Drosophila Src regulates anisotropic apical surface growth to control epithelial tube size

Abstract

Networks of epithelial and endothelial tubes are essential for the function of organs such as the lung, kidney, and vascular system. The sizes and shapes of these tubes are highly regulated to match their individual functions. Defects in tube size can cause debilitating diseases such as polycystic kidney disease (PKD) and ischemia^1,2. It is therefore critical to understand how tube dimensions are regulated. Here we identify the tyrosine kinase Src as an instructive regulator of epithelial tube length in the Drosophila tracheal system. Loss-of-function Src42 mutations shorten tracheal tubes while Src42 over-expression elongates them. Surprisingly, Src42 acts distinctly from known tube size pathways and regulates both the amount of apical surface growth and, with the conserved formin dDaam, the direction of growth. Quantitative 3-D image analysis reveals that Src42 and dDaam mutant tracheal cells expand more in the circumferential than the axial dimension, resulting in tubes that are shorter in length – but larger in diameter – than WT tubes. Thus, Src42 and dDaam control tube dimensions by regulating the direction of anisotropic growth, a mechanism that has not previously been described.

The Drosophila tracheal system is a gas exchange organ that arises from clusters of epithelial cells that invaginate, branch, and interconnect to form a network of tubes. After network formation, tracheal lumens dramatically increase their diameters and lengths³. Importantly, these expansions result exclusively from changes in cell shape and apical surface area since tracheal cell number does not change during expansion^3,4.

Multiple cellular processes are known to control tracheal tube size. Overexpression of the apical determinant Crumbs (Crb) increases both the length and diameter of late stage tracheal tubes⁵ and loss of the basal polarity protein Scribbled (Scrib) increases tube length^5,6. Mutations in planar cell polarity (PCP) genes also cause modest tube over-elongation resulting from increased apical cell surface⁷. The best-characterized size control pathway involves the apical extracellular matrix (aECM) whose assembly depends on a cell-cell junction termed the septate junction (SJ). After tracheal metameres fuse, a chitin-based aECM is secreted apically into the tracheal lumen^8–10. Putative chitin deacetylases Verm and Serp are secreted into the lumen to regulate aECM organization and are required to prevent over-elongation of tracheal tubes^11,12. Basolateral SJs direct apical secretion of Verm and Serp via an uncharacterized pathway¹². Consequently, mutations in almost all SJ components also cause tube over-elongation^6,13–15. To date, mutations that dramatically shorten the length of late-stage embryonic tracheal tubes have not been reported. Further, with the possible exception of Crb, the currently identified size-control proteins act as “permissive” rather than “instructive” factors, since their over-expression does not cause phenotypes opposite to their loss-of-function phenotypes^12,14,16.

Src-family kinases (SFKs)¹⁷ have been implicated in controlling the size of mammalian epithelial and endothelial tubes, though the mechanisms by which they do so have not been elucidated^18,19. We therefore examined the tracheal system of Drosophila embryos homozygous for null mutations in either of the two known SFKs, Src42 and Src64. Neither zygotic nor maternal/zygotic (M/Z) Src64 mutants showed tracheal defects or genetic interactions with a mutation in ATPα that elongates tracheal tubes¹³, and Src64 was not detected in tracheal cells using immunohistochemistry (Fig. S1 online). In contrast, by late embryogenesis (stage 16), zygotic loss of Src42 caused the dorsal trunk (DT), the largest and most easily analyzed tracheal tube, to be ~27% shorter than WT controls (Fig. 1b, d, q) and to have a “stretched” appearance in which the angle of the anterior transverse connective (TC) 1 relative to the DT was significantly increased in Src42 mutants over WT controls (Fig. 1e, f) and the normal ventral curve of the posterior DT metamere 10 was absent (Fig. 1g, h). Measurements revealed that Src42 elongation defects become apparent after tracheal metameres fuse midway through embryogenesis (Fig. 1a,c; 2jj) as the DT begins elongating³. Importantly, the elongation defect of Src42 mutants is independent of cell number, since Src42 and WT DTs have the same number of cells (metamere 8: WT = 24.1 ± 1.6, Src42 = 24.3 ± 2.1, p = 0.24, n = 8). These observations identify Src42 as one of the first genes required for increasing the length of late-stage embryonic trachea.

Figure 1. Src42 acts autonomously to control tracheal tube length.

(a–d) The dorsal trunk (DT, arrow) of Src42 null mutant embryos appears normal at early embryonic stage 14 (a, c), but is dramatically shorter than WT by stage 16 (b, d). Asterisks indicate endpoints for DT measurements. Lumens in a–o and p visualized by staining for 2A12 antigen. (e, f) DT shortening increases the angle between the anterior transverse connective (TC) and the DT in Src42 mutants (e, f) (n = 5, p = 6 × 10⁻⁹). (g, h) The angle of the posterior ventral curve is also increased in Src42 mutants (g, h) (n = 5, p = 1 × 10⁻⁵). (i, j) Src42 kinase activity is required for DT elongation, since tracheal-specific expression (btl-Gal4²⁴) of a WT Src42 transgene (Src42^WT) (i) but not a kinase-dead form of Src42 (Src42^KM) fully rescues Src42 DT elongation (j). (k–n) In WT embryos, Src42 is found throughout the cell periphery of tracheal cells (k). Src42 mutants completely lack Src42 labeling. (l). The Src42^KM and Src42^WT transgenes are expressed at similar levels in the trachea of Src42 mutants (m, n). (o) Src42^KM acts as a dominant negative in the tracheal system, as its over-expression inhibits DT elongation. (p) Conversely, over-expression of Src42^WT in the tracheal system of Src42 heterozygotes over elongates tracheal tubes. (q) Quantification of DT length in Src42 rescue experiments reveals statistically significant differences between Src42^KM and Src42^WT. (r) Allelic series using loss- and gain-of-function alleles and transgenes reveals that tracheal length correlates with Src42 levels. Note that over-expression of Src42^KM results in tracheal tubes that are comparable in length to null alleles. For q and r, error bars represent standard deviation, all p values are determined using Student’s t-tests, and n = 5 for all samples. Scale bars: 20 μm for a–d, i, j, o, p; 10μm for e–h; 5μm for k–n.

Figure 2. Src42 acts independently of the aECM-based tube size control pathway.

(a–c) Activated Src42 (pSrc) colocalizes with DE-cadherin (DE-cad) at the adherens junction (AJ). (d–f) pSrc shows some partial overlap with Crb at the subapical membrane. (g, h) Activated Src localizes apical to the septate junction (SJ), which is labeled with Coracle (Cor), a canonical SJ marker. (i, j). Src42 mutants show essentially no pSrc labeling, indicating that the pSrc antibody is specific for activated Src42. (k, l) Over-expression of Src42 results in a dramatic increase in pSrc. (m–r) Labeling for phospho-tyrosine (pTyr) shows that most tyrosine-phosphorylated proteins also localize apical to SJs (m, n). Src42 is required for much of the tyrosine phosphorylation at the apical membrane (m–p) and Src42 over-expression increases apical pTyr (q, r). (s–v) Src42 is epistatic to the aECM/SJ pathway, since Src42 mutants fully suppress tracheal over-elongation in aECM/SJ mutants, such as nrv2. (w–z) Src42 mutations do not disrupt SJ organization (note localized red Cor staining at the apical region of the lateral membranes, arrowheads in w and x) or Verm secretion (green lumenal staining) and suppress nrv2 tube elongation without restoring SJ organization (note diffuse Cor staining) or Verm secretion (note the lack of Verm staining) (y, z). Dashed lines mark the apical surface. (aa) The DT of Src42, nrv2 double mutants are the same length as Src42 single mutants. The length of the DT between TC5 and TC10 was measured because the over-elongation of aECM/SJ mutants is most apparent in the posterior DT. Error bars represent standard deviation. p values are from Student’s t-tests. For all samples, n = 5. (bb–ee) nrv2 null mutants do not show obvious changes in the levels or localization of pSrc (bb, cc) Src42 (dd), or pTyr (ee), indicating that the aECM does not directly regulate Src42. See Figure 2 for WT comparisons. (ff–ii) Over-expression of Src42 in a nrv2 background results in a loss of lumenal 2A12 and tracheal tubes that are cystic (ii). The enhanced phenotype indicates that Src42 and nrv2 are unlikely to function in the same genetic pathway. Dashed lines mark the apical cell surface. (jj) Measurement of DT length in WT, Src42, and nrv2 animals at two hour intervals shows that the DT of WT, nrv2, and Src42 are the same length at 10 h, just after tracheal metameres fuse (p = 0.46 for WT and Src42, p = 0.78 for WT and nrv2). However, by 12 h, Src42 DTs are shorter than WT (*, p = 0.044), while nrv2 mutants are not different from WT (p = 0.69). nrv2 mutants do not show significantly increased tube length compared to WT embryos until 12 h. Error bars represent standard deviation. p values are from Student’s t-tests. For hour 14 (stage 16), n = 5 for each genotype. For all other time points, n = 3 for each genotype. Scale bars: 10μm for a–z, bb–ii.

As SFKs can mediate cellular functions independent of their tyrosine kinase activity^20,21, we expressed either a WT (Src42^WT)²² or a kinase-dead form of Src42 (Src42^KM)²³ specifically in the trachea²⁴ of Src42 mutants. Src42^WT could completely rescue the DT elongation defect of Src42 mutants while Src42^KM could not, despite similar levels of expression (Fig. 1i–n, q). Furthermore, over-expression of Src42^KM, which acts as a dominant negative in other systems^23,25, shortened DTs (Fig. 1o, r). Together, these results indicate that the kinase activity of Src42 is required autonomously in trachea for DT elongation.

Over-expression of Src42^WT in the tracheal system resulted in severe DT fusion defects that precluded analysis of DT length (data not shown). However, over-expressing Src42^WT in a Src42 heterozygote did not disrupt fusion and increased tracheal DT length by ~8% (Fig. 1p, r). Combinations of loss-of-function Src42 alleles and over-expression constructs produced an allelic series in which the tracheal length ranged from ~27% too short to ~8% too long (Fig. 1r). Thus, Src42 acts instructively to regulate tube size.

Because the kinase activity of Src42 is required for regulating tracheal length, we determined the subcellular localization of activated (phosphorylated Tyr-400) Src42 (pSrc) in tracheal cells. As reported previously²³, while endogenous Src42 localizes throughout the plasma membrane of all DT cells (Fig. 1k), activated Src42 largely co-localizes with DE-cadherin (DE-cad) at the adherens junction (AJ) (Fig 2a–c, arrows) and partially co-localizes with Crb at the subapical membrane (Fig 2d–f, arrowheads). Activated Src42 does not co-localize with SJs (Fig. 2g, h). Notably, a large fraction of phosphotyrosine (pTyr) immunostaining that normally localizes to the AJ and subapical membrane (Fig. 2m, n) is absent in Src42 mutants (Fig. 2o, p), while Src42 over-expression results in a dramatic increase in pTyr content (Fig 2q, r). Thus, Src42 appears to be either directly or indirectly required for the majority of tyrosine phosphorylation in the tracheal system. Given the critical role pTyr is thought to play in junctional and cytoskeletal assembly and turnover^23,26,27, Src42 mutations have unexpectedly mild effects on tracheal morphology.

The elongated tracheal phenotype caused by Src42 over-expression is highly similar to that caused by polarity, SJ, and aECM loss-of-function mutations (Fig. 2u and ^{5,6,11,12,14,16}). To investigate whether Src42 acts in these pathways, we performed genetic epistasis experiments using null alleles of Src42 and several polarity, SJ, and aECM genes. Unfortunately, double mutant combinations of Src42 and the polarity genes scrib, yrt, and lgl resulted in severe defects in embryogenesis, making clear analysis of DT length difficult. However, the short-tracheal phenotype of Src42 was epistatic to all tested SJ and aECM mutations (Fig. 2s–v, aa, and Fig. S2m–p and Table S1a online). Importantly, Src42 suppressed over-elongation without improving aECM or SJ organization (Fig. 2w–z and S2a–p online). Consistent with this, Src42 over-expression resulted in DT over-elongation without significantly disrupting SJ integrity or aECM organization (Fig. S2q–t online). These results indicate that Src42 acts either downstream of the SJ/aECM tube size control pathways or independently in a parallel pathway.

To distinguish between “downstream” and “in parallel”, we assessed Src42 activity in SJ, aECM, and polarity mutants. If Src42 acts downstream of the SJ/aECM or polarity pathways, loss-of-function mutations in SJ, aECM or polarity genes should increase Src42 activity. However, we were unable to detect any changes in the levels or localization of pSrc, pTyr, or total Src42 in any tube expansion mutants (Fig. 2bb–ee and S2u–bb online), although elevated pSrc was readily detected when Src42 was over-expressed in the tracheal system (Fig. 2k, l). Moreover, over-expression of Src42 in a nrv2 null background (an essential SJ gene) resulted in a dramatically enhanced tracheal phenotype consisting of loss of lumenal 2A12 and a cystic lumen (Fig. 2ff–ii) suggesting that Src42 and nrv2 do not function in the same linear pathway. Consistent with this, Src42 and nrv2 act at separable developmental times. Src42 is required to initiate elongation between hours 10 and 12 of development, while nrv2 is required to restrict tube length after hour 12 (Fig. 2jj).

We also asked whether Src42 acts in the canonical PCP pathway to control tracheal length⁷. However, in contrast to Src42, mutations in the PCP genes dsh, fz, and sano cause elongated trachea (Fig. S3m and⁷), and the short phenotype of Src42 was completely epistatic to all tested PCP mutants (Table S1a online). Further, sano and dsh mutants did not show any changes in the levels of activated Src42 (Fig. S2u–x online and data not shown) and Src42 mutations did not alter the planar polarization of wing hairs, thorax bristles, or denticle belts (Fig. S3a–l online). Specific expression of a dominant negative Src42 construct in the thorax also did not affect planar polarization (Fig. S3h online) and caused only very subtle effects when expressed in a specific wing region (Fig. S3d′ online). Finally, while Dsh and other PCP genes have a planar polarized localization in other epithelia such as the epidermis, their distribution appears uniform in the trachea (Fig. S3q–t and⁷). Thus, Src42 acts either downstream or in parallel to PCP genes to control tube size. Together, these data indicate that Src42 acts in a pathway separate from known tracheal tube size control genes.

To understand the cellular functions of Src42 in tube size control, we investigated the morphology of WT and Src42 DT cells. We reasoned that short trachea could result either from a general failure to expand apical surface or from a failure to direct surface expansion anisotropically along the length of the tube, which would result in shorter but fatter tubes that were inappropriately expanded along the circumferential axis (Fig. 3a). To distinguish between these models, we measured both the length and diameter of WT and Src42 metamere 8 lumens. Interestingly, while Src42 lumens were ~37% shorter than WT controls (Fig. 3b,c,d), they were also ~11% larger in diameter (Fig. 3e). We calculated that the total apical surface area (ASA) of Src42 metameres is reduced by ~30% (Fig. 3f, see Materials and Methods), indicating that Src42 is required for increasing apical surface. However, the increased lumen diameter of Src42 mutants indicates that Src42 is also required for directing available surface expansion along the axial dimension.

Figure 3. Src42 is required for orienting anisotropic apical surface expansion along the longitudinal axis of tracheal tubes.

(a) Two models that could account for shortened DTs in Src42 mutants. Model 1, apical surface expansion is reduced, preventing anisotropic growth along the axial dimension. Model 2, apical surface expansion is misdirected along the circumferential dimension, producing shorter but fatter tubes. (b, c) Confocal projections of metamere 8 in WT (b) and Src42 (c) embryos labeled for DE-cadherin (DEcad) show that Src42 cells appears “shorter” than WT along the longitudinal axis of the tube. Arrowheads mark metamere ends. Scale bar: 5μm for b, c. (b′,c′) Cell tracings of b, c. (d,e) The length of metamere 8 (fusion cell to fusion cell) is shorter in Src42 mutants than in WT, but diameter is increased. (f) The calculated apical surface area of Src42 metamere 8 is smaller than WT. For d–f, error bars represent standard deviation. p values are from Student’s t-test. For all samples, n = 5. (g) Schematic showing the parameters measured for reconstructed cells. For details, see Materials and Methods. (h, i) A representative reconstruction of the apical surface of WT (h) and Src42 (i) metamere 8 cells showing the long axis of each cell (green line). The long axes of Src42 cells are misoriented towards the circumferential direction. (j) The median apical surface area of individual Src42 cells is smaller than WT. (k) The median length of cells along the longitudinal axis of the DT is shorter in Src42 than WT. (l) The median aspect ratio of WT and Src42 cells is not statistically different. (m). The median angle of the longest cell axis relative to the DT longitudinal axis is increased in Src42 mutants. For j–m, black bars represent the median, boxes delineate the upper and lower quartiles, whiskers indicate the highest and lowest values within 1.5 times the interquartile range, and dots represent outliers. p values are from Mann-Whitney U tests. For all box plots, cells were pooled from 4 embryos for n values of 57 cells (WT) and 67 cells (Src42).

To quantify these results at the single cell level, we determined ASA, axial length, aspect ratio, and cell orientation using 3D reconstructions of individual WT and Src42 tracheal cells (Fig. 3g–i and Materials and Methods). As observed for the whole metamere, the median ASA and axial length of individual Src42 cells was smaller than WT cells (Fig. 3j, k). Interestingly, despite the smaller size of Src42 cells, the median aspect ratios of WT and Src42 cells was not significantly different (Fig. 3l). However, the median angle of the longest cell dimension relative to the longitudinal tube axis was closer to 90° in Src42 than WT DT cells (Fig. 3m). Together, these data confirm that Src42 is required for controlling the amount of apical surface expansion, and for directing available expansion anisotropically along the longitudinal axis of the tube. These results establish that epithelial tube size can be controlled by regulated anisotropic growth of the apical surface (Fig. 3a, Model 2).

To define the molecular mechanisms by which Src42 acts, we tested loss-of-function mutations in candidate Src interactors for a short tracheal phenotype (Table S1b online). We found that dDaam, a conserved diaphanous-related formin that has been shown to bind vertebrate Src²⁸, displayed a mild, but significant short tracheal phenotype near the end of embryogenesis (stage 16) (Fig. 4m, 5l). Shortening was more apparent after hatching (Fig. 4a–c), raising the possibility that the weaker embryonic phenotype was due to maternal dDaam (failure of dDaam M/Z embryos to cellularize³⁰ precludes a direct test of this possibility). Like Src42, dDaam acts autonomously in the tracheal system since tracheal expression of Flag-tagged dDaam fully rescued the shortened DT of dDaam mutants (Fig. 4c, f). Remarkably, despite only causing modest reductions in DT length at embryonic stage 16, zygotic dDaam mutations were still able to completely suppress the over-elongation of SJ, aECM, apico-basal and PCP mutants (Fig. 4g–l and Table S1a online). Further, whereas Src42; scrib mutants had gross embryonic defects, dDaam; scrib mutants underwent largely normal development and had trachea with the dDaam short tracheal phenotype (Fig. 4n, o and Table S1a online). Thus, dDaam appears to act downstream or in parallel to all characterized tracheal tube size control pathways.

Figure 4. dDaam is required for tracheal tube elongation.

(a, b, c) The DT (yellow line) of dDaam first instar larvae is shorter than WT. (c) Tracheal specific expression of a Flag-tagged dDaam construct (Flag-dDaam) can fully rescue the dDaam elongation defect. Error bars represent standard deviation. p values are from Student’s t-tests. For WT, n = 6, for dDaam, n = 8, and for Flag-dDaam rescue, n = 7. (d) dDaam is localized at the apical surface of WT tracheal cells. (e) dDaam labeling is greatly reduced in dDaam mutants. (f) The Flag-dDaam rescue construct localizes correctly to the apical surface. (g–l) dDaam can suppress the DT over-elongation of nrv2 (m) The DT of dDaam; nrv2 double mutants is the same length as dDaam single mutants. Error bars represent standard deviation, n = 5 for all samples. (n, o) dDaam can suppress the over-elongation of scrib mutants. (p) Confocal projections of metamere 8 in dDaam embryos labeled for DE-cad. (q, r) Metamere 8 of dDaam mutants is shorter in length (q) but larger in diameter (r) than WT. (s) Unlike Src42 mutants, the calculated apical surface area of dDaam metamere 8 is not different from WT. For m, q–s error bars represent standard deviation, p values are from Student’s t-test, n = 5. (t, u) Representative reconstruction of the apical surface of WT (t) and dDaam (u) metamere 8 cells shows that the long axis (green line) of dDaam cells is oriented circumferentially. (v) The median apical surface area of individual dDaam tracheal cells is not significantly different from WT. (w) The median length of cells along the longitudinal axis of the DT is shorter in dDaam cells than WT. (x) The median aspect ratio of dDaam cells is slightly larger than WT. (y) However, dDaam cells are misoriented along the circumferential direction. For v–x, black bars represent the median, boxes delineate the upper and lower quartiles, whiskers indicate the highest and lowest values within 1.5 times the interquartile range, and dots represent outliers. p values are from Mann-Whitney U tests. For all box plots, cells were pooled from 4 embryos for n values of 57 cells (WT) and 59 cells (dDaam). Scale bars: 40μm for a, b; 10μm for d–l, n–p.

Figure 5. dDaam functions with Src42 to control tube size.

(a–c) dDaam colocalizes with pSrc at the subapical membrane of DT cells (arrows, b, c). However, dDaam is not expressed in fusion cells (asterisks). (d) Src42 co-immunoprecipitates with dDaam from embryo lysates. The blot is representative of three independent experiments (entire blot is shown in Fig. S5g). Anti-GFP antibody was used as a negative control. (e–j) dDaam is enriched in apical circumferential rings (arrows in e and h) that run orthogonally to the length of the tube (e). In Src42 mutants, dDaam still localizes to circumferential rings, but the rings are coarser and more broadly spaced (h). Images in e and h are superficial sections of the DT. Localization of Src42 is not affected in early stage 17 dDaam mutants (f, i). Nonetheless, early stage 17 dDaam mutants show a reduction in the levels of pSrc (g, j). Note that pSrc is not affected in fusion cells, where dDaam is not expressed (asterisks, j). Stage 17 dDaam embryos were imaged to account for maternal dDaam. (k) Quantification of apical pSrc (blue) and Src42 (green) levels in WT and dDaam early stage 17 DT cells shows that the levels of activated Src42 are reduced in dDaam mutants. (l) Measurement of dDaam and Src42 mutant DTs show that both are shorter than WT. dDaam; Src42 double mutants do not display a more severe elongation defect than Src42 single mutants. For k and l, error bars represent standard deviation, p values are from Student’s t-test, and n = 5 for all samples (m) A model for the interactions of aECM and the Src42/dDaam in controlling lumen size. Src42 is required for apical surface growth and both Src42 and dDaam direct growth along the longitudinal axis. aECM restricts tube length. Note that Src and dDaam are shown at the AJ, but the location of the pools of Src and dDaam that control the amount and direction of growth are not known. Scale bars: 10μm for a–c, f, g, i, j; 5μm for e, h.

Using quantitative imaging, we determined that, as in Src42 mutants, the apical surfaces of dDaam tracheal cells were misoriented, over-expanding in the circumferential direction to produce lumens that were shorter but larger in diameter (Fig. 4p–r, t–y). However, unlike Src42 mutants, the apical surface of metamere 8 and the median apical surface of individual tracheal cells were not dramatically altered in dDaam mutants (Fig. 4s, v). Thus, dDaam is required to orient apical surface growth, but is either not required to control the amount of expansion or there is sufficient maternal dDaam to allow normal apical surface expansion.

The similar tracheal phenotypes of Src42 and dDaam mutants suggested they act in the same pathway to control anisotropic surface growth. Consistent with this possibility, the short tracheal phenotype of the dDaam; Src42 double mutant was not enhanced compared to the Src42 single mutant, which is shorter than the dDaam mutant (Fig. 5l). In addition, dDaam colocalizes with pSrc at the AJ and subapical membrane of DT cells (arrows in Fig. 5a–c) and Src42 co-immunoprecipitates with dDaam from embryo lysates (Fig. 5d). Further, loss of dDaam decreased apical pSrc levels in DT cells (Fig. 5g, j, k), suggesting that dDaam acts upstream of Src42 in regulating tube-size. This is consistent with work in mammalian systems showing that Dia-related formins activate Src by binding to the auto-inhibitory SH3 domain^28,31,32. Interestingly, though there were no obvious changes in the localization or levels of Src42 in dDaam mutants (Fig. 5f, i, k), the apical circumferential rings of dDaam were coarser and more broadly spaced in Src42 mutant trachea (Fig. 5e, h), further evidencing functional interactions of these proteins.

Importantly, dDaam mutations do not completely eliminate phospho-Src staining. This result suggests that dDaam does not regulate all pools of Src42 in tracheal cells, which could explain why Src42 mutants cause defects in both surface growth and orientation of growth, whereas dDaam mutations affect only orientation of growth. Consistent with this possibility, over-expression of Src42 in dDaam trachea caused marked excess growth in the circumferential direction, dramatically increasing tube diameter without increasing length (Fig. S4 online). Together, these results support a model in which Src42 regulates surface expansion in a process that does not require dDaam, and that dDaam and Src42 act together in a complex that directs expansion along the longitudinal axis of the tube (Fig. 5m).

Given that Src42 and dDaam control directionality of cell growth in what is essentially a planar epithelium, we further investigated whether Src42 and dDaam act in the canonical PCP pathway. Although Src42 and dDaam do not act in PCP pathways outside the trachea (Fig S3a–l online), PCP genes could act upstream of Src42 and dDaam to control anisotropic growth in tracheal cells. Since a diagnostic feature of aberrant anisotropic growth is an increase in tube diameter when tube length is decreased, we determined tracheal diameter in embryos mutant for ft, a PCP gene whose loss causes short trachea. Tube diameter was unaltered (Fig. S3o, p online), indicating that canonical PCP genes act distinctly from Src42 and dDaam.

Surprisingly, loss of Src42 or dDaam did not cause isotropic (equal in all dimensions) growth of apical surface. Instead, apical surface growth was dramatically reoriented along the circumferential tube axis (Fig. 3l,m; Fig. 4x,y). Thus, Src42 and dDaam are required for orienting anisotropic growth rather than being essential for anisotropic growth per se. Interestingly, homologs of both Src42 and dDaam have been shown to be required for mitotic spindle orientation^33,34, suggesting that Src and Daam may have more broadly conserved roles in cellular anisotropy than have been previously suspected.

Our results reveal that Drosophila trachea use a mechanism of epithelial tube size control in which anisotropic growth of the apical surface drives increases in tube length or diameter. To our knowledge, this mechanism has not previously been described. Given that mammalian Src-family kinases have been implicated in control of tube size in the mammalian kidney¹⁹ and vascular system¹⁸, and the general conservation of basic cellular processes, these results have important implications for control of tube size in other systems.

Supplementary Material

1. Table S1 Tracheal phenotypes of examined mutants.

a. Tracheal length, Verm secretion, and aECM organization for all loss-of-function single and double mutants. All phenotypes are from stage 16 embryos, except for the trachea length of the indicated dDaam alleles, which are from early first instar larvae (L1).

*Qualitative assessment of trachea length. −−, very short; −, short; 0, wildtype; +, slightly long; ++, long; +++, very long. Numbers in parentheses indicate the length of measured trachea, (average ± standard deviation, n = 5.) as a percent of WT trachea length.

^§Tracheal length percentages shown are measured from transverse connectives 5 to 10.

^†Double mutant combinations between Src42 and polarity mutations resulted in head involution defects and dorsal closure defects, which produce shortened bodies and convoluted trachea. Due to these defects, proper staging of the embryos and correct normalization of trachea length are impossible.

ND, not determined.

b. Tracheal phenotypes of embryos mutants for candidate Src-interacting genes.

z, test zygotic for short trachea; e, epistasis test using SJ or aECM mutant; m, test m/z for short trachea; s, test for changes in localization or levels of gene product in Src42 mutants; DT, tracheal Dorsal Trunk.

2. Figure S1. Src64 does not regulate tracheal tube size.

(a, b) Src64 maternal/zygotic (M/Z) null mutations do not show obvious tracheal phenotypes. (c, d) However, forced expression of Src64 in the tracheal system using UAS-Src64 and the btl-gal4 driver can cause over-elongation of tracheal tubes in an otherwise WT background (c), and can rescue the short DT of Src42 mutants (d). (e–l) Src64 is not detected in the tracheal system (e, f, i, j). Src64 can be found in the midgut, demonstrating the specificity of the anti-Src64 antibody (g, h, k, l). Note that embryonic hindguts lack SJs and thus show diffuse Na,K-ATPase staining (magenta in k,l). (m–p) Src64 mutants cannot suppress the DT over-elongation or other defects in the aECM/SJ pathway mutant, Atpα. Dashed white lines indicate the tracheal apical surface. (q, r) Quantification of DT length reveals that Src64 mutant DTs are not different than WT DTs (q) and that Src64 cannot suppress the over-elongation of Atpα trachea (r). For q and r, error bars represent standard deviation, p values are from Student’s t-tests. For q, n = 5 for all samples. For r, n = 4 for all samples. Scale bars: 20μm for a–d; 10μm for e–p.

3. Figure S2. Src42 controls tube size independent of the aECM pathway.

(a, b) Lumenal secretion of Serpentine (Serp) is not affected in Src42 mutants. (c, d) Src42 suppresses the nrv2 tube elongation defect without restoring Serp secretion. Note the absence of Serp staining and the diffuse Cor localization. Dashed white lines indicate the tracheal apical surface. (e, f) Lumenal chitin organization is normal in Src42 mutants, with a clear gap between the apical surface (Crb) and the lumenal chitin fibrils (arrow, inset). (g, h) In aECM/SJ mutants like nrv2, the lumenal chitin is more diffuse and the gap is missing (inset, g). Src42 mutations suppress nrv2 tube elongation without restoring chitin organization (h). (i, j) The SJ barrier function is normal in Src42 mutants, as assessed by the exclusion of a 10kDa fluorescent dye injected into the body cavity from the tracheal lumen. (k, l) Dye readily leaks into the SJ mutant nrv2 (k). Src42 suppresses tube elongation without restoring SJ function (l). (m–p) Src42 mutations also suppress the DT over-elongation of the aECM mutant verm without restoring aECM organization. (q–t) Over-expression of Src42 does not dramatically alter the lumenal accumulation of Verm (q) or Serp (r), the organization of lumenal chitin (s), or the SJ barrier function (t), indicating that Src42-dependent tube expansion acts independent of the aECM/SJ pathway. (u–x) The levels and localization of activated Src42 (pSrc) are not different from WT in sano mutants. (y–bb) verm mutants do not show an obvious change in the levels or localization of pSrc (y, z), Src42 (aa), or pTyr (bb). See Fig. 2 for WT comparisons. Scale bars: 10 μm for a–bb.

4. Figure S3. Src42 and dDaam do not act in canonical PCP pathways to regulate tube dimensions.

(a–d) In WT animals, wing hairs are all oriented such that they point toward the distal end of the wing (a, a′). Mutations in the PCP gene dsh cause characteristic swirls in wing hair orientation (b, b′). In contrast, homozygous Src42^JP45 adult escapers do not show defects in hair PCP (c, c′). Similarly, expression of the dominant negative Src42^KM along the anterior-posterior border of the wing (asterisk in d) with the ptc-Gal4 driver does not dramatically disrupt wing hair orientation (d′). In some ptc>Src42^KM wings, very subtle defects in orientation were present (arrow in d′), though the hairs still pointed towards the distal end of the wing. The subtle effects are likely non-specific as expression of Src42^KM in the entire wing using ms1096-Gal4 resulted in extensive growth and morphogenesis defects (data not shown). (e–h) Though disrupted in PCP mutants (f), thorax bristle orientation was normal in Src42^JP45 adults (g) and in ms1096>Src42^KM adults (h). The ms1096-Gal4 line drives expression in the thorax and the entire wing. (i–l) In WT animals, an enrichment of F-actin localizes to the posterior edge of cell in the denticle belts of embryos (i, white arrows). Src42 and dDaam null alleles do not disrupt the orientation of the F-actin (j, k). In contrast, loss of the core PCP gene Van Gough (Vang) causes some F-actin accumulations to localize to the anterior edge of epidermal cells (l, yellow arrows). For i–l, all images are ventral views with anterior to the left. (m, n) The DT of dsh¹ embryos is not short like Src42 and dDaam mutants, but rather slightly convoluted (a). Over-expression of Dsh in the tracheal system did not alter tube dimensions (n). (o, p) Quantitative measurement of embryonic stage 17 metamere 8 lumens indicates that only ft mutants show a significant change in length from WT (asterisk in c, p < 0.05). Neither dsh nor ft mutants show any changes in tube diameter, indicating that canonical PCP components do not regulate anisotropic surface orientation. Error bars represent standard deviation. p values are from Student’s t-tests. For all samples, n = 5. (q–t) In epidermal cells of stage 16 embryos, GFP-tagged Dsh localizes to the lateral cell junctions (arrows indicate the absence of Dsh-GFP on junctions along the A-P axis, q and r). In contrast, Dsh-GFP is uniformly expressed throughout DT cells in the tracheal system and does not show a planar polarized distribution (g and h). Scale bars: 200 μm for a–h; 25 μm for a′–d′; 5 μm for I–l; 20 μm for m, n; 10 μm for q–t.

5. Figure S4. Src42 can increase apical surface expansion in dDaam mutants, but cannot correct orientation of growth.

(a–d) As compared to WT (a), dDaam mutants have shorter but wider tubes, indicating that growth of the apical surface is misoriented (b). Over-expression of Src42 increases apical surface, resulting in longer tracheal tubes (c). When Src42 is over-expressed in dDaam mutants, apical surface still expands, but only in the circumferential axis (d). For a–d, images are of metamere 8 and the dashed line highlights the length of the metamere. Arrowheads in (a) point to the transverse connectives. (e, f) Quantitative measurements of metamere 8 reveal the length of the DT is the same in dDaam embryos and dDaam embryos over-expressing Src42 (e). However, diameter increases dramatically (f). All measurements are from stage 16 embryos. Error bars represent standard deviation. p values are from Student’s t-tests. For all samples, n = 5. (g) Uncropped image of the immunoprecipitation gel from Figure 5d. Black arrowhead indicates IgG band and red arrowhead indicates Src42 band. Scale bar: 5 μm for a–d.