Oligodendrocyte precursor cells sculpt the visual system by regulating axonal remodeling

Many oligodendrocyte precursor cells (OPCs) do not differentiate to form myelin, suggesting additional roles of this cell population. The zebrafish optic tectum contains OPCs in regions devoid of myelin. Elimination of these OPCs impaired precise control of retinal ganglion cell axon arbor size during formation and maturation of retinotectal connectivity and degraded functional processing of visual stimuli. Therefore, OPCs fine-tune neural circuits independently of their canonical role to make myelin.

eage cells differentiated by 14 d.p.f., in strong contrast to hindbrain regions showing 56% differentiation (Extended Data Fig. 1d,e). Within the tectum, OPCs localized their soma either at the border between the neuropil and periventricular zone containing most tectal neurons or right within the neuropil (Fig. 1d-f and Extended Data Fig. 1f) and claimed non-overlapping territories, similarly to previous studies (Extended Data Fig. 1g and Supplementary Video 3) 10,14,15 . OPCs can be highly dynamic and potentially migrate, proliferate or differentiate. However, sparse labeling revealed that soma positioning of individual tectal OPCs remained largely stable, with a low rate of division and differentiation between 6 d.p.f. and 10 d.p.f. (Fig. 1f,g).
The appearance of OPCs coincided with the time when RGC axons arrive in the developing tectum and establish their terminal arborizations (Extended Data Fig. 2a). Time-lapse imaging revealed that RGC arbors dynamically interact with OPC processes and that interactions frequently preceded retractions of arbor tips (Fig. 2a, Extended Data Fig. 2b,c and Supplementary Videos 4-7). This correlation prompted us to ask whether OPCs influence the organization of RGC arbors. We tested this using an inducible nitroreductase (NTR)-mediated cell ablation system specifically targeted to OPCs (Tg(olig1:CFP-NTR)) (Fig. 2b,c and Extended Data Fig. 3a). Early OPC ablation from 2 d.p.f. when RGC axons arrive at the tectum led to formation of erroneous axon branches reaching outside the tectal neuropil as well as enlarged sizes of individual RGC arbors (Fig.  2e,f,h,i). To exclude that this phenotype was mediated indirectly by microglia clearing dying OPCs or by diverting microglial activities, which have an established role at eliminating synapses 16 , we carried out two controls: (1) genetic depletion of OPCs without causing inflammation by morpholino injection against oligodendrocyte transcription factor 2 (olig2) and (2) depletion of microglia using a morpholino against interferon regulatory factor 8 (irf8) (Fig. 2d and Extended Data Fig. 3b-d). Although olig2 morphants also exhibited ectopic branching and enlarged RGC arbors, none of these phenotypes was seen in irf8 morphants (Fig. 2g,j and Extended Data Fig.  3e,f). Therefore, erroneous RGC arborizations and enlarged RGC arbors resulted directly from the absence of OPCs.
After their formation, RGC arbors undergo a phase of developmental pruning during larval stages when retinotectal connectivity is refined 17,18 . To test whether OPCs also play a long-lasting role during the refinement of RGC arbors as they continue to persist and interact with each other, we carried out late OPC ablations starting from 7 d.p.f. when zebrafish have a functional visual system. Ablations were carried out analogously using NTR-mediated chemogenetics or by two-photon-mediated cell ablation to specifically eliminate OPCs from the tectum ( Fig. 3 and Extended Data Fig. 4). In control animals, individual RGC arbors underwent process remod-

NaTurE NEurOSCiENCE
eling with additions and eliminations of multiple neurites that lead to a net reduction in arbor size by about 14% between 7 d.p.f. and 10 d.p.f., similarly to previous reports ( Fig. 3c and Extended Data Figs. 4g and 5a-d) 18 . This reduction in arbor size was significantly decreased in OPC-ablated animals (1.7% after OPC laser ablation and 6% after OPC NTR ablation), with some arbors even increasing in size due to a reduction in neurite eliminations and an increase in neurite additions after laser-mediated OPC ablation ( Fig. 3c and Extended Data Figs. 4g and 5a-d). Despite changes in size, individual arbors remained stratified (Extended Data Figs. 4g and 5b,c). Furthermore, the effects on neurite remodeling were specific to axonal processes because remodeling of tectal neuron dendrites in the same tissue was unaffected upon OPC ablation, further corroborating that the effects observed did not result from unspecific collateral damage induced by our manipulations (Extended Data Fig. 5e-j).
It has been reported that enlarged RGC arbors impair visual processing 19 . To test if OPC ablation affects functional performance as the visual system matures, we first carried out prey capture assays in late ablations from 7 d.p.f. onwards. These experiments showed that about two times more paramecia remained uncaptured in OPC NTR-ablated animals, and about 70% more paramecia remained uncaptured after unilateral laser ablation of tectal OPCs (Fig.  3d,e and Extended Data Fig. 6a,b). This effect was specific to tectal OPCs as OPC laser ablation from telencephalic regions did not impair paramecium capture ( Fig. 3d and Extended Data Fig. 6a). None of our manipulations affected overall swimming activity, ruling out that gross locomotor defects account for reduced hunting (Extended Data Fig. 6c,d). Furthermore, we performed optomotor response (OMR) assays stimulated by moving gratings of different widths (Extended Data Fig. 6f,g and Supplementary Video 8).
Narrower gratings become increasingly difficult to resolve, leading to longer latencies and ultimate failure to elicit OMR (Fig. 3f). After global NTR-mediated OPC ablation, animals were able to robustly elicit OMR in response to wide gratings (10-mm width), but there was a significantly increased probability of failure to initiate OMR when narrow gratings (3.3 mm) were presented (Fig. 3g,h and Extended Data Fig. 6h-k). It is known that OMR does not primarily require the tectum but, rather, pre-tectal areas to where RGC axons can extend collaterals in addition to the tectum 20,21 and in which OPCs also reside (Extended Data Fig. 6e). Therefore, to test if OPCs within the tectum are of direct importance for functional sensory integration, we carried out in vivo calcium imaging of tectal neurons using Tg(elavl3:h2b-GCaMP6s) in response to light flashes in random positions around the visual field (Fig. 3i,j). Visual responses in OPC NTR-ablated animals were less reliable (Fig. 3k) and smaller in amplitude (Fig. 3l), whereas receptive field size and overall retinotopy along the tectum remained intact (Extended Data Fig. 7). Together, these data show that OPC ablation impairs visual processing.
In summary, our data reveal a physiological role for OPCs in fine-tuning the structure and function of neural circuits that is independent of their traditional role in myelin formation and which is mediated by regulating growth and remodeling of axon arbors. It remains an open question if regulation of arbor growth and remodeling by OPCs at different developmental stages might be mediated by different mechanisms, as OPCs constantly express growth-inhibitory molecules such as chondroitin sulfate proteoglycans 22 , but have also been reported to phagocytose axons 23 , similarly to how microglia can prune axons 24

NaTurE NEurOSCiENCE
changes in axon arbors translate into impaired neuronal connectivity underlying visual processing? Future work will reveal these mechanisms.

DNA microinjection for sparse labeling and generation of transgenic lines.
Fertilized eggs at one-cell stage were microinjected with 1 nl of a solution containing 5-20 ng μl −1 of DNA plasmid and 20 ng μl −1 of Tol2 transposase mRNA. Injected F 0 animals were either used for single-cell analysis or raised to adulthood to generate full transgenic lines. For this, adult F 0 animals were outcrossed with wild-type zebrafish, and F 1 offspring were screened for presence of the reporter transgene under a fluorescence stereo dissecting microscope (Nikon SMZ18). 'Tg(promoter:reporter)' denotes a stable transgenic line, whereas 'promoter:reporter' alone indicates that a respective plasmid DNA was injected for sparse labeling of individual cells. OPC ablation using two-photon lasers. OPCs were laser ablated from Tg(olig1:nls-mApple) using an Olympus FV1000/MPE equipped with a MaiTai DeepSee HP (Newport/Spectra Physics) and a ×25 1.05 NA MP (XLPLN25XWMP) water immersion objective. Continuous confocal scans using a 559-nm laser were taken to locate individual OPC nuclei in the optic tectum, which was identified by additional transgenic genetic markers (Tg(olig1:memYFP) or Tg(isl2b:EGFP)). Each cell was ablated using a 500-ms line scan across the cell nucleus using the MaiTai laser tuned to 770 nm (1.75-W output). The wavelength was incrementally increased for ablating OPCs in deeper tissue. After successful ablation, previously bright, round nuclei appeared dim, irregular or fragmented. The ablation procedure was repeated when cells did not show this signature. Unilateral OPC ablations took 60-90 min in the tectum. For analysis of axon remodeling after OPC laser ablation, surviving and/ or repopulating OPCs were ablated again on the second day.
Morpholino-mediated depletion of microglia and OPCs. Microglia were depleted from zebrafish embryos by microinjection of 4.5 pg of a previously published morpholino targeting the start codon of irf8 (5′-TCAGTCTGCGACCGCCCGAGTTCAT-3′) 34 . OPCs were depleted by microinjection of 7.5 pg of a previously published morpholino targeting the start codon of olig2 (5′-ACACTCGGCTCGTGTCAGAGTCCAT-3′) 35 . Both morpholinos were synthesized by Gene Tools.
Neutral red staining. Zebrafish embryos were incubated for 2.5 h in the dark in 2.5 µg ml −1 of neutral red solution (Sigma-Aldrich, n2889) diluted in Danieau's solution. Afterwards, embryos were washed three times for 10 min with Danieau's solution. Bright-field images of the head of the fish were taken using a Leica DFC300 FX Digital Color Camera.
Prey capture assay. Next, 2 ml of 0.3× Danieau's solution with 30 Paramecium multimicronucleatum were added to a 35-mm dish, along with a single zebrafish larva. The number of remaining paramecia was determined at hourly intervals for 2 h. To rule out batch-dependent effects resulting from 'natural' paramecia death followed by their disintegration, a control containing paramecia but no fish was run alongside each experiment. Spontaneous paramecia death occurred only rarely in 0-3%.
Locomotor activity assay. For all experiments, testing occurred between 9:00 and 17:00 using a randomized trial design to eliminate systematic effects due to the time of day. A tracking chamber was prepared by using a 35-mm Petri dish mold surrounded by 1% agarose situated in the center of a 85-mm Petri plate to eliminate mirroring that occurs at the wall of a plastic Petri dish. Single zebrafish were placed into the well filled with 2 ml of 0.3× Danieau's solution. The plate was positioned above an LED light stage to maximize contrast for facilitating zebrafish tracking (two dark eyes and swim bladder of zebrafish larva on a light background), with a high-speed camera (XIMEA MQ013MG-ON) equipped with a Kowa LM35JC10M objective positioned above the dish. The larvae acclimated to the recording arena for 5 min before the start of video acquisition. The center of mass of two eyes and swim bladder was taken as the center of the fish. Subsequently, video of spontaneous free swimming was recorded for 10 min at 100 Hz using a custom-written Python script and the Stytra package 36 .
Optomotor response assay. Zebrafish larvae were embedded in 1% agarose in a 35-mm Petri dish. After allowing the agarose to set, the dish was filled with 0.3× Danieau's solution, and the agarose around the tail was removed with a scalpel, leaving the tail of fish free to move (hereby referred to as a head-restrained preparation) 37 . Visual stimuli were presented on the screen from below using an ASUS P3E micro projector and an infrared light (Osram 850-nm high-power LED). The fish's tail was tracked using a high-speed camera (XIMEA MQ013MG-ON) and a 50-mm telecentric objective (Navitar TC-5028). A square-wave grating with variable spatial period and maximal contrast was achieved by the projector (black and white bars), and online tail tracking and stimulus control was performed using Stytra software 0.8.26 (ref. 36 ). Experiments were performed in closed loop, meaning that the behavior of fish was fed back to the visual stimulus to provide the fish with visual feedback. Therefore, when the fish swam, the grating accelerated backward at a rate proportional to swim powerthat is, [stimulus velocity] = 10 -[gain] × [swim power]. Swim power was defined as the standard deviation of the tail oscillation in a rolling window of 50 ms. To obtain a feedback that mimics the visual feedback that the animal would receive when freely swimming, the gain multiplication factor was chosen to result in an average fictive velocity of about 25 mm s −1 during the bout. When the fish was not swimming, [swim power] = 0, the grating moved in a caudal to rostral (forward) direction at a baseline speed of 10 mm s −1 . The stimulus scene was a square window that was centered on the head of fish and spanned a field of total 60 × 60 mm. For analysis, individual bouts were counted as episodes where the swim power was above 0.1 radian for at least 100 ms. Then, latency to first bout and total number of bouts were quantified for each trial (latency was set as a default value equal to the stimulus duration, when fish did not respond). Analysis was performed with custom scripts written in Python.
Zebrafish mounting for live-cell microscopy. Zebrafish larvae were anaesthetized with 0.2 mg ml −1 of 3-aminobenzoic acid ethyl ester (MS-222). For confocal microscopy, animals were mounted ventral side up in 1% ultra-pure low-melting-point agarose (Invitrogen) onto a glass-bottom 3-cm Petri dish (MatTek). For two-photon microscopy, embryos were mounted ventral side up in low-melting-point agarose on a glass coverslip. The coverslip was then flipped over on a glass slide with a ring of high-vacuum grease filled with a drop of 0.2 mg ml −1 of MS-222 to prevent drying out of the agarose. After imaging, the animals were either euthanized or released from the agarose using microsurgery forceps and kept individually until further use.
Immunohistochemistry. Samples were fixed overnight at 4 °C in a solution of 4% paraformaldehyde in PBS solution containing 1% Tween 20. After fixation, the samples were washed in the same solution without fixative and blocked for 1.5 h at room temperature in PBS buffer, 0.1% Tween 20, 10% FCS, 0.1% BSA and 3% normal goat serum. Primary antibody incubation was conducted at 4 °C overnight in blocking solution. Afterwards, samples were washed three times in PBS with 0.1% Tween 20 and then incubated with Alexa Fluor 633-conjugated secondary antibody. Stained samples were washed three times in PBS with 0.1% Tween 20 and subsequently mounted with ProLong Diamond Antifade Mountant (Thermo Fisher Scientific). Primary antibody rabbit anti-HuC/D (Abcam, ab210554) was used at a dilution of 1:100. Goat anti-rabbit secondary antibody (Thermo Fisher Scientific) was used at a dilution of 1:1,000. Images were obtained using a confocal microscope (Leica TCS SP8).

NaTurE NEurOSCiENCE
Confocal microscopy. Images of embedded zebrafish were taken with a Leica TCS SP8 confocal laser scanning microscope (LASX 3.5.2.18963). We used 458 nm for excitation of Cerulean and tagCFP; 458 and 488 nm for EGFP; 514 nm for EYFP; 561 nm for mApple and mScarlet; and 633 nm for AF633 and BODIPY630/650. For overview images and analysis of cell numbers based on nuclear transgenes, we used a ×10 / 0.4 NA objective (acquisition with 568-nm pixel size (x-y) and 2-μm z-spacing). For all other analyses, we acquired 8-bit or 12-bit confocal images using a ×25 / 0.95 NA water objective with 114-151-nm pixel size (x-y) and 1-μm or 1.5-μm z-spacing.

Analysis of contact-mediated retraction between RGC axons and OPC processes.
For analyzing dynamic interactions between RGC axon arbors and OPC processes, images were taken over 2 h within 2-min intervals and 1-μm z-spacing. Three-dimensional (3D) movies were subsequently generated for analyzing RGC retraction using Imaris software. Contact-mediated repulsion was classified as every event in which the tip of an extending RGC process directly opposed or apposed an OPC process, followed by RGC process retraction to resolve this apposition within the following seven frames. RGC processes, which changed from extension to retraction without such prior contact to OPC processes, were categorized as contact-independent retraction.
Analysis of axonal and dendritic arbor remodeling. Only RGC axons that arborized within the tectal neuropil and that could be traced back to the optic nerve were included for analysis. The neurites of periventricular interneurons (PVINs) extending to the superficial layers were randomly selected for analysis, as the dendrites of PVINs located in the superficial layers and their axons are located in deeper layers 38 . Individual axonal and dendritic arbors were analyzed using the segmentation tool of 3D tracing in the simple neurite tracer plugin in Fiji/ImageJ 39 . Each arbor was traced from its first branch point out to all branch tips, and each branch segment was counted from the branch point to the next branch point or branch tips. From this tracing, we extracted the measurement of total branch length and the total number of branch segments. The eliminated/added branch segment was obtained by comparing the tracing of the same neuron at two time points.
Lightsheet imaging for functional calcium imaging of tectal neurons. For lightsheet imaging, MTZ-treated Tg(olig1:CFP-NTR), Tg(elavl3:h2b-GCaMP6s) with late OPC ablation and MTZ-treated control Tg(elavl3:h2b-GCaMP6s) fish were embedded at the center of a custom-built plastic chamber using 2-2.5% low-melting-point agarose. The chamber was then placed onto the stage of a custom-built lightsheet microscope, previously described in ref. 40 . In brief, light from a 473-nm laser (Cobolt) was scanned with galvanometric mirrors (Sigmann Electronik) horizontally to create a ~5-µm-thick excitation sheet. The sheet was then moved vertically together with the light collection objective, controlled with a piezo controller (Piezosystem Jena). The eyes were protected from the incoming light by two plastic screens positioned at the conjugate plane of the scanning laser focus. Two orthogonal sheets were generated-one impinged on the brain from the side, the other one from the front-to ensure extensive coverage of the whole brain without hitting the eyes. The image was filtered with a band-pass 525/50 filter and acquired using an ORCA-Flash version 4.0 camera (Hamamatsu Photonics). The microscope was controlled using the Sashimi package (https://zenodo.org/ record/4122062#.YQl7yC0RpQI). We acquired volumes of ~130 × 400 × 340 µm (dorso-ventral, left-right and anterior-posterior axes, respectively) at a resolution of 15 × 0.6 ×0.6 µm and a frame rate of 3 Hz.
Visual stimuli were projected on a white plastic screen placed below the fish. The protocol consisted of a sequence of dark flashes on a red background, spanning 10° circular sectors of the area around the fish for a total of 36 different locations. Flashes were shown for 4 s, with a pause of 2 s of flat red background between them, and presented ten times each in a sequence randomized differently for every fish. The script for generating the experimental protocol in Stytra is available in the code repository. Synchronization between the imaging and the stimulus presentation was achieved by ZMQ-mediated triggering between Stytra and Sashimi.
Imaging data pre-processing for calcium imaging. Raw stacks from the lightsheet imaging were inspected, and fish with excessive drift were discarded, blindly to experimental group. Two of 25 fish were excluded, and all the remaining fish (n = 12 MTZ control fish and n = 11 OPC-ablated fish) were included for all subsequent analyses. The data were fed into suite2p for alignment and region of interest (ROI) segmentation. suite2p parameters were kept mostly at their standard values, adjusting values for cell side and temporal sampling frequency. The cell classification and the deconvolution steps were skipped in the suite2p pipeline, and Z-scored raw fluorescence extracted from all detected ROIs was used in all subsequent analyses. The script used for running the data pre-processing with suite2p is available in the code repository.
After alignment, a mask delineating the optic tectum was drawn manually for each fish using pipra 41 , to include only the ROI in this region for further analysis. However, responsiveness of all the best-scoring ROIs for this stimulus were located in the optic tectum, and the conclusions hold regardless of this selection criterion and the exact boundaries of the masks.

Analysis of calcium responses to visual stimulus.
To quantify responses of neurons to individual stimuli, the activity from individual ROIs was de-trended (subtracting the difference between the first and last points), Z-scored and chunked in a window of −2 s to 5 s from the stimulus onset. To compute the reliability score, we obtained the cross-correlation matrix across all stimuli repetitions and averaged all its off-diagonal values. For the average amplitude, the absolute value of the difference was calculated between the integral of the response in the 4 s of the stimulus and during the 2 s of the pre-stimulus pause. For the estimation of the receptive field size, 'visually responsive' (reliability score >0.5) ROIs were selected, and a Gaussian curve to the array of reliability scores was fit over stimulus positions. The variance of the Gaussian was taken as a measure of the width of the tuning curve. All statistical comparisons were performed using the Mann-Whitney U-test (with the implementation in scipy.stats.ranksums 42 ).
Image and data presentation. Images were analyzed with Fiji and Imaris. Morphology reconstructions were carried out with the Imaris FilamentTracer module. Data were prepared and assembled using GraphPad Prism 7, 8 and 9, Fiji and Adobe Illustrator CS6 and 2021.
Statistics and reproducibility. For analyses that involved cohorts of animals or treatment groups, zebrafish embryos of all conditions were derived from the same clutch and selected at random before treatment. No additional randomization was used during data collection. For time-course analyses of OPCs and RGC, zebrafish were screened for single-cell labeling before imaging, and all animals with appropriate expression were used in the experiment. Two fish with excessive drift were discarded for calcium imaging, blindly to experimental group; no other data were excluded from the analyses. We selected sample sizes based on similar sample sizes that were previously reported 18,37,43,44 . No statistical analysis was used to pre-determine sample sizes. Data collection and analysis were not performed blinded to the conditions of the experiments unless stated. Analysis was performed using Microsoft Excel and GraphPad Prism. All data were tested for normal distribution using the Shapiro-Wilk normality test before statistical testing. In the figures, bar graphs are shown as mean ± s.d.; line plots are shown as mean ± s.e.m. or median ± interquartile range (IQR); box and whisker plots are expressed as median ± IQR, minimum and maximum values; and violin plots represent the median ± IQR, minimum and maximum values. For statistical tests of normally distributed data that compared two groups, we used unpaired t-tests. Non-normally distributed data were tested for statistical significance using the Mann-Whitney U-test (unpaired data). For multiple comparisons test, one-way ANOVA was used for parametric data and the Kruskal-Wallis test for non-parametric data (both with Benjamini-Krieger and Yekutieli correction). Repeated measurements were tested using two-way ANOVA with Sidak's correction. We used Fisher's exact test to analyze contingency tables.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data underlying this study will be made available upon reasonable request. Raw data for functional analysis have been deposited at https://doi.org/10.5281/ zenodo.5894603.

Code availability
The codes used for running and analyzing behavioral and imaging experiments are available at https://doi.org/10.5281/zenodo.5894770.  Last updated by author(s): Jan 21, 2022 Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability The data that support the findings of this study are available from the corresponding author upon reasonable request.