TMEM161B regulates cerebral cortical gyration, Sonic Hedgehog signaling, and ciliary structure in the developing central nervous system

Significance By evaluating children with cortical folding malformations, we identified TMEM161B, a gene with previously unknown function that is part of no known protein superfamily. In this work, we show that disrupting Tmem161b in utero is sufficient to lead to gyration abnormalities in a ferret model. We also characterized a Tmem161b null mouse that demonstrated developmental abnormalities including sequelae of abnormal Sonic Hedgehog signaling as well as alterations of primary ciliary structure. These connections help provide hypotheses for the cellular functions of Tmem161b, but also implicate Sonic Hedgehog signaling in promoting the normal folding of the human cortex.


TMEM161B Orthologs
AAs  (A) PLAC-seq data demonstrating interactions between local enhancers and the TMEM161B transcription start site. Enhancers sites described in Figure 2A show physical interaction with the TSS of TMEM161B in human radial glial or fetal neuron cells sequenced, confirming the enhancergene interaction in the developing CNS.
(B) Expression data of TMEM161B, TMEM161B-AS1 in bulk RNAseq of human brain. TMEM161B is more highly expressed in the developing frontal cortex than it is in the mature brain, data analyzed from Jaffe et al. 2015, Nature Neuroscience.
(C) Breeding strategy for generating Tmem161b KO animals. Tmem161b null mice were generated through excision of exon 4 of Tmem161b via use of EIIa-cre breeder against the Tmem161b tm2a/+ line described in STAR methods. This line also allows the reporting of Tmem161b transcriptional activity through b-galactosidase staining as demonstrated in Figure 2C. Null mice were confirmed via bulk-RNA sequencing as seen in Figure S3A and Figure S2D.
Expression of TMEM161B in human fetal brain Figure S3: ScRNAseq expression of TMEM161B in human fetal cortex.
UMAP plot of single cell RNA expression of TMEM161B in human fetal brain. 28 Single cell RNA sequencing of human fetal cortex at gestational week 17, TMEM161B is diffusely expressed across cell types. Dot plot of cell-type RNA expression of TMEM161B in human fetal brain summarizes expression across cell types. TMEM161B shows more expression in progenitor cell types relative to mature ones and is present in outer radial glial cells implicated in cortical folding such as HOPX+ oRGs. NPC = neural progenitor cell, oRG = outer radial glial cell,    IUE performed at E14.5, and analysis at P7, where final distribution of GFP+ cells was quantified across 10 laminar bins of the P7 cortex divided evenly from pial surface to bottom of cortex was described in Figure 4. Knockdown of Tmem161b at E14.5 led to an altered distribution of GFP+ cells at P7 (2-way ANOVA of plasmid condition X cortical region showed an interaction effect, p<0.001, with post Sidak multiple comparison tests, corrected for multiple testing). Co-staining these same slides with Satb2 and Ctip2 as layer markers revealed a negligible amount of GFP+/Ctip2+ costained cells, but a depletion of GFP+/Satb2+ double marked cells in the Tmem161b knock-down condition relative to controls (t-test, n=3 control, n=4 knockdown, p=0.0154). Additionally, there were fewer overall GFP+ cells in the knockdown condition relative to the control condition, suggesting that disruption of Tmem161b led to fewer overall progeny (t-test, p=0.0004). (B) Example of Image Quantification of IF of Arl13b+ puncta at ventricular zone of E14.5 mouse embryos. Although Tmem161b WT and KO mice show the same number of Arl13b+ puncta on IF, these puncta are different in size/shape analysis of the images under super-resolution confocal microscopy demonstrate that Tmem161b KO cilia show increased circularity (suggesting shorter length), and a decreased major/minor axis ratio suggesting the same. These differences are better appreciated on electron microscopy as above.

Supplementary Video 1 Related to Figure 5E
XZ Plane video of inset region showed in Figure 5E. Green signal corresponds to Tmem161b-citrine, and red signal to Acetylated tubulin. Note the green signal as puncta at multiple ciliary tips as well as ciliary bases.

Supplementary Video 2 Related to Figure 5F
Video of z-stack from maximum intensity projection displayed in Figure 5F of ciliated IMCD3 cells with transfected humanized TMEM161B-GFP, co-stained for Acetylated tubulin. Green signal corresponds to Tmem161B-GFP fusion protein, and red signal to Acetylated tubulin. Note accumulation of GFP+ signal at ciliary tips in this field.

Clinical Case Summaries of Individuals with biallelic TMEM161B variants Family A Individual: 09DG00538
09DG00538 was born to a consanguineous Caucasian couple at term by spontaneous vaginal delivery following an uneventful pregnancy. Birth weight was 3.5 kg (0.2 SD), she was noted to have congenital microcephaly, and seizures commenced within 1 hour after birth necessitating neonatal intensive care. Brain MRI revealed extensive bilateral polymicrogyria and ventriculomegaly, with some heterotopic subependymal grey matter nodules. Her epilepsy was never controlled despite treatment with many antiepileptic drugs. Examination at 3.5 years old showed axial hypotonia with spastic quadriplegia and, apart from severe microcephaly and related craniofacial changes, there was no gross dysmorphism. The family history is notable for a brother with microcephaly who died at 4.

Supplementary Discussion
Potential cellular roles for TMEM161B The ciliary phenotype noted on SEM was dramatic but apparently not catastrophic in its effect; the phenotype of the Tmem161b KO mouse is not as severe as those that obliterate function of the primary cilium entirely 1,2 , and only a proportion of cilia observed at the moment of fixation exhibited ballooning in our KO embryos, estimated at 10-80% depending on the field observed.
Thus, one hypothesis of how pathogenic variants in TMEM161B lead to the ciliary abnormalities is that TMEM161B may play a role in the proteostasis of transmembrane receptors that are normally regulated through exquisitely precise ciliary trafficking. The SEM of abnormal primary cilia in Tmem161b KO mice is reminiscent of, though more dramatic than, abnormal cilia found in mutants of ciliary trafficking proteins, such as Ift140 3 , Ift144 4 , Dync2h1 4 , or the Golgi protein Golgb1 5 , all phenotypes of which are due to material accumulating in the ciliary tip.
Another possibility is that Tmem161b may be involved in the machinery that regulates actindependent ciliary vesicle exocytosis, defects in which can lead to abnormal Shh signaling, while not impairing an initial Smo translocation into the primary cilium. 6 Receptors and other transmembrane proteins that do not undergo retrograde transport in the cilium can be released through ciliary exocytosis, and poisoning actin remodeling in ciliated cells responding to ciliary GPCR ligands with cytochalasin D causes accumulation of receptors in the ciliary tip with consequent ballooning. 6 Early ciliary disassembly of the ciliary tip is based on similar mechanisms 7 , so the ciliary abnormalities seen in our Tmem161b null embryos could be related to a failure of scission of either ciliary exocytic vesicle release or ciliary decapitation. These hypotheses (trafficking defect, exocytic vesicle release defect, and ciliary tip release defect) are not mutually exclusive, and future studies using live cell imaging might help describe the changes in ciliary dynamics in TMEM161B null cells and help illuminate the specific cellular function of TMEM161B.

Airyscan Super-Resolution Confocal Microscopy
Coronal cryosections of E14.5 and E18.5 mouse cortex were immunostained for primary cilia marker Arl13b and imaged using a Zeiss LSM 980 confocal microscope with Airyscan 2. Imagining was at the apical surface of the dorsolateral corner of the lateral ventricles to maintain consistency of the anatomical locations. The Z-stack function was used to image a 5 µm thickness, and automatic Airyscan processing was performed after acquisition. All images were captured at the same imaging settings of laser intensity and detector gain. Sample images were prepared in ImageJ and Zen software.
Primary Cilia Quantification Image analysis was performed automatically using ImageJ software. Maximal intensity projection and automated thresholding was performed on the image. A mask of all the cilia was generated on the thresholded image using the Analyze Particles function in ImageJ. To exclude noise pixels and overlapping cilia, a size range of 0.05 3 µm2 was selected. ImageJ automatically determines an ellipse fit for each cilium and the output contains the area and the perimeter of the cilium, and the major axis length and mi calculated by the ratio between the length of the major and minor axis of the fitted ellipse, a larger value indicates more elongated cilium.

Analysis of scRNA-seq data
To facilitate cross-species comparison of TMEM161b expression, an integrated fetal brain scRNAseq atlas was created from the following publicly available datasets: The dataset from Polioudakis et al., showed strong clustering by tissue donor, and, as recommended by Seurat developers, was first integrated across donors using the Seurat integration pipeline. This integrated dataset as well as all other datasets above was then processed individually using a consistent quality-control pipeline, including filtering of features expressed in <3 cells, cells that express an unusually high or low number of features, and cells with >5% mitochondrial transcripts. scTransform pipeline was used for normalization, variance stabilization and regression of confounding sources of variation, such as batch, percent mitochondrial expression, difference between cell-cycle phases, and percent ribosomal expression.
After processing each standard integration pipeline to create the final integrated fetal brain scRNA-seq atlas. Dimensionality reduction was performed to visualize the atlas, and visualization of covariates such as species and dataset of origin showed that clustering was not being driven by origin effects.
Unsupervised clustering was performed using the standard Seurat pipeline, and cell types were annotated based on a comprehensive, multi-person approach utilizing both marker gene expression Human Genetics Sequencing DNA extracted from peripheral blood of enrolled subjects was analyzed by exome sequencing (ES). ES and data processing for Family A was performed by the Genomics Platform at the Broad Institute of Harvard and MIT (Cambridge, MA, USA). ES was performed on DNA samples (>250 ng of DNA, at >2 ng/ l) using Illumina exome capture (38 Mb target). The ES pipeline included sample plating, library preparation (2-plexing of samples per hybridization), hybrid capture, sequencing (150 bp paired reads), sample identification quality control check, and data storage. The hybrid selection libraries cover >90% of targets at 20x and a mean target coverage of ~100x. The ES data was de-ES data were processed through a pipeline based on Picard, using base quality score recalibration and local realignment at known indels. The BWA aligner was used for mapping reads to the human genome build 37 (hg19). Single nucleotide polymorphism (SNPs) and insertions/deletions (indels) were jointly called across all samples using Genome Analysis Toolkit (GATK) HaplotypeCaller package version 3.4. Default filters were applied to SNP and indel calls using the GATK Variant Quality Score Recalibration (VQSR) approach. Lastly, the variants were annotated using Variant Effect Predictor (VEP). For additional information please refer to Supplementary Section 1 of the paper describing ExAC. 8 The variant call set was uploaded on to Seqr and analysis was performed to identify candidate variants according to ACMG guidelines. ES was performed for Family B and Family C via similar methods.
Candidate variants identified via ES were evaluated by manual review for their rarity, predicted effects, and review of literature. For each family, following the emergence of TMEM161B as the most likely candidate gene, confirmation of the variants by Sanger sequencing was performed on original subject samples in each family, including first degree relatives, when available, to confirm segregation of TMEM161B alleles with phenotype. In silico prediction of the effect of missense variants was performed using Combined Annotation Dependent Depletion (CADD; https://cadd.gs.washington.edu), MutationTaster (www.mutationtaster.org), SIFT (https://sift.bii.astar.edu.sg), and PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2). The splice region variant was characterized using Alamut® Visual Plus v1.3 (Sophia Genetics) by MaxEntScan, NNSPLICE, and SpliceSiteFinder-like.
Molecular Cloning/shRNA Validation shRNA plasmids were ordered from VectorBuilder, using their shRNA plasmid generation tool on a pRP background, to include shRNA sequences specific to mouse Tmem161b. The efficiency of these plasmids was evaluated in two different mouse cell lines before use in experiments. SL2 cells were grown as adherent cultures to 60% confluency. Scramble shRNA, shRNA Tmem161b-M2 (used in in utero and SL2 assays), and shRNA Tmem161b-M3 (used in SL2 assays) plasmids were transfected onto the N2A or SL2 cultures with Lipofectamine 3000 reagent (Thermo Fisher) according to manufacturer protocol. The cells were checked under fluorescence for GFP reporter expression to estimate transfection efficiency at 48 hours. The transfection efficiency was approximately 50-60% when the cells were harvested after 96 hours of incubation, using a Qiagen RNEasy RNA extraction kit. mRNA was isolated from the RNA extracts through SuperScript VILO cDNA synthesis using poly-A priming. qPCR was performed on diluted cDNA using PowerUP SYBR Green Master Mix with 5 technical replicates in qPCR to estimate the amount of TMEM161B transcript in the Scramble vs. TMEM161B shRNA conditions ( Figure S4B).
Ferret Histology/Immunofluorescence Ferret kits (postnatal day 21, P21) were deeply anesthetized with ketamine and xylazine by intraperitoneal injection, and transcardially perfused with PBS followed by 4% paraformaldehyde (PFA) in PBS. Dissected brains were post-fixed for 48 hr at 4°C, and then cryoprotected by immersion in a series of sucrose solutions (from 10% to 20% and 30% sucrose solutions in PBS, changed every 24 hr). Ferret brains were cut frozen on a sliding microtome (Leica SM2010 R) at 50 -037-246). Imaging of ferret brains was performed in a LSM980 confocal microscope at 8-bith depth with 5X/0.16 (Magnification/Numerical Aperture) objective and 1.0 digital zoom.
Tile scan images of coronal sections were acquired at 1024 x 1024 pixel resolution at 0.25 s/pixel acquisition speed. Imaging of neuronal density and laminar position was performed at 2780 x 2780 pixel resolution at 0.68 s/pixel acquisition speed, with 2 average per frame. Histological analyses were conducted at P21, an age in which neurogenesis and neuronal migration are largely completed in both somatosensory and motor regions (Smart and McSherry, 1986, Journal of Anatomy). For analysis of gross anatomical features, >10 coronal sections were examined for each ferret brain sample. Analysis of dorsal cortical surface was restricted to the somatosensory and motor cortices since the electroporated cells were predominantly found in these regions. Local gyrification was calculated as previously described (Matsumoto et al., 2020, eLife). Briefly, we calculated the ratio of the size of the gyri in the electroporated hemisphere compared to the size of the corresponding gyri in the contralateral, non-electroporated hemisphere; quantification of sulcal depth was also estimated as a ratio relative to the corresponding sulcus in the contralateral, non-electroporated hemisphere (Fig. S5). We focused our analyses on the posterior sigmoid gyrus, the coronal gyrus, the coronal sulcus, and the suprasylvian sulcus, where the somatosensory and motor cortices lie, since electroporated cells were predominantly found in this region. Analysis of Satb2+/mCherry+ neurons across layers was restricted to the primary somatosensory cortex to consistently examine laminar organization across brain sections and reduce the variability inherent to regional differences.

Mouse IF
Mouse embryos were harvested at ages between E11.5-E18.5 and were dissected in cold PBS to isolate tissues of interest (brains, or thorax for spinal cords). This tissue was drop fixed in 4% paraformaldehyde overnight (14-18 hours), and then rinsed in PBS several times before being placed overnight in 30% sucrose solution. Tissues were frozen in OCT and cryo-sectioned at 12-18 m. Slides were rinsed with PBS X5, and then blocked in a solution with 8% normal donkey serum + 0.3% BSA and 0.3% Triton X-100 for 45 minutes at room temperature before being treated with primary antibodies (see below) overnight at 4°C diluted in the blocking solution. Slides were rinsed with PBS X5 and then treated with secondary antibodies conjugated to Alexa fluorophores + DAPI diluted in blocking solution for 2 hours at room temperature before being rinsed, dried, and cover slipped with Prolong Diamond mounting medium. Slides were imaged on a (Zeiss LSM 550) confocal microscope.
Protein Modeling/Evolutionary Analyses Remote homologs of human TMEM161b (https://www.uniprot.org/uniprot/Q8NDZ6) were harvested from sequence databases by iterative PSIBLAST 10 searches, and also aligned by the EVcouplings server (https://evcouplings.org). 11 More sensitive, proteome-level searches for distant TMEM161 family relatives were conducted by Hhpred at the MPI Bioinformatics toolkit (https://toolkit.tuebingen.mpg.de) 12 that also calculated PSIPRED secondary structure profiles 13 for the superfamily. The protein chains of TMEM161 family members typically displayed 9 transmembrane (TM) segments by Phobius analysis, 14 with the N-terminal TM1 showing a weak or equivocal signal peptide nature by SignalP5. 15 The threedimensional structure of TMEM161b was first predicted by the top-ranking CASP14 servers trRosetta 16 and tFold, 17 convergently revealing a uniquely complex fold that did not bear any similarity to other membrane protein structures in the PDB 18 by DALI analysis. 19 A possible functional site in the TMEM161b fold was suggested by ConSurf analysis 20 that mapped an evolutionarily conserved pocket on the luminal face of the fold. The 3D structural models of TMEM161b and a phylogenetically broad range of homologs were greatly improved by ColabFold access (https://github.com/sokrypton/ColabFold) 21 to new AlphaFold2 22 and RosettaFold 23 deep learning-based programs, that employ end-to-end neural network algorithms with near-crystallographic accuracy. 24 This phylogenetically diverse ensemble of TMEM161b-related structures was used to locate conserved epitopes and residues in the unique fold, to help interpret mutations in human TMEM161b linked to disease and drive functional analysis. Structures manipulated and viewed with PyMOL (www.pymol.org).
Protein Modeling/Evolutionary Analyses Remote homologs of human TMEM161b (https://www.uniprot.org/uniprot/Q8NDZ6) were harvested from sequence databases by iterative PSIBLAST 10 searches, and also aligned by the EVcouplings server (https://evcouplings.org). 11 More sensitive, proteome-level searches for distant TMEM161 family relatives were conducted by Hhpred at the MPI Bioinformatics toolkit (https://toolkit.tuebingen.mpg.de) 12 that also calculated PSIPRED secondary structure profiles 13 for the superfamily. The protein chains of TMEM161 family members typically displayed 9 transmembrane (TM) segments by Phobius analysis, 14 with the N-terminal TM1 showing a weak or equivocal signal peptide nature by SignalP5. 15 The threedimensional structure of TMEM161b was first predicted by the top-ranking CASP14 servers trRosetta 16 and tFold, 17 convergently revealing a uniquely complex fold that did not bear any similarity to other membrane protein structures in the PDB 18 by DALI analysis. 19 A possible functional site in the TMEM161b fold was suggested by ConSurf analysis 20 that mapped an evolutionarily conserved pocket on the luminal face of the fold. The 3D structural models of TMEM161b and a phylogenetically broad range of homologs were greatly improved by ColabFold access (https://github.com/sokrypton/ColabFold) 21 to new AlphaFold2 22 and RosettaFold 23 deep learning-based programs, that employ end-to-end neural network algorithms with near-crystallographic accuracy. 24 This phylogenetically diverse ensemble of TMEM161b-related structures was used to locate conserved epitopes and residues in the unique fold, to help interpret mutations in human TMEM161b linked to disease and drive functional analysis. Structures manipulated and viewed with PyMOL (www.pymol.org).

Mouse Generation
The C57BL/6N-A tm1Brd -Tmem161b tm2a(EUCOMM)Hmgu /BayH mice (referred to as Tmem161b tm2a ) were originally generated by Baylor College of Medicine as part of their commitment to the International Mouse Phenotyping Consortium (IMPC), RRID: MMRRC_041541-UCD. These mice were bred against a pan-cre expressing line (e2a-Cre) to generate the Tmem161b tm2b(EUCOMM)Hmgu allele line (referred to as Tmem161b tm2b ). Initial phenotyping was conducted at The Mary Lyon Centre, MRC-Harwell. Some primary phenotyping data may be found at www.mousephenotype.org. LacZ Staining E12.5 For LacZ staining all solutions were at pH 8.0. Embryos were dissected into ice-cold phosphate buffered saline (PBS) and exsanguinated by severing the umbilical vessels. After washing in PBS, embryos were fixed in ice-cold 4% paraformaldehyde (PFA) prepared in PBS for 20 minutes. For whole-mount staining, embryos passed through three, 30-minute PBS washes before being immersed in X-gal stain (MgCl2 2mM, IGEPAL CA-630 0.02%, potassium ferrocyanide 5mM, potassium ferricyanide 5mM, Sodium deoxycholate 0.01%, X-gal 1mg/mL, in PBS) and incubated at 4°C for 24 hours. Following staining embryos were rinsed and passed through two, 1-hour PBS washes, before post-fixing in 4% PFA overnight. For sections, after initial 20-minute fix, embryos were washed in PBS for 40 minutes, then transferred to 30% sucrose in PBS for cryoprotection overnight, frozen in OCT and stored at -80°C. 30µm sections were air dried, fixed in 4% PFA for 10 mins, passed through three, 5-minute PBS washes, before incubation in X-gal stain (as above, but with the addition of 12µg/mL 4-Nitro blue tetrazolium chloride) at 37°C for 48 hours.