Ten‐m3 plays a role in the formation of thalamostriatal projections

The importance of the thalamostriatal pathway for a myriad of brain functions is becoming increasingly apparent. Little is known about the formation of this pathway in mice. Further, while Ten‐m3, a member of the Ten‐m/teneurin/Odz family, is implicated in the proper wiring of mature thalamostriatal projections, its developmental time course is unknown. Here, we describe the normal development of thalamostriatal projections arising from the parafascicular nucleus (PFN) and show a role for Ten‐m3 in its formation. Ten‐m3 is expressed in both the PFN and the striatum by embryonic day 17 (E17). By postnatal day 3 (P3), it had a patchy appearance in the striatum, overlaid on a high dorsal–low ventral expression gradient in both structures. In wild‐type mice, axons from the PFN begin to innervate the striatum by E17. By P3, terminals had ramified but were not confined to any striatal subregion. By P7, the axons had begun to avoid striosomes. The first indication of clustering of thalamic terminals within the striatal matrix was also seen at this time point. The compartmental targeting and clustering of PFN projections became more apparent by P10. Analysis of Ten‐m3 knockout mice showed that while the early developmental progression of the thalamostriatal pathway is conserved, by P10 differences emerged, with a loss of topographic precision and the absence of terminal clustering. No evidence of the involvement of EphA7 downstream of Ten‐m3 was found. Overall, our results suggest that Ten‐m3 plays a role in the consolidation and refinement of thalamic axons to a specific subregion of the striatal matrix.

Despite its functional importance, the normal development of this pathway is not well characterized.Brief reports in cats, opossums, and rats (Fisher et al., 1983;Martin et al., 1989;Vercelli et al., 2003) suggest that thalamostriatal projections form embryonically. Likewise, the molecular mechanisms underlying the formation of this circuit are largely unknown, although an interaction between semaphorin 3E and the plexin D1 receptor has been reported to play a role (Ding et al., 2011).Another promising candidate is Ten-m3, a member of the teneurin/Ten-m/Odz family of glycoproteins (Tran et al., 2015).
The involvement of Ten-m3 in the formation of the thalamostriatal pathway is suggested by its expression in a patchy pattern within the striatal matrix.Further, an overall high dorsal to low ventral gradient has been observed within both the striatum and the PFN.Notably, deletion of Ten-m3 disrupts both the topography and clustering of thalamostriatal projections from the PFN in adult mice (Tran et al., 2015).EphA7 expression is also present in a patchy pattern in the striatal matrix (Janis et al., 1999;Tai et al., 2013).It is currently unclear if or how these expression patterns are linked.
The aims of the current study are to investigate a potential role for Ten-m3 in the initial formation of thalamostriatal projections.To address this, we first characterized the expression pattern of this molecule during embryonic and early postnatal development.We also assessed the normal development of the thalamostriatal pathway, how this relates to Ten-m3's expression pattern, as well as whether the time course of arrival and/or patterning of these projections was impacted by Ten-m3 deletion.Finally, we explored a potential relationship between Ten-m3 and the axonal guidance molecule EphA7 in the striatum.

METHODS
All procedures were performed on mice, approved by the Animal Ethics Committee of the University of Sydney, and conformed to NHMRC guidelines (Protocol: 5424).C57/Bl6 mice were obtained from the Animal Resources Centre (ARC; Perth, Australia).Ten-m3 KO (Ten-m3 −/− ) mice and wildtype (WT: Ten-m3 +/+ ) littermates were maintained on a Bl6/Sv129 cross, as described previously (Leamey et al., 2007).

In situ hybridization
In situ hybridization for Ten-m3 was described previously (Leamey et al., 2007;Tran et al., 2015).EphA7 in situ hybridization was performed on adjacent sections using the same protocol.Briefly, in situ hybridization was performed on 15-μm fresh-frozen coronal and sagittal sections using digoxigenin-labeled sense and anti-sense riboprobes.Primers were designed against a 200-bp region of the gene (Ten-m3: NM_023456; EphA7: NM_010141.3).The reaction was developed using an alkaline phosphatase-catalyzed reaction with a bromochloro-indolyl phosphate (Roche) and nitroblue tetrazolium (Roche) substrate solution.

Western blotting
To determine potential changes in EphA7 protein levels in the absence of Ten-m3, western blotting was performed on striatal tissue from 8-10 WT and Ten-m3 KO mice.Western blotting was performed on striatum that was isolated from the overlying cortex, using rabbit anti-EphA7 antibody (Abcam; 1:100) and compared to a rabbit β-actin monoclonal IgG (cell signaling technology; 1:1000) control.Membranes were washed and imaged using the BioRad ChemiDoc system.Images of Western blots were analyzed for relative intensity using ImageLab software (BioRad).
To determine the location of the PFN in fixed tissue, a standard Nissl stain was performed (as described previously [Tran et al., 2015]) across consecutive coronal sections spanning the rostrocaudal length of the cerebral cortex for each age point.By measuring the distance of the PFN from the rostral pole of the cortex in Nissl-stained sections, and assuming that tissue shrinkage due to histological processing is uniform across a given axis, we calculated the rostrocaudal and dorsoventral position of the PFN in unprocessed tissue.These coordinates are provided in Table 1.
For DiI crystal placement, the brain was exposed and a scalpel was used to make a coronal cut at the appropriate rostrocaudal location of the PFN.A small DiI crystal was placed at the measured dorsolateral location, at the lateral edge of the fasciculus retroflexus (fr), which was often visible.Crystals were placed into dorsolateral PFN in both hemispheres, as the thalamostriatal projection from the PFN is strictly ipsilateral (Smith et al., 2004).Brains were incubated in 1% paraformaldehyde at 37˚C for 3-6 weeks to allow for dye transport.
After the appropriate transport period, brains were embedded and cut at 80 μm in the sagittal plane on a vibratome.Sections were mounted onto glass slides and coverslipped with 50% glycerol in 0.1 M phosphate buffer containing 4′,6-diamidino-2-phenylindole.
In some cases, immunohistochemistry for tyrosine hydroxylase (TH) was performed on DiI-labeled sections to determine the development of the compartmentalization of thalamostriatal terminals.These procedures were described previously (Lee et al., 2008;Tran et al., 2015).

Image collection and analysis
Wide-field images were taken using the AxioCam HR camera and AxioVision Rel.4.7 software on a Zeiss Deconvolution Microscope.High-power, confocal z-stacks were taken using the Leica SPE-II confocal microscope.All sections that were taken as multiple images were merged using the "reposition only" or "interactive layout" modes in the photomerge function of Adobe Photoshop CS3 to form a single panoramic image.For consistency, all analyses were performed on sections from lateral striatum using Image J (NIH).
For postnatal DiI experiments, single sections in lateral striatum from distinct injections at P10 (5 × WT, 5 × KO) were used for analysis.Sections approximately corresponded to 2.64 mm from the midline in the adult mouse brain (Paxinos & Franklin, 2001).Analysis was performed on images taken on the DiI channel.The perimeter of the striatum was traced using the ImageJ "polygon" tool and the enclosed area was determined.Images were subsequently thresholded at a consistent level and processed to produce 100 × 100 matrices and heat maps using R (R project), as described previously (Tran et al., 2015).Heat maps were produced by summing matrices across all cases.Graphs plotting the distribution of label along the dorsoventral axis were generated by summing pixels across y and x axes, respectively, in the cumulative matrices for each age point.The number of positive pixels was summed for each case and averaged across samples for each genotype.
Alignment of adjacent sections labeled for Ten-m3 and EphA7 (n = 3) was performed as described previously (Tran et al., 2015).Briefly, images were warped using IR-tweak (www.sci.utah.edu)according to the location of the striatal border, anterior commissure, and blood vessels, to control for morphological changes resulting from tissue processing.Images were then overlaid using the "linear dodge" function in Adobe Photoshop CS3.To calculate the area of overlap between Ten-m3 and EphA7, images were thresholded and total positive-pixel areas were determined for each gene separately.The "add" function was used to determine the overlapping Ten-m3 and EphA7-positive pixels.This area was calculated as a percentage of the total pixel area for each gene separately.
To determine potential changes to the distribution and intensity levels of EphA7 expression in Ten-m3 KO mice, in situ hybridization was also performed on P3 KO tissue.Intensity analysis was performed to compare expression levels between corresponding sections 3 × WT and 3 × KO.Images were inverted and the background was subtracted before performing thresholding.The polygon tool was used to trace the perimeter of the striatum, and the total pixel count within this contour was determined.

Statistical analyses
Graphs were plotted and data were analyzed using Microsoft Excel, SPSS (IBM), Matlab (Mathworks), or the statistical program R (R Project).Unless otherwise specified, error bars indicate standard error of the mean (SEM).Two-tailed, unpaired t-test, Kolmogorov-Smirnov (K-S) test comparing two distributions, or mixed-model analyses of variance (ANOVAs) were used where appropriate.A significance value of α = 0.05 was assumed for all statistical testing.

Ten-m3 expression in the striatum and PFN during development
We previously reported that Ten-m3 is present in an overall high dorsal to low ventral gradient in the striatum and a topographically corresponding high dorsal to low ventral gradient in the PFN by P3 (Tran et al., 2015).In the striatum, the Ten-m3-positive regions were arranged in a patchy distribution that overlapped with the terminals of thalamostriatal axons.The time course of expression of Ten-m3 in the thalamostriatal pathway and its relationship to the formation of this projection are unknown.To gain more insight into the temporal and spatial relationship between Ten-m3 expression and thalamostriatal projections, we first used in situ hybridization to characterize expression patterns of Ten-m3 in the striatum and PFN across a number of late embryonic and early postnatal time points.
Ten-m3 mRNA expression could not be detected in the striatum or the PFN at E15 (not shown).By E17, expression was detectable and was highest at the dorsal and lateral borders of the striatum with a patchy appearance (Figure 1A).The patchy distribution is highlighted with arrowheads and can be seen more clearly in the inset within Figure 1A.No staining was seen in the sense control (Figure 1B).Overall, Ten-m3 expression increased by P3 as did the high dorsal to low ventral pattern (Figure 1C).In the coronal plane, clear Ten-m3-positive patches could be observed in parallel bands oriented along the dorsomedial-ventrolateral axis of the striatum (Figure 1C, arrowheads).In the sagittal plane, Ten-m3 expression was likewise observed to have a patchy appearance, where it appeared to show highest expression in dorsal and caudal regions (Figure 1D; see also Tran et al., 2015).The patchy expression and dorsal-to-ventral gradient were evident until at least P10 (not shown).
The time course of Ten-m3 expression in the PFN was similar to that observed in the striatum.No expression could be discerned at E15 but was detected at E17 (approximate location indicated by arrow, Figure 1E).At E17, Ten-m3 expression was evident surrounding the fr, particularly dorsolaterally.Expression in this region was at similar levels to the dorsal part of the dorsal lateral geniculate nucleus (asterisk) with no expression seen in the sense control (Figure 1F).The borders of the PFN were more obvious by P3 (delineated by dashed line, Figure 1G).Although Ten-m3 covered the entire PFN by P3, it appeared to be higher in dorsal versus ventral regions at this age (Figure 1G; see also Tran et al., 2015).The sense control showed no staining in this region (Figure 1H).A similar pattern was maintained until at least P10 (not shown).
Thus, Ten-m3 expression emerges between E15 and E17 in both the PFN and the striatum, exhibiting a high dorsal to low ventral expression pattern by P3 in both structures.Within the striatum, there is a clear patchy distribution of Ten-m3 mRNA that is apparent until at least P10.We next investigated how this expression pattern relates to the formation of the thalamostriatal pathway.

Normal development of thalamostriatal projections in wild-type mice
The normal development of thalamostriatal projections has not been previously described in mice.Accordingly, we used DiI tracing in fixed tissue to characterize the early development of this pathway in WTs.Since thalamostriatal axons extend into the striatum in a rostralward direction in rodents (Vercelli et al., 2003), we used sagittal sections to visualize this projection.
Thalamic axons were observed traversing the striatum from E16 (data not shown) but did not send branches into the nucleus at this stage.By E17 (Figure 2A), however, the first collaterals were observed entering the striatum.These projections extended at approximately right angles to their parent fibers to innervate the structure (Figure 2B, arrows).
By P3, thalamic terminals had ramified extensively and were mostly concentrated in rostral and dorsal regions of the striatum (Figure 2C).Label within this area tended to be fairly uniform.Notably, small regions completely devoid of terminal label (within the otherwise labeled area) clearly seen at later stages were not apparent at this age.This indicates that the confinement of thalamic terminals within the striatal matrix compartment observed in adult (Herkenham & Pert, 1981;Sadikot et al., 1992;Xu et al., 1991) had not yet commenced.Double labeling with TH (Figure 2D, green) to mark striosomes confirmed this: thalamostriatal terminals (Figure 2E) and striosomes (Figure 2F) overlapped considerably (Figure 2G), suggesting that thalamic afferents are not confined to the striatal matrix in early postnatal development.Within the matrix, little evidence of the characteristic clustering of terminals reported in adults could be observed (Deschênes et al., 1996;Parent & Parent, 2005).
By P7, thalamostriatal terminals were dense and innervated rostral and dorsal striatum (Figure 2H).In addition, distinct "holes" were noted within otherwise densely labeled regions (Figure 2H, arrow).Double labeling with TH (Figure 2I, green) showed that the "holes" correlated with striosomes.Thus, in contrast to P3, thalamostriatal terminals at this age (Figure 2J) mostly avoided striosomes (Figure 2K), with only a few fibers overlapping with TH-positive regions (Figure 2L), suggesting that refinement mechanisms are well in play by the end of the first postnatal week.While some patchy labeling (indicative of thalamostriatal terminal clustering) could be seen within the matrix (Figure 2H, arrowhead), it was not consistent across all sections at this time point.
By P10 (Figure 2M), thalamostriatal terminals covered most of the dorsal striatum and appeared adult like.Distinct holes with sharp boundaries were visible suggesting a high level of compartmentalization of terminals to the striatal matrix (Figure 2M, arrow).Further, there was clear evidence of a patchy arrangement of thalamic terminals, suggesting that clustering had commenced (Figure 2M, arrowheads).

Development of thalamostriatal projections in Ten-m3 KO mice
To determine whether Ten-m3 may be involved in the development of the thalamostriatal pathway, we examined these projections in Ten-m3 KO mice at corresponding time points sampled in WTs (see above).
Bundles of axons originating from the PFN traversed the striatum at E16 (data not shown).As in WTs, by E17 (Figure 3A), the first collateral branches extending at approximately right angles from their parent fibers could be observed projecting into various regions of the striatum in Ten-m3 KOs (Figure 3B, arrows).
The distribution of thalamostriatal terminals in KOs was mostly comparable to that detected in WTs during the first postnatal week.At P3, KO terminals were dense (Figure 3C), and appeared quite uniform within the labeled region.Doublelabeling sections with TH (Figure 3D, green) showed extensive overlap between terminals (Figure 3E) and striosomes (Figure 3F) at this age (Figure 3G), suggesting that thalamic axons were not confined to the striatal matrix during their initial ingrowth.
By the end of the first postnatal week, terminal label was dense and predominantly concentrated in rostral and dorsal striatum (Figure 3H).As in WTs, distinct "holes" were observed within the labeled region at this age (Figure 3H, arrows).Double labeling for TH (Figure 3I, green) revealed an exact overlap between holes within the label (Figure 3J) and striosomes (Figure 3K).Only a few fibers could be detected within TH-positive regions (Figure 3L).Thus, as in WTs, compartmentalization of thalamostriatal terminals appeared to commence by P7 in Ten-m3 KOs.Within the matrix, terminal label was quite uniform, with no evidence of clustering.Curiously, label also appeared to extend further along the dorsoventral axis of the striatum, rather than remaining confined within the dorsal region as observed in WTs.
By P10 (Figure 3M), thalamostriatal terminals in KOs covered much of the striatum and appeared to spread into more ventral regions compared to WTs (compare Figure 3M with Figure 2M).As in WTs, distinct holes in the label with sharp boundaries reminiscent of striosomes were clearly seen (Figure 3M, arrows), although unbranched axons traversed through these regions.Unlike WTs, however, the patchy pat-terning of terminals within the matrix was not apparent, suggesting that thalamostriatal projections are not clustered in KOs at this age.

The topographic precision of thalamostriatal projections is altered in Ten-m3 KO mice
To further assess the differences in the patterning of thalamostriatal projections between WTs and KOs, we compared the cumulative labeling across multiple subjects (n = 5 for both WTs and KOs; see Section 2) of terminals at P10, the time point at which genotypic differences first emerged.Heat maps revealed that dense thalamostriatal terminals targeted the dorsal half of the striatum in WTs (Figure 4A).In KOs, however, axons appeared to extend much more ventrally (Figure 4B).Quantitatively, the average distribution of terminal label at P10 was significantly different along the dorsoventral axis (Figure 4C; p = 3.046 × 10 -8 , K-S test) between genotypes, while the total amount of label was indistinguishable (Figure 4D; WT: 166,549 ± 64,232 [mean ± SEM], n = 5; KO: 171,527 ± 43,774 [mean ± SEM], n = 5; p = .951,t-test).These findings suggest that the observed genotypic differences in patterning were due primarily to a change in the distribution, rather than the amount of thalamostriatal projections targeting the striatum.

Ten-m3 does not act via EphA7 in the striatum
The mechanisms underlying the apparently multimodal roles of Ten-m3 in topographic mapping and terminal organization are currently unknown.Previously, we proposed that Ten-m3 may act as a direct adhesive target to control the organization of thalamostriatal terminations (Tran et al., 2015).The intracellular domain of teneurin proteins can also be cleaved and translocated to the nucleus to affect gene expression (Bagutti et al., 2003;Drabikowski et al., 2005;Kenzelmann et al., 2008;Nunes et al., 2005).Consistent with this, the expression of a number of genes is altered in visual structures of Ten-m3 KO mice (Glendining et al., 2017).One interesting downstream candidate is EphA7, which is significantly downregulated in the visual pathway of Ten-m3 KOs (Glendining et al., 2017) and is also expressed in a patchy pattern in the striatal matrix (Janis et al., 1999;Tai et al., 2013).
We compared EphA7 striatal expression in WT and Ten-m3 KO.In situ hybridization for EphA7 in WTs showed bands/patches of expression in the striatum (Figure 5A).These patches or stripes of intense EphA7 expression tended to be elongated along the dorsomedial-ventrolateral axis, separated by unlabeled regions.There was no obvious difference in EphA7 expression in the striatum of Ten-m3 KOs.The distribution of signal appeared identical to WTs (Figure 5B).Analysis confirmed no change in signal intensity (Figure 5C; WT: 7018 ± 775 [mean ± SEM], n = 3; KO: 5668 ± 544 [mean ± SEM], n = 3; F(1,4) = 1.3, p = .32,repeated measures ANOVA).A western blot analysis of EphA7 protein in isolated striatal samples similarly showed no change in EphA7 expression levels between WT and Ten-m3 KOs (not shown).
The similarity in expression patterns of Ten-m3 and EphA7 was striking.To further investigate the potential relationship between EphA7 and Ten-m3, we compared the distribution of mRNA signal of these two genes in sections of P3 WT mice (Figure 5).As shown previously, both genes were expressed in bands/patches throughout the striatum (Figure 5D,E).Overlaying Ten-m3 and EphA7 revealed a complex relationship between the two genes (Figure 5F).Although very similar, they largely did not overlap.Rather, EphA7 expression was highest in regions of moderate Ten-m3 levels (Figure 5D-F, arrows), and high Ten-m3 mRNA signal was observed adjacent to regions of peak EphA7 expression (Figure 5D-F, arrowheads).Hence, high Ten-m3and EphA7-positive bands appeared exactly adjacent to one another.Quantitatively, the overlap represented 22.4% ± 1.6% (mean ± SEM) and 20.1% ± 3.1% (mean ± SEM) of Ten-m3 and EphA7 expression, respectively (Figure 5G), suggesting a largely complementary relationship between these genes.

DISCUSSION
Development of the thalamostriatal pathway is not well characterized.To our knowledge, this study is the first to report the major developmental events of this pathway in mice.We show that the first inputs from the PFN to the striatum arose as branches of thalamofugal axons at approximately E17.These projections ramified to become notably more robust by P3.While the normal refinement of these collateral branches to the striatal matrix compartment observed in adults had not commenced at P3, it could be observed by P7 and was more obvious by P10.Likewise, the adult-like patchy arrangement of these terminals within the matrix was first observed at P7 and became more obvious by P10.Furthermore, we investigated a molecular candidate for the development of thalamostriatal projections: Ten-m3.We show that expression of Ten-m3 in the striatum and PFN both temporally and spatially correlated with the major developmental events of the thalamostriatal pathway.In the absence of Ten-m3, thalamostriatal projections did not cluster and were topographically less precise during development.Finally, we investigated the possibility that the actions of Ten-m3 may, in part, be the consequence of functions of EphA7, a downstream target of Ten-m3 in the visual pathway (Janis et al., 1999;Tai et al., 2013).We found an intriguing partially overlapping expression pattern between Ten-m3 and EphA7 in the striatum.EphA7 expression was not altered in the absence of Ten-m3, however, suggesting that EphA7 does not play a major role in the altered thalamostriatal circuitry observed in Ten-m3 KOs.

Technical considerations
We used DiI crystals to trace the thalamostriatal projection in developing mice due to the small size and inaccessibility of the PFN in young animals.Although we are certain that the placement of these crystals targeted the PFN, as only cases where there was evidence of placement just lateral to the fr were included for analysis, the amount of spread of DiI was difficult to account for.Due to the lipophilic properties of DiI, it is readily taken up by neurons of adjacent cells or processes (Godement et al., 1987;Honig & Hume, 1989), such as those of adjacent thalamic structures, some of which also project to the striatum.Hence, it is likely that some other thalamostriatal axons were also labeled.Furthermore, the bulk nature of these tracing studies may have prevented the visualization of terminal clustering at its inception (see below).The technical requirements for DiI tracing also made it impossible to perform in situ hybridization to directly compare the distribution of Ten-m3 to the terminal label.The intense labeling of not only terminals but also axons of passage made meaningful quantification of subtler aspects of terminal organization, such as terminal clustering, difficult to assess.Accordingly, these observations are described only qualitatively.

Normal thalamostriatal development in mice
The first appearance of thalamostriatal collaterals at late embryonic ages is consistent with previous studies in cats, opossums, and rats (Fisher et al., 1983;Martin et al., 1989;Vercelli et al., 2003).Thalamofugal axons traverse the striatum, where the first collateral branches form and innervate the striatum at E17.Many of these axons extend to innervate cortical regions dorsal to the striatum.
Our data suggest that thalamostriatal terminals are not confined to the striatal matrix at P3.By P7, however, initial signs of matrix-specific targeting were evident and became even more apparent by P10.This suggests that these terminals initially target the entire striatum before undergoing refinement to confine themselves to the striatal matrix.Interestingly, the timing of this shift corresponds roughly to when terminals of striosome-specific corticostriatal projections begin clustering within this subregion (Nisenbaum et al., 1998).Further studies will be required to determine if and how the timing of terminal refinement of projections from the thalamus and cortex influences one another.
VGlut2, a marker of thalamostriatal terminals, initially exhibits expression within striosomes before shifting to the matrix by P3-P7 (Nakamura et al., 2005).Preferential expression of VGlut2 within the matrix, however, is not as marked as that observed for thalamostriatal terminals in the current study.This may be due to the fact that VGlut2 labels all thalamostriatal terminals, not just those originating from the PFN.Interestingly, cholinergic interneurons, a major postsynaptic target of thalamostriatal projections from the PFN (Lapper & Bolam, 1992;Smith et al., 2004), have also been reported to show a shift in distribution between striosome and matrix compartments.Cells expressing choline acetyltransferasean established marker of these neurons-show a shift in preference for striosomes to the intermediate zone, a region of the matrix immediately surrounding striosomes, by P7 in rats (Van Vulpen & Van Der Kooy, 1996).It is interesting to note that this timing coincides with the refinement of thalamostriatal terminals to the matrix compartment observed in the present study.
The characteristic patchy organization of thalamostriatal projections from the PFN was first observed in WTs at P7 and became more apparent by P10.It is not clear from our studies if this is the earliest manifestation of this patterning of thalamic terminals in the striatum.As noted above, it is possible that the tracing techniques used here may not have been sensitive enough to show terminal clusters at the time of their earliest formation.Since thalamostriatal clusters have been shown to originate from single PFN neurons in adults (Deschênes et al., 1996;Parent & Parent, 2005), further studies incorporating focal injections to label smaller numbers of neurons could better assess the early development of this patterning.

Ten-m3 expression in the striatum and PFN
Ten-m3 expression in the striatum and PFN at E17 (and not E15) correlates with the first arrival of thalamostriatal branches.Furthermore, the dorsal-to-ventral gradient of Ten-m3 in both structures (Tran et al., 2015), along with the patchy expression in the striatal matrix, correlates spatially with the topographic distribution and clustering of thalamostriatal terminals described previously (Deschenes et al., 1996;Parent & Parent, 2005;Sadikot et al., 1992).
We previously proposed that the patchy expression of Ten-m3 may act to direct the clustering of thalamostriatal terminals within the matrix (Tran et al., 2015).Patchy Ten-m3 expression is observed as early as E17.In con-trast, the patchy patterning of thalamostriatal terminals did not become evident until P7-10.This time delay suggests that the Ten-m3-positive patches may play a role in the refinement, stabilization, and clustering of thalamostriatal synapses, rather than in guiding axons to these patches per se.A role in synapse formation is consistent with what has been reported in limbic circuits (Zhang et al., 2022), but is somewhat different to the role Ten-m3 appears to play in the development of the visual pathway (Glendining et al., 2017), where it has a clear impact on axonal guidance.It should also be noted that, at least in the hippocampus, presynaptic Ten-m3 mediates synapse formation via heterophilic interactions with latrophilins postsynaptically, rather than homophilic interactions with other Ten-m3 molecules (Zhang et al., 2022).Whether this extends to the thalamostriatal pathway is currently unknown.Analysis of this pathway using a conditional Ten-m3 would be required to address this possibility.

Ten-m3 is important for the development and topography of thalamostriatal projections
The first thalamostriatal branches occur at E17 in Ten-m3 KOs, as they do in WTs, suggesting that the collateralization of these projections is conserved in these animals.The early development of this projection was surprisingly well conserved in Ten-m3 KOs, with no clear differences emerging until P10.The differences observed at this age point were consistent with those previously reported in adults with both a lack of clustering and a loss of topographic precision along the dorsoventral axis (Tran et al., 2015).This indicates that the differences in thalamostriatal organization seen here are maintained into adulthood.Further it suggests that the differences seen in adulthood, which were associated with functional changes, arise directly from these aberrant developmental processes.
The corresponding Ten-m3 gradient and thalamostriatal topography along the dorsoventral axis suggest a direct role for the gene in the development of this organization.The homophilic adhesive properties of Ten-m3, along with other teneurins, have been suggested to govern these axon guidance and synaptic organization roles (Beckmann et al., 2013;Berns et al., 2018;Dharmaratne et al., 2012;Hong et al., 2012;Leamey et al., 2007;Mosca et al., 2012) and could be at play here.

Ten-m3 function in thalamostriatal development is not mediated by EphA7
In addition to direct adhesive properties, teneurins are hypothesized to perform indirect functions by altering gene expression.This is believed to result from the cleavage and translocation of the intracellular domain to the cell nucleus (Bagutti et al., 2003;Kenzelmann et al., 2008;Nunes et al., 2005;Rubin et al., 1999).A number of axon guidance cues have previously been reported as downstream targets for Ten-m3 (Glendining et al., 2017).We investigated one of these candidates, EphA7, which has been shown to be expressed in the striatum (Janis et al., 1999;Tai et al., 2013).
The striking pattern of Ten-m3-positive patches was paralleled by a similar expression of EphA7 in the striatum.Surprisingly, these genes marked mostly distinct, but adjacent, groups of neurons.Previous work proposed that EphA7positive patches mark a matrisome compartment linked to striatofugal connections in rats (Tai et al., 2013).Matrisomes are striosome-like subregions within the matrix only previously identified by afferent and efferent organizations in primates and cats (Desban et al., 1989;Flaherty & Graybiel, 1993;Gimenez-Amaya & Graybiel, 1991;Jimenez-Castellanos & Graybiel, 1989).Hence, our results suggest that Ten-m3 may mark a novel matrisome compartment associated with the thalamostriatal circuit.
The significance of the intriguing spatial relationship between Ten-m3 and EphA7 is unclear.Similar interdigitation of matrisomes formed by corticostriatal terminals from somatosensory cortical regions has been reported (Flaherty & Graybiel, 1993;Malach & Graybiel, 1986).Intriguingly, inputs from ipsilateral somatosensory cortex to the striatum have been shown to avoid EphA7-positive patches (Tai & Kromer, 2014).Cortical projections representing the same body part from ipsilateral and contralateral somatosensory cortex are also reported to interdigitate in distinct matrisomes (Malach & Graybiel, 1986).These observations suggest a high-order integration of functions for striatal compartments, such as "reordering" corticostriatal inputs into different combinations (Ragsdale & Graybiel, 1990) and integrating inputs with those of similar outputs.It would be of interest to determine the role of the Ten-m3-positive compartment in such high-order integration functions.It is possible that Ten-m3 and EphA7 could chemically define functionally distinct, parallel corticostriatal loops to mediate diverse functions (Mandelbaum et al., 2019).Further studies are required to determine the significance of the relationship between Ten-m3-positive cells receiving thalamostriatal input, EphA7positive striatofugal neurons, and corticostriatal projections in rodents.

CONCLUSION
Overall, our data provide evidence of a role for Ten-m3 in the development of the thalamostriatal projection.This molecule is expressed in corresponding patterns within the PFN and striatum from the initial formation of the pathway.Our data suggest an important role for Ten-m3 in the refinement and clustering of thalamostriatal projections to the matrix and in topographic precision along the dorsoventral axis.These roles may be significant for the arrangement of these projections within functional subregions of the striatum and for the highly ordered compartmental integration of information with respect to other afferent and efferent systems.

C O N F L I C T O F I N T E R E S T S T A T E M E N T
The authors declare no conflicts of interest.

D A T A AVA I L A B I L I T Y S T A T E M E N T
Data will be made available upon reasonable request to the authors.

F
I G U R E 1In situ hybridization reveals Ten-m3 mRNA distribution in the developing striatum and parafascicular nucleus (PFN).(A, B, E, F) Coronal sections through the striatum (A, B) and PFN (E, F) at embryonic day 17 (E17).Ten-m3 mRNA (A, E) was first observed in both structures at this time point.Expression was observed as patches in the striatum (A; arrowheads), which were more apparent in the dorsal region.The patches are seen more clearly in the inset (boxed region).Sense controls of nearby sections show no staining (B, F).Although the boundaries of the PFN could not be clearly discerned at E17, Ten-m3 expression was seen in the thalamus, in close proximity to the fasciculus retroflexus (fr; area of presumptive PFN demarcated by arrow).Expression in this region is similar to that in the dorsal part of the dorsal lateral geniculate nucleus (*).(C, D, G, H) Ten-m3 expression in the striatum and PFN showed an overall high dorsal to low ventral distribution in both structures at postnatal day 3 (P3).In coronal section of the striatum (C), Ten-m3-positive cells formed patches and bands, oriented approximately parallel to the dorsolateral border of the nucleus, along the dorsomedial (DM) to ventrolateral (VL) axis (C; arrowheads).Patches and bands of Ten-m3 expression were also evident in the sagittal plane (D; arrowheads) and tended to be higher in dorsal and caudal striatum.In the PFN, Ten-m3 was expressed throughout the nucleus (borders delineated by dashed line), being highest dorsally and lowest ventrally (G).Sense control (H) shows no staining.Scale bars indicate 500 μm.Scale bars in panels (B), (C), (F), and (H) also apply to panels (A), (D), (E), and (G), respectively.In panels (A-C) and (E-H), dorsal is up and lateral is right as indicated in panel (B).For panel (D), dorsal is up and caudal is right, as indicated.

F
Development of thalamostriatal projections in wild-type (WT) mice.(A, B) Sagittal sections show thalamic axons traversing the striatum (boundary delineated by dashed line) at embryonic day 17 (E17).A higher power image of the area indicated in panel (A) shows the first thalamic branches forming at approximately right angles to the parent axons extending into the striatum at this age (B, arrow).(C-G) At postnatal day 3 (P3), thalamic terminals branched within rostral and dorsal striatum.Double labeling for thalamic terminals and striosomes (mu-opioid receptor, green; D) showed little compartmentalization.A higher power image of the area indicated in panel (D) shows terminals (E) and striosomes (F) overlapping considerably (G), suggesting that the confinement of terminals to the matrix observed in adults is yet to be initiated at P3. (H-L) At P7, thalamic terminals were more abundant.Clear evidence of "holes" (H, arrow) in otherwise labeled regions was present.Double labeling for thalamic terminals and striosomes (mu-opioid receptor, green; I) indicated confinement of terminals to the matrix.A higher power image of the area indicated in panel (I) showed terminals (J) mostly avoided striosomes (K), with only a few overlapping fibers (L).Some evidence of clustering of thalamostriatal terminals (H, arrowhead) could also be seen.(M) Thalamostriatal terminals were extensive by P10 and showed clear evidence of clustering (arrowheads) and "holes" with sharp boundaries (arrow), suggesting a high level of compartmentalization. Scale bar in panel (C) applies to panel (D).Scale bar in panel (H) applies to panel (I).Scale bars in panels (A), (C), (H), and (M) indicate 500 μm.Scale bars in panels (B), (E-G), and (J-L) indicate 100 μm.For all images, dorsal is up and rostral is left.

F
Development of thalamostriatal projections in Ten-m3 KO mice.Sagittal sections through the striatum in Ten-m3 KO mice.(A, B) Thalamic axons traversed the striatum (A, boundary delineated by dashed line) at embryonic day 17 (E17).A high-power image of the area indicated in panel (A) shows the first thalamic branches extending into the striatum at this age (B, arrows).(C-G) At postnatal day 3 (P3), thalamic terminals ramified in much of the striatum with little evidence of clustering (C).Double labeling for thalamic terminals and striosomes (mu-opioid receptor, green; D) showed little evidence of compartmentalization.A high-power image of the area indicated in panel (D) shows terminals (E) and striosomes (F) overlapping considerably (G).(H-L) At P7, thalamic terminals were denser and filled much of the striatum."Holes" were present within otherwise labeled areas (H, arrows).Double labeling for thalamic terminals and striosomes (mu-opioid receptor, green; I) indicated that confinement of terminals to the matrix had commenced.A high-power image of the area indicated in panel (I) shows that terminals (J) mostly avoided striosomes (K), with only a few overlapping fibers (L).Within the matrix, there was no evidence of clustering of thalamic terminals.(M) Thalamostriatal terminals were extensive by P10, with distinct "holes" with sharp boundaries (arrows) suggesting a high level of compartmentalization.No evidence of clustering of thalamostriatal terminals was apparent.Scale bar in panel (C) applies to panel (D).Scale bar in panel (H) applies to panel (I).Scale bars in panels (A), (C), (H), and (M) indicate 500 μm.Scale bars in panels (B), (E-G), and (J-L) indicate 100 μm.For all images, dorsal is up and rostral is left.

F
I G U R E 4 Developing thalamostriatal projections are altered in Ten-m3 KO mice.(A, B) Heat maps representing cumulative label in WTs (A) and KOs (B) at postnatal day 10 (P10).(C) Distribution of label along the rostrocaudal axis from cumulative heat maps in (A) and (B).Distribution of label was significantly different between genotypes along the rostrocaudal axis (C, ***p = 3.046 × 10 -8 , K-S test).(D) The average total pixels were similar between genotypes at P10.For column graphs, data indicate mean ± SEM.KO, knockout; WT, wild-type.

F
EphA7 expression is not impacted by Ten-m3 deletion.(A, B) Coronal sections show in situ hybridization for EphA7 mRNA in WT and Ten-m3 KO mice.(A, B) EphA7 was expressed in a pattern of bands and patches in WTs (A).This expression pattern was conserved in Ten-m3 KO mice (B).The patchy pattern is visible at higher power in the insets (boxed regions).(C) The level of EphA7 expression (as measured by pixel counts) was not significantly altered in Ten-m3 KOs compared to WTs. (D-F) Comparison of Ten-m3 and EphA7 expression in the striatum.(D) Ten-m3 expression was distributed in patches and bands (arrowheads).(E) EphA7 was expressed in a similar pattern of bands (arrows).(F) Overlay of panels (D) and (E).There was a partial overlap between Ten-m3 and EphA7 expression.Ten-m3 expression was highest in areas of modest EphA7 expression (arrowheads).(G) Signal overlap of Ten-m3 and EphA7 expression, respectively.Relatively low levels (∼20%; see text) suggest a largely complementary relationship between these genes.KO, knockout; WT, wild-type.
Heidi Tran was supported by an Australian postgraduate research award Open access publishing facilitated by The University of Sydney, as part of the Wiley -The University of Sydney agreement via the Council of Australian University Librarians.
Coordinates for tracer placement (from rostral pole of cortex).