX2 deletion mutation experiments, Alexa Fluor 488 and Alexa Fluor 594 were used, respectively (Supplementary Table 18), at a dilution of 1:1,000 for 1 h at room temperature. Cells were then washed three times with PBS and once with H2O, followed by mounting with VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories). This was performed by placing a drop of mounting medium on a glass slide and flipping the coverslip with the cells onto the drop of mounting medium. Images were taken at 40× magnification using an EVOS M5000 Imaging System and the cytoplasmic area per cell was quantified using CellProfiler73.

OTX2 TurboID and mass spectrometry

TurboID experiments were performed as previously described by Cho et al.74. NEBuilder HiFi DNA Assembly (New England Biolabs) was used to introduce the human OTX2 sequence into the plasmid pCAG-H2B-mAG-P2A-3xHA-TurboID3. The vector was linearized with KpnI and ClaI and assembled with PCR-amplified 3xHA-TurboID (FW_KpnI_Turbo and RV_Turbo_GGSGG primers) and PCR-amplified hOTX2 (using reverse-transcribed RNA (Norgen) recovered from HDMB03 cells and RNEB_ClaI_N-Otx2 and FNEB_Link_N-OTX2 primers) to generate pCAG-H2B-mAG-P2A-3xHA-TurboID-OTX2. PCR-amplified sequences in the final plasmid were verified by Sanger sequencing. The primers used were as follows: FW_KpnI_Turbo (5′-AGAACCCTGGACCTGGTACCATGTACCCGTATGATGTTCCGG-3′); RV_Turbo_GGSGG (5′-GCCTCCAGATCCGCCCTTTTCGGCAGACCGCAGACTG-3′); RNEB_ClaI_N-OTX2 (5′-CGAGCTCTAGATCATCGATTTACAAAACCTGGAATTTCCACGAGGATGTCTG-3′); and FNEB_Link_N-OTX2 (5′-GGCGGATCTGGAGGCATGATGTCTTATCTTAAGCAACCGCCTTAC-3′).

For the TurboID transfection studies, 5 × 105 HDMB03 cells per well were seeded into 6-well ultra-low attachment plates and transfected using Lipofectamine 3000 (Thermo Fisher Scientific) with 12 μg plasmid, as previously described by Hendrikse et al.3. Tumourspheres were cultured for 72 h in StemPro Medium (Life Technologies) and then treated with 500 μM biotin-d (B4639; Sigma–Aldrich) for 15 min. Tumourspheres were then washed in 1× PBS and lysed with 1 ml 1× RIPA (20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA-Na2, 1 mM EGTA, 1% NP-40, 1% sodium deoxycholate, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na3VO4 and 1× Halt Protease and Phosphatase Inhibitor (Thermo Fisher Scientific)) for 10 min3. Samples were then sonicated over 2 × 10 s pulses at 30% load to shear DNA (an FB50 sonicator with microprobe; Thermo Fisher Scientific) and clarified by centrifugation at 16,000g and 4 °C for 10 min. Supernatants were quantified using the Bradford protein assay. Streptavidin–sepharose bead slurries (50 µl aliquots) were washed once in 1 ml 1× RIPA. Washed beads were resuspended in 1 ml 1× RIPA containing 2 mg protein supernatant and rotated at 4 °C overnight. The next day, samples were centrifuged for 5 min at 1,000g and bead pellets were washed twice in 1 ml 2% sodium dodecyl sulfate solution (in double-distilled water) and then three times (8 min each time) in 1 ml Wash Buffer (50 mM Tris-HCl (pH 7.4) and 8 M urea) at room temperature. Samples were resuspended in Storage Buffer (285 μl of ammonium bicarbonate (50 mM) and 15 μl of 1 mM biotin) and stored on ice.

An Orbitrap Exploris 480 instrument (Thermo Fisher Scientific) was used to obtain mass spectrometry data, as described by Hendrikse et al.3 Briefly, all mass spectrometry raw files were processed with Proteome Discoverer (version 2.20.388) at the Manitoba Centre for Proteomics and Systems Biology. The files were searched for tryptic peptides against the human UniProt protein database (December 2020) using SEQUEST with standard Orbitrap settings, as previously described3. In addition, up to two missed cleavages were permitted, with a parent and fragment mass tolerance of 0.02 Da and 15 ppm. A fixed modification of cysteine carbamidomethylation was applied. Also, variable modifications including deamidation (at Asn and Gln), amino-terminal acetylation, oxidation (at Met and Trp), phosphorylation (at Ser, Thr and Tyr), ubiquitylation (at Lys), double oxidation (at Met and Trp) and biotinylation (at Lys) were permitted. The results were filtered by 1% FDRs at both the peptide and protein levels. SAINTexpress (version 1.0.0)47,48 was used to calculate the probability of each potential proximal protein interaction from background contaminants using default parameters (n = 4 OTX2 biological replicates and n = 3 control biological replicates). A complete list of OTX2-interacting proteins is provided in Supplementary Table 1.

Co-IP

HDMB03 MB cells were grown as tumourspheres for 3 d, dissociated and lysed with immunoprecipitation lysis buffer (25 mM Tris (pH 7.4), 150 mM NaCl, 1% Triton X-100 and 5 mM EDTA) plus 1× protease inhibitor complex. Samples were then pre-cleared for 2 h and rotated at 4 °C. Immunoprecipitation was carried out overnight with 2 μg OTX2 antibody or immunoglobulin G. For co-IP on HDMB03 cells overexpressing wild-type OTX2 or OTX2 lacking its DNA-binding homeodomain, immunoprecipitation was carried out with a haemagglutinin antibody. Magnetic beads were blocked with 1% bovine serum albumin solution for 1 h, washed and then added to the protein/antibody complexes for 3 h with rotation at 4 °C. Beads were then washed five times with immunoprecipitation lysis buffer. Complexes were eluted by boiling for 5 min in 1× sodium dodecyl sulfate sample buffer. Interaction with LASR complex members (RBFOX2, ILF2, ILF3, MATR3, DDX5, HNRNPM, HNRNPC and HNRNPH) was then determined by immunoblotting. The antibody conditions are described in Supplementary Table 18.

Subcellular fractionation

Subcellular fractionation was performed using previously established protocols17,30. HDMB03 cells were lysed in cytoplasmic lysis buffer (CLB; 340 mM sucrose, 10 mM Tris-HCl (pH 7.9), 0.1 mM EDTA, 3 mM CaCl2, 1 mM dithiothreitol (DTT), 2 mM MgCl2, 0.5% NP-40 and protease and phosphatase inhibitors) on ice for 10 min and the cytoplasmic fraction was removed following centrifugation at 3,500g for 15 min. Cellular pellets were then washed with CLB wash buffer (CLB lacking NP-40) followed by centrifugation at 800g for 3 min. Soluble nuclear fractions were removed by lysing the cells for 5 min on ice with nuclear lysis buffer (10% glycerol, 3 mM EDTA, 20 mM HEPES (pH 7.9), 150 mM KOAc, 1.5 mM MgCl2, 1 mM DTT, 0.1% NP-40 and protease and phosphatase inhibitors) followed by centrifugation at 15,000g for 30 min. The RNA fraction was then isolated by lysing cells with nuclease incubation buffer (10% glycerol, 150 mM HEPES (pH 7.9), 1.5 mM MgCl2, 1 mM DTT, 150 mM KOAc and protease and phosphatase inhibitors) plus 10 mg ml−1 RNase, whereas the DNA fraction was isolated by lysing cells with nuclease incubation buffer plus DNAse I (1 U) for 30 min at 37 °C. RNA and DNA fractions were then recovered by centrifugation at 20,000g for 30 min. Samples were analysed by sodium dodecyl sulfate polyacrylamide gel electrophoresis with the antibodies described in Supplementary Table 18.

Intracerebellar transplantation and drug treatment

For the in vivo morpholino studies, HDMB03 and MB3W1 cells were treated with 1 or 2.5 μM control or PPHLN1-Mo for 5 d in culture. Cells were dissociated and 1 × 105 (HDMB03) or 2 × 105 (MB3W1) cells were injected into the cerebellums of 7- to 9-week-old male NOD-SCID mice. NOD-SCID mice were housed in individually ventilated cages (Tecniplast). Irradiated feed was used and bedding was sterilized by steam autoclave. The room ambient temperature was 21–23 °C with a relative humidity target of 50% (the range was 30–60%). Beginning at 06:00 am, the light cycle was 12 h on and 12 h off. The endpoint was defined as a 20% weight reduction from peak body weight and/or significant ataxia and ruffled fur. MRI was performed on a cryogen-free FlexiScan 7T system (MR Solutions). For tumour size calculations from MRI images, the ImageJ freehand tool was used to determine the volume from 18 serial sections (300 µm thickness) from each sample. The sum of all of the slices was then used to calculate an overall tumour volume, as previously described55. At the endpoint, tumours were prepared for immunohistochemistry using a mitochondria-specific antibody to detect human cells as well as SOX2 using methods previously described15. The primary and secondary antibody sources and concentrations are listed in Supplementary Table 18.

Digital spatial profiling on MB xenograft tumour samples

One Ctrl-Mo and one PPHLN1-Mo formalin-fixed, paraffin-embedded xenograft tumour sample were processed following the GeoMx DSP Slide Preparation User Manual (MAN-10087-04). Briefly, slides were baked at 60 °C for at least 1 h, deparaffinized using CitriSolv d-limonene and then rehydrated. Antigen retrieval was performed using 1× citrate buffer at a pH of 6.0 in a pressure cooker for 15 min at 100 °C on high pressure. Slides were blocked using NanoString blocking buffer W for 1 h. Slides were then incubated overnight at 4 °C with the following ultraviolet-photocleavable, barcode-conjugated antibody panels against a total of 43 targets and six control targets from NanoString Technologies: GeoMx Neural Cell Profiling Panel Human Protein Core for nCounter; GeoMx Cell Death Panel Human Protein Module for nCounter; GeoMx PI3K/AKT Signaling Panel Human Protein Module for nCounter; and GeoMx MAPK Signaling Panel Human Protein Module for nCounter. At the same time, the samples were incubated with morphology antibodies consisting of a 1:100 dilution of Ki67–Alexa Fluor 647 (12075S; Cell Signaling Technology) and a 1:500 dilution of MAP2–Alexa Fluor 532 (NBP1-92711AF532; Novus Biologicals). The slides were washed and stained with SYTO 13 (S7575; Thermo Fisher Scientific) for 15 min before loading onto a GeoMx DSP microscope (NanoString Technologies). Fluorescence images were scanned at 20× and regions of interest (ROIs) along the tumour border and in the tumour core were selected. Oligos from antibodies were cleaved and collected into 96-well plates. Oligos were dried down completely by incubating overnight at room temperature with a permeable plate seal and then rehydrated and hybridized with NanoString HybCode barcodes. Samples were hybridized overnight for at least 16 h in a thermal cycler at 67 °C with a heated lid at 72 °C. The resulting oligos hybridized to barcode tags were detected and counted using an nCounter Prep Station and Digital Analyzer (NanoString Technologies) the following day. The digital counts of each antibody for each ROI were generated for data analyses and analysed using NanoString GeoMx software. The software has built-in quality control analysis and three ROIs that did not pass quality control metrics due to high binding density were excluded from downstream analysis. The data were normalized by scaling to the negative control immunoglobulin G probes, constituting a signal-to-noise ratio. Counts were further scaled to the number of nuclei per ROI and the normalized counts from the Ctrl-Mo and PPHLN1-Mo samples were calculated.

Statistics and reproducibility

Prism 8.0 software (GraphPad) was used for all of the statistical analyses. No data were excluded from the analyses. NOD-SCID mice for the animal studies were randomly assigned into treatment groups; however, the remaining experiments were not randomized. Blinding was performed during MRI imaging and tissue preparation for immunohistochemistry. For all of the other experiments, the investigators were not blinded. No statistical methods were used to predetermine sample size, but our sample sizes are similar to those reported in previous publications15,55. Brown–Forsythe tests were performed to assess the homogeneity of variances for cell culture-based studies. The statistical tests used for all of the experiments are noted in the figure captions.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.