Re‐evaluating the actin‐dependence of spectraplakin functions during axon growth and maintenance

Abstract Axons are the long and slender processes of neurons constituting the biological cables that wire the nervous system. The growth and maintenance of axons require loose microtubule bundles that extend through their entire length. Understanding microtubule regulation is therefore an essential aspect of axon biology. Key regulators of neuronal microtubules are the spectraplakins, a well‐conserved family of cytoskeletal cross‐linkers that underlie neuropathies in mouse and humans. Spectraplakin deficiency in mouse or Drosophila causes severe decay of microtubule bundles and reduced axon growth. The underlying mechanisms are best understood for Drosophila’s spectraplakin Short stop (Shot) and believed to involve cytoskeletal cross‐linkage: Shot's binding to microtubules and Eb1 via its C‐terminus has been thoroughly investigated, whereas its F‐actin interaction via N‐terminal calponin homology (CH) domains is little understood. Here, we have gained new understanding by showing that the F‐actin interaction must be finely balanced: altering the properties of F‐actin networks or deleting/exchanging Shot's CH domains induces changes in Shot function—with a Lifeact‐containing Shot variant causing remarkable remodeling of neuronal microtubules. In addition to actin‐microtubule (MT) cross‐linkage, we find strong indications that Shot executes redundant MT bundle‐promoting roles that are F‐actin‐independent. We argue that these likely involve the neuronal Shot‐PH isoform, which is characterized by a large, unexplored central plakin repeat region (PRR) similarly existing also in mammalian spectraplakins.


INTRODUCTION
Axons are the slender, up to 2 m long processes of neurons that form the biological cables wiring our bodies (Prokop, 2020). Their de novo formation during development, regeneration, or brain plasticity is implemented at growth cones (GCs), the amoeboid tips of extending axons (Harrison, 1910;Ramón y Cajal, 1890). GCs navigate by sensing spatiotemporally patterned chemical and mechanical cues along their paths which are translated into orchestrated morphogenetic changes leading to axon extension (Franze et al., 2013;Sanes et al., 2019;Tessier-Lavigne & Goodman, 1996). These morphogenetic changes are mediated by the cytoskeleton, in particular, actin and microtubules (MTs; Dent et al., 2011;Lowery & van Vactor, 2009;Prokop et al., 2013;Tanaka & Sabry, 1995): F-actin in the GC periphery is required for explorative protrusive activity and mechanotransduction leading to the directional stabilization of MTs which, in turn, implement the actual growth events (e.g., Buck & Zheng, 2002;Geraldo et al., 2008; A. C. Lee & Suter, 2008;Qu et al., 2019;Suter & Forscher, 2001). When MTs in GCs arrange into bundled loops or spools, they seem to suppress such interactions in the periphery and slow down axon growth (Dent et al., 1999).
The Drosophila spectraplakin Short stop (Shot) is a close orthologue of dystonin and ACF7/MACF1; in neurons, Shot is required for axon and dendrite growth, neuronal polarity, axonal compartmentalization, synapse formation, and axonal MT bundle maintenance (S. Lee et al., 2000;Prokop et al., 1998;Reuter et al., 2003;Voelzmann et al., 2017). In Shotdeficient neurons, MT bundles in axon shafts and GCs frequently disintegrate into disorganized, curled, criss-crossing arrangements (from now on referred to as MT curling). This dramatic MT phenotype can be rescued when reinstating actin-MT cross-linking activity of Shot, through a mechanism where Shot guides the extension of polymerizing MTs along the axonal cortex into parallel bundles (Alves-Silva et al., 2012;Hahn et al., 2021;Sánchez-Soriano et al., 2010). The C-terminal MT interaction involved in this function of Shot is quite well described. In contrast, little is known about Shot's N-terminal interaction with neuronal F-actin networks; for example, how it is influenced by different forms of F-actin networks which can present as sparse cortical F-actin rings in the axon shaft (Leterrier et al., 2017;Qu et al., 2017;Xu et al., 2013) or dense lattice-like or bundle-forming F-actin networks in GCs (Dent et al., 2011).
Here, we have gained new understanding of Shot's F-actin interaction. First, we show that Shot function does not simply depend on F-actin: it rather appears to involve a wellbalanced interplay of low-affinity CH domains with F-actin networks, where any changes can trigger alterations in Shot's functional output; this phenomenon directs the formation of MT spools relevant for axon growth regulation. In the axon shaft, Shot is required for MT bundle maintenance through the above-mentioned guidance mechanism depending on Factin/MT/Eb1 cross-linkage. In addition, we provide strong indications that Shot performs actin-independent bundlemaintaining functions that act redundantly to F-actin/MT/Eb1 guidance. We argue that these functions are mediated by the neuronally enriched Shot-PH isoform. Shot-PH is the only isoform displaying an evolutionarily conserved plakin repeat region (PRR; Hahn et al., 2016;Röper & Brown, 2003;Voelzmann et al., 2017), which is functionally unexplored and might hold the key to uncharted mechanisms of axon biology and architecture. F I G U R E 1 Different Shot constructs and their localization. (a) Illustration of different Shot isoforms as a function of different start sites (A*-D*) and splice-in of different exons (X, PRR); different domains and motifs are color-coded (CH, calponin homology; PD, plakin domain; PRR, plakin repeat region; SRR, spectrin repeat region; EFH, EF-hand; GRD, Gas2-related domain; MtLS, MT tip localization sequence which forms the Eb1-binding motifs); positions of the epitope used to generate the Shot-C antibody (Strumpf & Volk, 1998), the kakP2 P-element insertion (blocking the a* and b* start sites) and the break-point of the V104 inversion (deleting the Ctail) are indicated in red. (b-j) Different UAS-constructs expressing modified Shot versions, color-coded as in (A) and GFP indicated by a yellow star; newly generated constructs are indicated by asterisks, origins of all other constructs are provided in the Methods section. (b'-j') Primary neurons at 6-8 h in vitro (HIV) cultured on glass which express the respective constructs on the left and are stained for actin (red), tubulin (green), and GFP (blue); wild-type reference neurons are not shown but can take on any of the shapes displayed in B'-J' (see examples in Figures 2a, 4b, 6a, 8a, 8g, and Figure S1a). (b″-j″) GFP channel are shown in grayscale. In all images, asterisks indicate cell bodies, arrow heads the axon tips; scale bar in (A) represents 10 μm in all images. (k and l) Graphs display the distribution of axon length phenotypes (K) and frequency of spools in neuronal growth cones (GCs) (L) taken from neuron populations expressing the same constructs as displayed in b″-j″. Number of neurons analyzed are shown in orange, median values in blue (k only), black numbers within columns in (L) indicate the percentage of neurons with spool-containing GCs; black/gray numbers on the right of each plot/bar indicate the p-values obtained via Mann-Whitney rank sum tests in (K) (Kruskall-Wallis analysis of variance [ANOVA] test results shown above) and chi-square tests in (l). Data were normalized to wild-type controls performed in parallel to all experiments (red dashed lines)

Roles of Shot's actin-binding domain in gain-of-function experiments
To assess F-actin dependency of Shot function, we first took a gain-of-function (GOF) approach. For this, we targeted the expression of transgenic Shot constructs to primary Drosophila neurons which were fixed after 6 h in vitro (HIV) and analyzed for two phenotypes: we quantified the length of axons and the number of neurons showing growth cones with bundled loops (referred to as ''spools''; Figure 1b',g',h')as opposed to GCs with ''pointed'' (Figure 1c',i',j') or ''disorganized'' MTs (Figure 1f';Sánchez-Soriano et al., 2010;Teng et al., 2001). Neuronal expression of Shot-PE::GFP (a GFP-tagged version of the best-studied PE isoform; Hahn et al., 2016; Figure 1a,b) caused a reduction in axon length at fixation stage to ∼80% and doubled the number of MT spools in GCs compared to wild-type controls (Figure 1b',k,l). In contrast, Shot-PC::GFP (another natural isoform which lacks CH1; Figure 1a,c) failed to induce either of these phenotypes; instead it showed a slight trend to suppress spool numbers below control levels ( Figure 1c,l), as observed in previous studies (Sánchez-Soriano et al., 2010). The finding suggests that an interaction with F-actin is essential for spool formation, since the lack of CH1 in the Shot-PC isoform (Figure 1c) eliminates F-actin interaction (concluded from previous localization and binding studies; S. Lee & Kolodziej, 2002). Accordingly, spool induction can also be suppressed when depleting F-actin with the drug latrunculin A (LatA; Figure 2b,d; Sánchez-Soriano et al., 2010). F I G U R E 2 Impact of drug-induced F-actin inhibition on Shot-PE function. (a-c) Primary neurons at 6-8 h in vitro (HIV) on glass treated with DMSO (control), LatA or CK666 as indicated and stained for GFP (green), tubulin (red), and actin (blue); grayscale images below show single channels as indicated; asterisks indicate cell bodies, arrowheads the tips of axons; scale bar in (a) represents 10 μm in all images. (d) Frequency of neurons with growth cones (GCs) that contain spools (examples of neurons in a-c are assigned to their respective data columns via color-coded squares); orange numbers indicate the sample numbers (number of neurons analyzed), white numbers within columns the percentage of neurons with GCs that contain no spools; numbers on the right of each graph indicate the p-values obtained via chi-squared tests. Data were normalized to wild-type controls performed in parallel to all experiments (dashed red line) Shot-PE and Shot-PC not only differ in the presence/absence of CH1 they also display different lead sequences that flank CH domains N-terminally (blue A* vs. yellow C* in Figure 1a-c; Hahn et al., 2016). Both lead sequences lack any informative homologies or motifs but may still have modifying impacts on CH domain functions (Yin et al., 2020). Therefore, we generated Shot-PE variants containing the A* lead sequence but lacking single or both CH domains (Figure 1d-f). When expressing these variants, Shot-PE-ΔABD::GFP (lacking both CH domains) caused phenotypes almost identical to those of Shot-PC (Figure 1f',k,l), thus corroborating former claims that the actin-binding capability of Shot-PC is negligible (S. Lee & Kolodziej, 2002).
In contrast, single CH domain deletions generated surprising results. As mentioned above, CH1 is the main actinbinding domain of the tandem, and we expected therefore that Shot-PE-ΔCH2 would have actin-binding, hence spoolinducing, capability whereas Shot-PE-ΔCH1 would behave like Shot-PC or Shot-ΔABD. However, the opposite was true: deleting the functionally less prominent CH2 caused robust axon elongation to ∼120% and failure to induce extra spools, suggesting that the CH1 domain alone fails to mediate actin-binding properties (Figure 1e,k,l). In contrast, Shot-PE-ΔCH1 expression had a trend toward extra spool formation and shorter axons, suggesting modest actin-binding properties although CH1 was absent (Figure 1d',k,l). Since Shot-PE-ΔCH1 and Shot-PC only differ in the presence of either the A* or the C* lead sequence (Figure 1c,d), our results might hint at potential regulatory roles: for example, the C* lead sequence of Shot-PC might inhibit residual actin affinities of CH2 but not the A* sequence of Shot-PE, thus explaining why Shot-RE-ΔCH1 seemed to display more activity than Shot-RC and Shot-RE-ΔABD.

F-actin is required for Shot construct localization
To gain more understanding of these phenotypes, we performed localization studies. Shot-PE::GFP was strongly enriched at the distal ends of axons, mostly at the actinenriched GCs; this was consistent with its spool-inducing activity ( Figure 1b″). Also, Shot-PC::GFP and Shot-PE-ΔABD::GFP were distally enriched in axons (Figure 1c″,f″), suggesting that their inability to induce spools was not due to their physical absence but rather their functional impairment.
Also, Shot-PE-ΔCH1::GFP was enriched in distal axon segments ( Figure 1d″). This localization was consistent with its spool-inducing tendencies (potentially mediated by residual F-actin affinity of CH2; see above). In contrast, the Shot-PE-ΔCH2::GFP construct was retained at or actively localized to proximal axon segments (Figure 1e'), consistent with the absence of its spool-inducing activity (Figure 1l).
Taken together, except Shot-PE-ΔCH2::GFP all Shot constructs localized distally, including those lacking the F-actinbinding CH domains. This seemed to contradict further findings that Shot-PE::GFP lost its tip localization upon removal of F-actin with LatA (''GFP'' in Figure 2b)-similarly observed also with the F-actin-inhibiting drug cytochalasin D (CytoD; Figure S1b). This actin dependence in the absence of CH domains did not involve C-terminal domains of Shot: the GFP-tagged Cterminus (Shot-EGC::GFP; comprising EF-hand motifs and the MT-binding GDR and Ctail; Figure 1j) localized homogeneously along axonal MTs, and did not induce extra spools or axon shortening (Figure 1j-l; Alves-Silva et al., 2012). Instead, we focused on the N-terminal plakin domain because Shot-PE-Δplakin::GFP had been reported to display transient localization defects in developing embryonic motor nerves (Bottenberg et al., 2009). However, like most other constructs, Shot-PE-Δplakin::GFP displayed distal localization in primary neurons (Figure 1i″), but it failed to induce robust spool formation or axon shortening (Figure 1k,l; consistent with its partial deficits in supporting axon growth in vivo; Bottenberg et al., 2009).
Taken together, our data suggest complex regulations at the N-terminus. We propose that two domains can mediate F-actin association: CH domains through direct binding, and the plakin domain (which contains a SRC Homology three motif of protein interaction; ''SH3'' in Figure 1a) through association with independent factors that are localized at GCs through F-actin (e.g., transmembrane proteins; see Discussion). In this scenario, distal localization of Shot could be mediated by either the CH domains or the plakin domain alone, but its spool-inducing function would depend on both domains in parallel; this would explain why single deletion of either the plakin or the CH domains abolishes Shot's spoolinducing activity but not its localization.

Qualitative or quantitative changes of F-actin interaction influence Shot's MT-regulating roles
As explained above, we propose that Shot interacts with Factin networks through both the plakin and CH domains. This raises the question of whether Shot uses F-actin as a mere anchor or whether its function is influenced by changes in the quantity and quality of F-actin networks. To address this, we first introduced quantitative and qualitative changes to F-actin networks by manipulating actin nucleation, that is, the process of seeding new actin filaments.
In Drosophila primary neurons, nucleation is performed primarily by the formin DAAM and the Arp2/3 complex Prokop et al., 2011); of these, Arp2/3 is expected to contribute branched networks that are qualitatively different from those nucleated by formins (Blanchoin et al., 2014). Arp2/3-mediated actin nucleation can be specifically inhibited by CK666 (Hetrick et al., 2013). When applying 100 nM CK666 for 2 h, we observed a reduction in filopodia numbers to 72 ± 5% (p Mann-Whitney < .001, n = 80), indicating successful Arp2/3 inhibition and a reduction in F-actin abundance . Under these conditions, Shot-PE::GFP was still recruited to the distal axon, but its spool-inducing activity was strongly suppressed (Figure 2c,d). This finding supports our hypothe-sis that quantitative and/or qualitative changes of F-actin networks impact MT regulatory roles of Shot.
To further challenge this notion, we decided to exchange the two CH domains of Shot for conceptually different actinbinding domains taken from other proteins. For this, we chose the 17 residue actin-binding motif Lifeact (Life) from the Saccharomyces cerevisiae protein Abp140 (Riedl et al., 2008), and the C-ERMAD domain of Moesin (Moe; Kiehart et al., 2000;Millard & Martin, 2008). When extrapolating from binding studies reported for CH domains of α-actinin (closely related to those of Shot; Figure S2), we expected that Shot's CH domains bind F-actin modestly, whereas Life should bind F-actin more robustly in a phalloidin-like manner (Lemieux et al., 2014). In contrast, Ezrin's actin-binding domain (closely related to Moe; Fritzsche et al., 2013Fritzsche et al., , 2014 had been shown to dissociate even faster from F-actin than αactinin's CH domains, consistent with observations that fulllength Moesin does not strongly co-localize with F-actin in embryonic chick neurons or PC12 cells (Amieva & Furthmayr, 1995;Marsick et al., 2012). We, therefore, predicted a gradual impact of the different actin-binding domains on Shot localization and/or function in the hierarchical sequence Life > Shot CH1+2 ≥ Moe.
We first analyzed the localization of the different actinbinding domains fused to the N-terminal lead sequence of Shot-PE (GFP::A*::CH1+2, GFP::A*::Life, GFP::A*::Moe; Figure S3) by transfecting them into Drosophila primary neurons. Like GFP controls, also GFP::A*::CH1+2 and GFP::A*::Moe were distributed fairly homogeneously throughout entire neurons, consistent with their expected low affinity for F-actin ( Figure S3a-c). In contrast, GFP::A*::Life showed the expected robust, phalloidin-like staining ( Figure  S3d). None of the three fusion constructs caused any obvious MT phenotypes ( Figure S3e). We next replaced both CH domains in Shot-PE::GFP with Life or Moe (Figure 1g,h) and generated transgenic flies using the same genomic landing site as utilized for other transgenic constructs in this study (see Methods); this makes sure that the expression strength was comparable between constructs (Bischof et al., 2007). When targeted to primary neurons, Shot-PE-Moe::GFP behaved like the ΔCH1 and Δplakin constructs: it was enriched along MTs in distal axons accompanied by mild axon shortening and a trend toward increased spool formation (Figure 1g″,k,l). In contrast, Shot-PE-Life::GFP localized strongly in GCs but also along axons (Figures 1h″ and 3; Figure S4) and caused axon shortening and spool induction to similar degrees as Shot-PE::GFP (Figure 1k,l). However, other subcellular features were strikingly novel: (1) 38% of Shot-PE-Life::GFP-induced MT spools in GCs had a ''tennis racket'' appearance with many MTs projecting diffusely through the center of spools ( Figure 3a and ''white arrows'' in Figure S4); (2) a number of neurons showed unusual MT bundles in close proximity  Figure S4); (3) about 60% of axonal MT bundles were split into two parallel portions that were decorated with strong Shot-PE-Life::GFP staining, and closely accompanied by F-actin staining that was unusually strong for axon shafts (Figure 3b,c and ''white arrowheads'' in Figure S4); these constellations suggested that the hybrid construct firmly cross-links and alters the sub-cellular arrangement of MTs and F-actin while taking on an unusual localization itself (Figure 3 and Figure S4; see Discussion). The aberrant localization of Shot-PE-Life::GFP and its dominant MT phenotypes were clearly abolished when treating neurons with LatA, thus demonstrating the F-actin dependence even of this powerful hybrid construct (Figure 3e-g).
Taken together, our GOF analyses suggested that the quality and quantity of F-actin networks can regulate Shot's MT bundle-inducing function. The low affinity of Shot's CH domains seems ideally tuned to read those differences in Factin: high abundance of F-actin induces spools in GCs, while increases in Shot's F-actin affinity (Shot-PE-Life) cause MT rearrangements (split bundles) even in axon shafts where Factin networks are usually sparse (Leterrier et al., 2017;Qu et al., 2017;Xu et al., 2013).

Shot's axon length regulation involves MT spool formation in GCs and MT bundle maintenance
Our key readout for Shot GOF was the formation of MT spools in GCs. MT spools have been suggested to inhibit axon growth (Dent et al., 1999;Sánchez-Soriano et al., 2010). Accordingly, we found a strong negative correlation between spools and axon lengths when plotting the data from our overexpression experiments (black dots in Figure 4a); also neurons without Shot GOF plotted onto this curve (Figure 4f,g and orange dots in [a]), including untreated wild-type neurons, neurons treated with LatA (less spools, enhanced axon length), or neurons lacking the F-actin-promoting factor Chickadee (Chic, the sole profilin in Drosophila; Gonçalves-Pimentel et al., 2011; slightly less spools, modest increase in axon length). Also, spool formation in neurons without Shot GOF seems to be mediated by Shot, as was suggested by shot mutant neurons where spool numbers were strongly reduced compared to wild-type (Figure 4f; Sánchez-Soriano et al., 2010).
However, shot mutant neurons do not plot onto the correlation curve (blue dots in Figure 4a): instead of showing axon extension that would usually correlate with the absence of spools, their axons were very short. Furthermore, combinatorial studies revealed that the short axon phenotype of shot overrides LatA-or chic-induced axon elongation (Figure 4e,g). These short axon phenotypes of shot seemed to mirror the occurrence of MT disorganization in shot mutant neurons, where axonal bundles lost their parallel arrangements and took on curled, criss-crossing appearances  , (R4) (Hahn et al., 2021); red arrow heads point at specific functional lesions in the different conditions. Explanations: in wild-type neurons, CytoD eliminates cortical actin and weakens MT polymerization (pale Eb1 with dashed outline), not strong enough to affect parallel MT arrangements but leading to MT gaps (interrupted green line; b); in the absence of Shot, MTs curl (c) and MT networks shrink (they become (referred to as MT curling; Figure 4c). Like the axon length phenotype, axonal MT curling was not influenced by LatA treatment or loss of Chic (Figure 4e,g,h), thus demonstrating a further parallel between both phenotypes.

Shot seems to work through two redundant mechanisms in MT bundle maintenance
Previous work had demonstrated that Shot prevents MT curling through a guidance mechanism involving F-actin/Eb1/MT cross-linkage: via its N-terminus Shot binds cortical F-actin and via its C-terminus to MTs and Eb1-thus guiding the extension of polymerizing MTs along the axonal cortex into parallel bundles; this F-actin/Eb1/MT guidance mechanism is supported by numerous structure-function, loss-of-function, and pharmacological and genetic interaction studies (details in Figure 5; Alves-Silva et al., 2012;Hahn et al., 2021;Qu et al., 2019;Sánchez-Soriano et al., 2009).
The F-actin/Eb1/MT guidance mechanisms would predict that removal of cortical F-actin from wild-type neurons (which can be achieved with the F-actin-inhibiting drug CytoD, but less so with LatA or loss of Chic; Qu et al., 2017) should mimic the shot mutant MT curling phenotype. However, CytoD application to wild-type neurons failed to cause MT curling; instead it caused a deficit in MT polymerization leading to gaps in MT bundles (likely due to loss or shortening of MTs, which therefore fail to overlap; Figures S1b, S6b, and S5b; Qu et al., 2017)-which may also explain why loop suppression upon CytoD application ( Figure S1) does not enhance axon growth as observed with LatA (Sánchez-Soriano et al., 2010).
The fact that CytoD failed to mimic the MT curling phenotype of shot mutant neurons (Figure 6b vs. c) might indicate vulnerable to lack of actin-promoting effects causing more severe loss of MTs; stippled green line; d); guiding function is fully re-instated by targeted expression of Shot-PE (e; expression constructs red encircled with a green GFP dot at their ends); Shot-PE fails to guide MTs in the absence of actin, but it protects MT polymerization (f); Eb1 deficiency eliminates the F-actin/MT/Eb1 guidance mechanism, and might even be involved in the alternative mechanism of Shot (arrow of pale PH toward MT plus end; g; see Discussion); MT curling upon reduced Eb1 levels (Eb1↓) can be rescued with Shot-PE expression (h); MT curling caused by loss of Shot (or Eb1; see Ref. 4) cannot be rescued with Shot-PE variants that lack Ctail or Eb1-binding SxIP motifs (i; see Figure 7c) or the CH1 domain (j); absence of the same domains in shot V104 (k) or shot kakP2 (l) does not cause MT curling. We propose that the presence of the Shot-PH isoform (faintly shown in a, b, k, and l) protects axons against loss of actin or F-actin/MT/Eb1 guidance mechanism, that is, conditions which cause severe curling in the other experimental settings (c, f, g, i, and j)  Figure 5 that F-actin/Eb1/MT guidance is not the only mechanism through which Shot contributes to MT bundle maintenance. For example, Shot might work through further isoforms beyond Shot-PE (the only isoform shown so far mediating MT bundle maintenance; Figure 5e,f,h-j). To test this possibility, we used Shot-deficient mutant neurons in which the MT curling phenotype was rescued by Shot-PE::GFP so that Shot-PE was the sole isoform present in these neurons (Figures 5e and 6e). These neurons were normal in appearance. However, when treated with CytoD, strong MT curling was induced (Figures 6f and 5f). This indicated that Shot-PE-mediated F-actin/Eb1/MT guidance is not sufficient when F-actin is removed from axon shafts, and might suggest that these neurons lack some additional bundle-maintaining functions of Shot that are actin-independent.
In contrast to Shot-PE::GFP, the Shot-PC::GFP or Shot-PE-ΔCtail::GFP variants failed to rescue MT curling in shot mutant neurons (Figure 5i,j), arguing that the CH1 and Ctail domains of Shot are essential for F-actin/Eb1/MT guidance. In shot kakP2 and shot V104 mutant animals, these two domains are specifically missing from all isoforms of endogenous Shot proteins (details in Figures 1a and 7; Bottenberg et al., 2009;Gregory & Brown, 1998). Therefore, these two alleles are expected to eliminate the endogenous F-actin/Eb1/MT guidance function but potentially leave other functions of Shot intact (''PH'' in Figure 5k,l).
When analyzed in whole embryos, both mutant alleles clearly caused hypomorphic loss-of-function mutant phenotypes: shot kakP2 strongly affected the nervous system (Bottenberg et al., 2009;Gregory & Brown, 1998), whereas shot V104 defects seemed to restrict to non-neuronal tissues ( Figure S6). We next cultured primary neurons from these embryos and measured the degree of MT curling in the axon shaft, which is the area where the guidance mechanism is expected to make its prime contributions. In shot 3 null mutant neurons used as positive controls, severe MT curling occurred along axon shafts; in contrast, shot V104 mutant neurons showed no obvious phenotypes, and shot kakP2 revealed only a trend toward curling (Figure 8a-d,f). For shot V104 mutant neurons, we repeated the experiment, but this time culturing them on concanavalin A which is a more challenging condition causing greater mechanical strain . When challenged this way, shot V104 mutant neurons displayed robust MT curling. This suggests that loss of the F-actin/MT/Eb1 guidance mechanism weakens the overall machinery of MT bundle maintenance: under modest conditions, its absence can be F I G U R E 7 The shot V104 breakpoint removes the Ctail. (a) View of the 2R polytene chromosome (Lindsley & Zimm, 1992) indicating the mapped breakpoint in 50C (orange arrow) and potential sites of the second breakpoint in the centromeric region of 2R (orange arrowheads) suggested by the mapping positions of several clones with matching sequences (when using the BLAST function in flybase.org and the blue sequence in (b) as query); clones with matching sequences: DS03708 (42A4-42A5), BACR04E10 (41C-41D), BACR07J16 (41C-41C), BACR05A24 (41C-41D), BACR05A24 (41C-41D), and BACR03D04 (40D-40D). (b) Alignment of the wild-type and V104 mutant genomic sequences of shot indicating the breakpoint (yellow arrow) in position 13,868,412 (primary assembly 2R: 13,864,237-13,925,503 reverse strand) and the newly fused sequence in shot V104 (blue) likely derived from the other end of the inversion that would usually be situated near the position of the second breakpoint (orange arrows in a). (c) Schematic of the Shot-PE protein (FBtr0087618) drawn to scale and indicating domain/motif borders (colored numbers below; compare Figure 1a) as well as exon borders (stippled vertical lines, gray numbers, exon numbers indicated between lines); the V104 breakpoint is situated in intron 22/23. (d) The predicted V104 protein is truncated behind the GRD (yellow arrow) potentially reading into intronic sequences (gray). Comparison of the V104 sequence at the breakpoint (highlighted yellow) with sequences of GRDs from normal Shot and other GRD-containing proteins (listed in gray; taken from Alves-Silva et al., 2012) strongly suggest that the truncation does not affect the final α-helix and amino acid changes occur behind the GRD. (e-g) Ventral nerve cords of stage 16 embryos (cx, cortex containing cell bodies; np, neuropile containing synapses and as-/descending tracts; dashed yellow lines demarcate outlines of the ventral nerve cord; t, trachea) stained with the Shot-C antibody against the C-terminal part of the spectrin masked by other bundle-maintaining functions of Shot, but this becomes insufficient when the mechanical challenge is increased.

Roles of Shot in axonal MT regulation involve various isoform-specific actin-dependent and -independent functions
Spectraplakins are well conserved across the animal kingdom; they are essential cytoskeletal regulators in neurons, linked to severe MT curling in mammals and Drosophila alike (Dalpe et al., 1998;Eyer et al., 1998;Sánchez-Soriano et al., 2009;Voelzmann et al., 2017). Many mechanistic insights were gained using Drosophila Shot as a model, and Factin/Eb1/MT guidance has emerged as a central theme that is consistent also with roles in non-neuronal cells (Kodama et al., 2003;Ricolo & Araujo, 2020). Here, we refined our understanding of Shot's actin dependency during MT regulation, while also proposing the co-existence of actinindependent functions involved in MT bundle promotion.

Shot's roles in spool formation are regulated by F-actin
Our findings suggest that F-actin is an important instructor of Shot's MT-regulating roles. For example in GCs, Shot is an essential regulator of spool formation in an F-actin-dependent manner: (1) spools can be suppressed when depleting Factin (LatA, CytoD; Figures 2d and S1b; Sánchez-Soriano et al., 2010), (2) when changing the properties of F-actin networks (CK666; Figure 2D), or (3) when changing Shot's actin-binding properties as observed with Shot-PC, Shot-PE-ΔABD, Shot-PE-ΔCH2, Shot-PE-Moe, and Shot-PE-Life ( Figure 1l). In contrast, F-actin networks of the axon shaft are far less prominent (Xu et al., 2013), which seems sufficient for F-actin/Eb1/MT guidance but not enough to induce prominent changes to MT bundles even when overexpressing Shot-PE. In contrast, Shot-PE-Life was able to induce abnormal MT bundle split in the shaft (Figures 3 and S4), suggesting that increased F-actin affinity is sufficient to tip the balance in an F-actin-sparse environment and change the MT-regulating behavior of Shot.
repeat rod (Figure 1a; Strumpf & Volk, 1998); staining reveals the presence of protein in wild-type (e), absence in homozygous shot null mutant embryos (f) and presence in hemizygous shot V104/MK1 mutant embryos where reduced expression is due to the absence of one gene copy (V104 is over the MK1 deficiency); scale bar in E represents 20 μm in (E-G) F I G U R E 8 Phenotypes of shot kakP2 and shot V104 mutant primary neurons. (a-d, g, h) Images of neurons at 6-8 h in vitro (HIV) of different genotypes (wt, wild-type; 3, shot 3/3 ; kakP2, shot kakP2/kakP2 ; V104, shot V104/Df(MK1) ) cultured on glass (a-d) or ConA (g and h) and stained for tubulin (green), actin (magenta) or GFP (blue); grayscale images on the right show only the tubulin channel; asterisks indicate cell bodies, arrowheads the tips of axons, dashed white lines demarcate axon shafts, curved arrows areas of MT curling; scale bar in (a) represents 20 μm in (a-d) and 10 μm in (g) and (h). (e, f, i, and j) Quantifications of axon length (e and i) and microtubule (MT) curling (measured in MT disorganization index [MDI]; f and j), both normalized to wild-type controls (red dashed line); numbers of neurons analyzed are indicated in orange, mean ± SEM in blue and results of Mann-Whitney rank sum tests are shown in black/gray Taken together, these experiments suggest that proper Shot function requires well-balanced interaction with F-actin networks, and the spectacular phenotypes we observe with Shot-PE-Life (Figure 3 and Figure S4) suggest that our findings can be turned into new genetic tools to investigate how changes in cytoskeletal organization may impact on neuronal architecture, dynamics, and even physiology.
Our experiments with Shot-PE-Life have demonstrated a clear F-actin-dependence of the induced MT phenotypes (Figure 3f,g). They also suggested that this construct was able to induce ectopic F-actin in axon shafts (Figure 3 and Figure  S4), potentially reflecting mutual regulation mediated through Shot. This may involve known roles of the Shot C-terminus in promoting F-actin nucleation : for example, the strong localization of Shot-PE-Life along axon shafts might trigger a positive feedback loop by nucleating more F-actin which then enhances Shot-PE-Life localization.
In normal Shot-PE, direct binding through the CH domains does not appear sufficient to trigger changes in spool-inducing functions of Shot, and also the plakin domain appears functionally involved. To our knowledge, the only plakin domainbinding factors reported so far are transmembrane adhesion factors including integrins and collagen XVII at mammalian hemidesmosomes (Aumailley et al., 2006) and potentially the N-CAM homologue Fasciclin II in Drosophila neurons (Voelzmann et al., 2017). Since the localization of such adhesion factors is dependent on F-actin (Woichansky et al., 2016), they might provide a potential second route through which Factin can influence Shot activity.
In summary, we have built a case for regulatory impacts of F-actin networks on Shot function which, in turn, trigger MT network changes that impact on axon growth; this is best exemplified by the negative correlation between spool formation and axon growth (Figure 4a; Dent et al., 1999;Sánchez-Soriano et al., 2010).

Shot displays prominent F-actin-independent roles in axons
Shot also plays major roles in maintaining MT bundles in axon shafts. We confirmed here the importance of F-actin/Eb1/MT guidance for parallel bundle arrangements (Alves-Silva et al., 2012;Hahn et al., 2021;Figures 5, 6, and 8). We believe that roles of F-actin in this context are merely permissive with little regulatory potential because F-actin networks in axon shafts appear sparse and far less dynamic compared to GCs. However, these permissive roles are nevertheless important as clearly demonstrated by MT curling in shot mutant neurons which correlates with drastic growth reduction. We suggest that curling in the axon shaft diverts MT polymerization away from the axon tip, thus reducing their likelihood of reaching the GCs and contribute to axon growth events.
In addition to F-actin/Eb1/MT guidance, we also presented strong arguments for additional functions of Shot in MT bundle maintenance that are independent of this form of crosslinkage (Figure 5b,k,l). Considering the enormous importance that MT bundles have for the long-term survival of axons Prokop, 2020), it would make biological sense to have redundant mechanisms to maintain these bundles and prevent axonopathies (Prokop, 2021).
In our view, the best candidate to mediate F-actinindependent functions of Shot is the unique Shot-PH isoform. Shot-PH is highly expressed in the nervous system, has a C*-type N-terminus (non-F-actin-binding like Shot-PC; Figure 1a), and stands out as the only isoform containing a large central PRR ( Figure S7b; flybase.org reference: FBgn0013733; Hahn et al., 2016;Röper & Brown, 2003;Voelzmann et al., 2017).
PRRs are conserved in mammalian dystonin and ACF7/MACF1 (Voelzmann et al., 2017), but very little is known about their role or potential binding partners. The PRR of Drosophila Shot plays regulatory roles at epithelial adherens junctions through unknown mechanisms (Röper & Brown, 2003). In mammals, the PRR-containing isoform MACF1b was shown to associate with the Golgi (Lin et al., 2005). However, we struggle to imagine mechanisms through which Golgi-related mechanisms in the cell body could maintain MT bundles in the axon under conditions where F-actin/Eb1/MT guidance is abolished. In our view, investigating the potential roles and mechanisms of PRRs in axons would therefore have great potential to deliver new mechanisms that can advance our understanding of axon maintenance and architecture (Prokop, 2020).
As a first step to study PRRs, we generated flies carrying a CRISPR/Cas9-mediated PRR deletion. Unfortunately, shot ΔPRR mutant flies displayed unexpected splicing defects resulting in a strong loss-of-function mutant allele (details in Figure S7); while being potentially interesting for molecu-lar geneticists that work on splicing mechanisms, this allele was unsuitable for our purposes. An alternative strategy could be to identify PRR-binding or PRR-associating proteins (Lin et al., 2021), and then use versatile Drosophila genetics in combination with our culture model (Prokop et al., 2013) to establish their potential involvement in bundle maintenance.
Among the PRR-interacting proteins, we would expect to find also Eb1-binding proteins or even Eb1 itself (note that PRR contains a potentially Eb1-interacting SNLP motif as similarly found in the Ctail; Figure 7c; Honnappa et al., 2009). A link from the PRR to Eb1 could explain an important conundrum posed by the current data: loss of Eb1 causes MT curling, but the deletion of the Eb1-binding Ctail from all Shot isoforms does not (shot V104 ; Figures 5g vs. 5k, 7, and 8d-f)the PRR might be the missing puzzle piece establishing alternative links to Eb1.
Taken together, we propose a system of redundant Shotmediated mechanisms that promote axonal MT bundle architecture-in addition to other factors expected to be involved, such as classical MAPs or mitotic kinesins (Guha et al., 2021;Hahn et al., 2019;Prokop, 2020). Such robust redundancy makes sense when considering the enormous importance of these MT bundles for axonal longevity (Prokop, 2021). We believe that the study of Shot-PH can establish new investigative paths toward a more profound understanding of axon architecture, thus bridging a gap in the field that may provide important explanations for a wide range of axonopathies and new avenues for their treatment.

Drosophila primary neuronal cell culture
Neuronal cell cultures were generated as detailed elsewhere Voelzmann & Sánchez-Soriano, 2021). Embryos were dechorionated for 1.5 min in 50% domestic bleach, correct stages (usually stage 11; Campos-Ortega & Hartenstein, 1997), and genotypes were selected under a fluorescent dissecting microscope, transferred to sterilized centrifuge tubes containing 100 μl of 70% ethanol, washed in sterile Schneider's medium containing 20% fetal calf serum (Schneider's/FCS; Gibco), and, eventually, homogenized with micropestles in 1.5 ml centrifuge tubes containing 21 embryos per 100 μl dispersion medium . They were left to incubate for 4 min at 37˚C. Dispersion was stopped with 200 μl Schneider's/FCS, cells were spun down for 4 min at 650 g, supernatant was removed, and cells were resuspended in 90 μl of Schneider's/FCS; 30 μl drops were placed in culture chambers and covered with cover slips. Cells were allowed to adhere to cover slips for 90-120 min either directly on glass or on cover slips coated with a 5 μg/ml solution of concanavalin A, and then grown as a hanging drop culture at 26˚C usually for 6-8 h.
Transfection of Drosophila primary neurons was executed as described previously (Qu et al., 2019). In brief, 70-75 embryos per 100 μl dispersion medium were used. After the washing step and centrifugation, cells were resuspended in 100 μl transfection medium (final media containing 0.1-0.5 μg DNA and 2 μl Lipofectamine 2000 [L2000; Invitrogen]), incubated following manufacturer's protocols (Thermo Fisher, Invitrogen) and kept for 24 h at 26˚C. Cells were then treated again with dispersion medium, resuspended in culture medium and plated out as described above.

Drug application and immunohistochemistry
For drug treatments, solutions were prepared in cell culture medium from stock solutions in DMSO. Cells were treated for 4 h with 200 nM latrunculin A (Biomol International), 0.4 μg/ml cytochalasin D (Sigma), or 100 nM CK666 (Sigma), respectively. For controls, equivalent concentrations of DMSO were diluted in Schneider's medium.

Stage 17 embryo dissections
Dissection of late stage 17 embryos (stages according to Campos-Ortega & Hartenstein, 1997) was carried out as described in great detail elsewhere (Budnik et al., 2006). In brief, embryos were dissected flat in PBS on Sylgard-coated cover slips with the help of sharpened tungsten needles and Histoacryl glue (Braun, Melsungen, Germany), followed by 1 h fixation in 4% paraformaldehyde, 1 h wash in PBT, and the same histochemical staining steps as mentioned above using the following antibodies: anti-FasII (1D4 2F3, DSHB; mouse, 1:20; Van Vactor et al., 1993), anti-GFP (see above), and anti-Synaptotagmin (rabbit polyclonal; 1:1,000; Littleton et al., 1993). Embryos were cut out from the glue using razor blade splinters or the tungsten needles and embedded in glycerol.

Imaging and image analysis
Standard imaging was performed with AxioCam 506 monochrome (Carl Zeiss Ltd.) or MatrixVision mvBlueFox3-M2 2124G digital cameras mounted on BX50WI or BX51 Olympus compound fluorescent microscopes. Measurements from images were carried out in the fixed preparations using ImageJ (segmented line and freehand selection tools). Only neurites at least twice the length of the soma diameter were analyzed using α-tubulin staining and measuring from the edge of the cell body to the tips of the axons (excluding MTs in filopodia); in cases where neurites branched, the longer branch was measured, in cases where two neurites extended from a single cell, the longer value was taken. The degree of disorganized MT curling in axon shafts was established either as binary readout (% of neurons with disorganization) or as "MT disorganization index" (MDI) described previously (Qu et al., 2019; in short: the area of disorganized curling was measured with the freehand selection tool in ImageJ; this value was then divided by axon length (see above) multiplied by 0.5 μm (typical axon diameter, thus approximating the expected area of the axon if it were properly bundled); in this study, MDI measurements were restricted to the axon shaft, that is, from the cell body to the base of GCs (white dashed lines in Figures 6 and 8). Filopodia numbers were counted per neurite. GCs containing looped MT bundles (spools) were classified according to previous publications (Sánchez-Soriano et al., 2010). Graphpad Prism was used to describe data, perform statistical tests, and generate final graphs. Data were usually not normally distributed, and the median was determined for axon length; since MDI measurements contain many zero-value data, the mean and standard error of the mean (SEM) had to be used to obtain meaningful numbers. For statistical analyses, the chi-square test was used when comparing percentages, Kruskal-Wallis one-way analysis of variance (ANOVA) test to compare groups, and Mann-Whitney rank sum tests (indicated as p MW ) to compare pairs of data. For the correlation, r-and p-values were determined via nonparametric Spearman correlation analysis (tests showed that data are not distributed normally). The data used for our analyses will be made available on request from the authors.

Electron microscopy
Procedures followed protocols published in detail elsewhere (Budnik et al., 2006 (Warming et al., 2005). The GalK cassette was originally inserted into M-6-attB-UAS-1-3-4 shot-RE-borne shot-RE by using similar recombineering steps with GalK which had been amplified with primers that added the same homology arms as mentioned above.
The completed constructs in M-6-attB-UAS-1-3-4 vector were amplified in Epi300 competent cells (EpiCentre) in LB-Chloramphenicol medium, adding CopyControl solution (EpiCentre) 2 h before the miniprep. Amplified constructs were used to generate transgenic flies (outsourced to Best-Gene, Chino Hills, CA 91709, US) using PhiC31-mediated site-specific insertion using a specific attB landing site on the third chromosome (PBac{y + -attP-3B}CG13800 VK00031 ; Bloomington Stock Center #9748; Alves-Silva et al., 2012). This same landing site was used for all constructs to avoid position effects and achieve equal expression levels of all constructs (Bischof et al., 2007).

qRT-PCR analysis of shot ΔPRR mutant embryos
For RNA isolation, at least 10 Drosophila third instar larvae were placed in Trizol (Invitrogen) and homogenized using a pestle. Total RNA was isolated using the NucleoSpin RNA II kit (Macherey & Nagel), and RNA concentration was analyzed via a NanoDrop spectrophotometer (Thermo Scientific). For first strand cDNA synthesis, 500 ng of total RNA was transcribed using the QuantiTect RT Kit (Qiagen). Real-time PCR was performed with 1 μl cDNA per reaction using the Power SYBR Green PCR Master Mix (ThermoFischer Scientific) as detection dye. Experiments were performed with the BioRad C1000 Thermal Cycler. cDNA samples were run in triplicate, and the average CT was used to analyze the expression levels via the −2ΔΔCT method. Experiments were repeated with independently isolated RNA samples. Actin 5C (Act5C, act) and Ribosomal protein L32 (RpL32, rp49) were used as reference genes. Expression analysis was performed using BioRad C1000 System software and GraphpadPrism. The following oligonucleotides were used for real time PCR analysis ( Figure S7a

Sequencing the shot V104 breakpoint
The chromosomal breakpoint of shot V104 was described to be in a 373 bp region between 73,398 and 73,771 bp of the shot locus (Strumpf & Volk, 1998). We used an inverse PCR approach to determine the exact chromosomal breakpoint of shot V104 . For this, genomic DNA of 200 homozygous shot V104 embryos was isolated (Berkeley Drosophila Genome Project protocol; https://www.fruitfly.org/about/ methods/inverse.pcr.html) and restricted with Sau96I. The restricted DNA was purified, diluted 10:1, and ligated into circular fragments. Using primer pairs designed to face toward the unknown region covering the breakpoint (fw: CCT-GCTTTCAAACTAACATCCTGC; rev: CTGGCTGAATG-GCAATTAAAGG), the circular DNA fragment containing the shot V104 breakpoint region was amplified using a High Fidelity PCR Kit (Eppendorf and Roche). PCR products were gel-extracted, cloned into pDrive (Qiagen), and sequenced. The sequencing of one inverse PCR fragment showed a perfect alignment with wild-type genomic DNA until bp73.681 followed by an adenine and thymine-rich region. Using BLAST (flybase.org), we identified this region as part of the centromeric region of chromosome 2R (Figure 7). The breakpoint was confirmed via PCR (Figure 7) using the following sequence-specific primers: • forward sense primer: TCTACGCTTGCGCTGCC-CGCTCGCC (binding the wild-type shot region 100 bp upstream of the breakpoint); • reverse antisense wt1: TTTGTACGCATTGGCATGGCA-GATG (binding the wild-type region before the breakpoint); • reverse antisense wt2: GGCAGATGCACAGATGCATT-TATATACGC (binding the wild-type region directly after the break point); • reverse antisense mutant 1: TGTTAGTTCTTATACAA-GAAGATTCAATAAATAAAAGC (in the putative new shot V104 sequence after the breakpoint).

A C K N O W L E D G M E N T S
Work reported here was made possible through the support of A.P. by the BBSRC (BB/C515998/1, BB/I002448/1, BB/M007553/1) and the Wellcome Trust (084561/Z/07/Z, 077748/Z/05/Z, 087820/Z/08/Z, 092403/Z/10/Z), the BBSRC to Natalia Sánchez-Soriano (BB/R018960/1) and Leverhulme Trust to Ines Hahn (ECF-2017-247), and by parents as well as the Faculty of Life Sciences to Yue Qu. The Manchester Bioimaging Facility microscopes used in this study were purchased with grants from the BBSRC, The Wellcome Trust and The University of Manchester Strategic Fund. The Fly Facility has been supported by funds from The University of Manchester and the Wellcome Trust (087742/Z/08/Z).
Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study.

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

AU T H O R C O N T R I B U T I O N S
Cell culture experiments, image analysis, and cloning of constructs: Yue Qu and Kriti Gupta. Embryo analyses: Juliana Alves-Silva. Breakpoint analysis and CRISPR design: Ines Hahn. Cloning of constructs, qRT-PCR, and developing transfection procedure: Jill Parkin. Cell culture experiments and image analysis: Natalia Sánchez-Soriano. Electron microscopy, embryo analyses, image analysis, and writing the manuscript including figure compositions: Andreas Prokop.

D A T A AVA I L A B I L I T Y S T A T E M E N T
The data used for our analyses will be made available on request from the authors.