Regulation of Archease by the mTOR-vATPase axis

The Drosophila tracheal system consists of an interconnected network of branched tubes. The system originates from ten pairs of epithelial sacs composed of roughly 80 cells each. During embryogenesis, these sacs, which have invaginated from the surface epithelium, undergo successive rounds of branching (Ghabrial et al., 2003; Schottenfeld et al., 2010). In primary branching, tip cells actively migrate and pull trailing cells along behind them; in this manner, typical sacs give rise to six primary branches (Ribeiro et al., 2004; Ghabrial and Krasnow, 2006; Caussinus et al., 2008). The largest tubes are multicellular and smaller tubes are autocellular (surrounding a lumenal space and sealing by formation of autocellular junctions) or seamless (elaborating an internal apical membrane that extends through the branches of stellate-shaped terminal cells) (Samakovlis et al., 1996a,b). Secondary and tertiary branching are accompanied by specification of ‘fusion’ cells that mediate anastomoses with neighboring branches, and terminal cells extend blind-ended seamless tubes to target tissues (Samakovlis et al., 1996a,b). Embryonic tracheal terminal cells are the ultimate cells in tracheal branches and are connected to the network via intercellular junctions with their neighboring autocellular stalk cells. Terminal cells initially form a single intracellular blind-ended tube, but, subsequently, during larval life, sprout dozens of long, highly branched ‘seamless’ tubes (Samakovlis et al., 1996a; Ghabrial et al., 2003; Schottenfeld et al., 2010). These tubes are initially liquid-filled, but undergo gas-filling at molts (Manning and Krasnow, 1993; Tsarouhas et al., 2007). During each instar, new larger tubes are constructed around the pre-existing gas-filled tubes, and become gas-filled themselves at the molt. In terminal cells, the new tubes are not only larger, but include new branches targeting additional tissues.

The apical (lumenal) membrane domain of terminal cells must expand dramatically to permit new tubes to ramify adequately on internal tissues, thus permitting gas exchange. Mutations that restrict the generation of apical membrane in terminal cells, such as vATPase and Rheb mutations, or that directly affect the generation of apical polarity, such as aPKC and par-6 mutations, cause compensatory growth of the neighboring wild-type stalk cells, which lengthen, invade and branch within the mutant terminal cells (Francis and Ghabrial, 2015). Mutations in two vATPase genes were isolated based on the presence of a gas-filling defect in what appeared to be the terminal cell tube, adjacent to the position where the intercellular junction connecting terminal (seamless tubes) and stalk (autocellular tube) cells would normally be found. In fact, the tube lacking gas-filling is actually that of the invading stalk cell – the failure to gas-fill may be related to the remodeling that occurs as an autocellular tube replaces the seamless tube, or may reflect the fact that liquid clearance typically occurs during molts and newly made tubes remain liquid filled in-between molts (Manning and Krasnow, 1993). Because the autocellular tubes of the stalk cell penetrate dozens of microns into the compromised terminal cell, and branch, the result is pairs of terminal and stalk cells with multiple intercellular junctions marking the connections of stalk cell-derived autocellular tubes – surrounded by terminal cell cytoplasm – and terminal cell-derived seamless tubes.

In a previously described forward genetic screen, we identified mutations in four loci that caused related gas-filling defects (Ghabrial et al., 2011). Upon further characterization, one of these loci, lotus, was distinct, preventing the formation of a stable connection between autocellular and seamless tubes (Song et al., 2013), whereas mutations in the other three loci induced compensatory growth from neighboring wild-type stalk cells, as described above (Francis and Ghabrial, 2015). We identified two of those loci as components of the vATPase: oak gall (Vha26) and conjoined (Vha13). Here, we identify the remaining locus, disjointed, as encoding the Drosophila Archease. Although evolutionarily ancient, relatively few studies have focused on Archease (Canaves, 2004; Auxilien et al., 2007; Desai et al., 2014, 2015; Jurkin et al., 2014; Popow et al., 2014; Kosmaczewski et al., 2015; Song et al., 2015; Poothong et al., 2017; Duan et al., 2020). Compelling work has established a requirement for Archease as a co-factor for the RNA ligase, RtcB (Popow et al., 2014). Together, Archease and RtcB play an essential role in transfer RNA (tRNA) maturation and in nonconventional splicing of X-box binding protein 1 (Xbp1) mRNA (Desai et al., 2014, 2015; Jurkin et al., 2014; Popow et al., 2014; Poothong et al., 2017).

Nascent tRNAs undergo three processing events to become mature tRNAs ready to function in translation: (1) removal of the 3′ extension sequences, (2) intron splicing and ligation and (3) aminoacylation. Archease and RtcB are required for the second of these tRNA maturation steps. Unlike mRNA splicing, tRNA splicing is not mediated by the spliceosome, but is instead carried out by complexes containing only protein. Not all tRNAs contain an intron, but throughout the animal kingdom roughly 6-7% of tRNA genes (slightly higher in zebrafish and frogs, and closer to ∼25% in budding and fission yeast) contain an intron (Schmidt and Matera, 2020). The tRNA splicing endonuclease (SEN/TSEN) complex initiates the splicing process by cleaving the tRNA intron (Schmidt and Matera, 2020). Plants and fungi differ from metazoans in using a ‘healing and sealing’ ligation pathway, rather than a direct ligation pathway (Schmidt and Matera, 2020). The tRNA ligase complex consists of the core tRNA ligase (HSPC117/RtcB) and ASW (also known as C2orf49), CGI-99 (also known as C14orf166 or RTRAF), FAM98B and the DEAD-box helicase DDX1 (Kroupova et al., 2021). Archease (ARCH or ZBTB8OS) is a co-factor for RtcB and is required for full activity of the tRNA ligase (Popow et al., 2014); however, its interaction with RtcB is transient (Kroupova et al., 2021). Unlike the ‘healing and sealing’ pathway, the direct ligation pathway produces not only a spliced tRNA, but also a circular intronic RNA, tricRNA, of poorly understood function. In yeast, splicing of tRNAs occurs in the cytoplasm, in association with mitochondria, whereas in eukaryotes pre-tRNA splicing occurs in the nucleoplasm (reviewed by Chatterjee et al., 2018).

The tRNA ligase complex and Archease also are required for the nonconventional splicing of a transcription factor in the cytoplasm (Jurkin et al., 2014). Xbp1 is a transcription factor that induces a suite of unfolded protein response (UPR) genes under conditions of endoplasmic reticulum (ER) stress. In the absence of ER stress, Xbp1 mRNA encodes for a nonfunctional protein, Xbp1m. Upon ER stress, the nuclease activity of the Inositol-requiring enzyme 1 (IRE1) stress sensor is activated, and cleaves the Xbp1 mRNA. The mRNA fragments are then ligated by Archease/RtcB (Jurkin et al., 2014), resulting in a change in the reading frame of the transcript, and consequent translation of a functional Xbp1s transcription factor. Nonconventional splicing of the Xbp1 mRNA is thought to be cytoplasmic, as the mRNA is localized to the ER membrane, where IRE1 cleaves it to initiate nonconventional splicing (Back et al., 2006; Uemura et al., 2009; Yanagitani et al., 2009; Wang et al., 2015; but also see Wang et al., 2015).

Taken together, requirements for RtcB/Archease at the cytoplasmic face of the ER, and in the nucleus, implies that the tRNA ligase complex/Archease can shuttle between the nucleus and the cytoplasm. Indeed, the tRNA ligation complex is known to shuttle, and has been found both in the nucleus and in association with cytoplasmic RNA granules (Pérez-González et al., 2014; Duan et al., 2020). Published data have described a broad cytoplasmic distribution of Archease, at least in cell culture (Jurkin et al., 2014; Duan et al., 2020). Here, we characterize the requirement for Archease in tracheal terminal cell morphogenesis, identify a primarily nuclear localization for Archease, and uncover regulation of Archease subcellular localization by TOR-vATPase pathway activity.

A gap in gas-filling at the stalk cell:terminal cell interface is characteristic of a group of mutations – oak gall (okg), conjoined (cnj), lotus and disjointed – that we first identified in a forward genetic screen (Ghabrial et al., 2011). Whereas mutations in okg and cnj were found to cause a gas-filling gap in terminal cells as a result of invasion by neighboring wild-type stalk cells, mutations in lotus disrupted the connection between seamed and seamless tubes in the terminal cell (Song et al., 2013; Francis and Ghabrial, 2015). We determined that lotus encodes Nsf2, and is therefore required for vesicle trafficking, whereas okg and cnj each encode a subunit of the vacuolar ATPase, and were determined to act downstream of Tor in regulating apical membrane domain size. We chose to further characterize the single mutant allele of disjointed (dsj), the last remaining member of this phenotypic class (Fig. 1A-B″). We found that, in addition to the transition zone gas-filling defects (59% of terminal cells), 24% of dsj terminal cells entirely lacked gas-filling and 17% were normally gas-filled (n=41). Examination of terminal cell tubes revealed that they are continuous (like okg and cnj mutants).

Given their striking phenotypic similarity, we next investigated whether disjointed mutant terminal cells induced compensatory stalk cell branching, similar to that documented for cells mutant for members of the TOR-vATPase axis (Francis and Ghabrial, 2015). In wild-type animals, terminal cells connect to neighboring stalk cells at a single intercellular connection 96% of the time (Fig. 1C) and 4% made two connections to the same terminal cell. The intercellular connection includes septate (functionally equivalent to vertebrate tight junctions for paracellular barrier function) and adherens junctions. Consistent with the phenotype previously described for mutants in the TOR-vATPase axis (Francis and Ghabrial, 2015), in dsj mosaic animals, 72% (n=54) of mutant terminal cells had two connections to their neighboring stalk cell, and an additional 4% of dsj terminal cells were invaded by long extensions of stalk cell autocellular tube terminating in three or more intercellular connections (Fig. 1C-E′).

To understand better the cell biological and molecular functions of dsj in the context of the terminal cell-to-stalk cell interface, we mapped the mutation (Fig. 2A). Sequence analysis of candidate genes within the candidate interval revealed a G-to-A transition in the third exon of the gene CG6353, which encodes the Drosophila ortholog of human Archease/ZBTB8OS (Fig. 2B). This missense mutation is predicted to convert a highly conserved glycine residue to glutamic acid, creating a loss-of-function allele (G₁₁₅E). Archease proteins across kingdoms have been implicated in DNA or RNA metabolism (Auxilien et al., 2007; Desai et al., 2014), with more recent work identifying Archease as a co-factor for RtcB, an RNA ligase required for tRNA maturation and nonconventional splicing of Xbp1 mRNA (Desai et al., 2014; Popow et al., 2014).

Two additional experiments confirmed gene identity. Expression of an inducible wild-type Archease cDNA construct rescued terminal cell gas-filling and morphology in dsj mutant cells (Fig. 2C), and depletion of CG6353 by RNAi in otherwise wild-type larvae reproduced the defects observed in dsj mutant terminal cells (Fig. 2D). Terminal cell-specific depletion of CG6353 resulted in a substantial increase in the number of terminal cells connected to neighboring stalk cells by two intercellular junctions (43%, n=76), and resulted in a striking number of terminal cells (37%) invaded by three or more autocellular tubes (a condition which has never been observed at all in wild type). The increased frequency of the more severe phenotype upon dsj knockdown, compared with the dsj point mutation, suggests the existence of maternally deposited dsj mRNA, or that the point mutation in dsj is hypomorphic, or both. Consistent with maternal deposition of Archease, dsj homozygous animals survive until the second larval instar; however, we could not further test this hypothesis as maternal depletion of dsj resulted in defective oogenesis. Using the dominant female sterile technique (Chouet al., 1993), we found that females with germline clones mutant for Archease were sterile. Taken together, these data establish that dsj encodes Archease.

The rescuing transgene we generated was N-terminally tagged with the HA epitope. Although, we were unable to identify a canonical nuclear localization signal in Archease, we observed strong nuclear localization of tagged Archease in rescued tracheal cells (Fig. 3A,A′). Although this result was consistent with a requirement for Archease in tRNA maturation (which occurs in the nucleus), it conflicts with previously published data from (HeLa) cells suggesting an exclusively, or largely, cytoplasmic localization for Archease (Jurkin et al., 2014; Duan et al., 2020).

Because dsj displays phenotypes similar to okg and cnj (which encode two vATPase subunits), we sought to determine whether vATPase activity regulates dsj. We found that, in contrast to a robust nuclear signal in wild-type terminal cells, Archease was absent from cnj terminal cell nuclei (Fig. 3B-B″ yellow caret) and instead accumulated in the cytoplasm. In internal controls, (neighboring heterozygous cells in the same larvae), Archease remained nuclear (Fig. 3B, white carets). We and others have reported that the vATPase acts genetically downstream of the TOR pathway (Gleixner et al., 2014; Francis and Ghabrial, 2015). As Archease regulates tRNA maturation, which is necessary for protein synthesis, we sought to determine whether TOR pathway activity was required for Archease localization to the nucleus. To this end, we generated Rheb mosaic animals and examined the homozygous mutant terminal cells. Although some Archease was still localized to the small Rheb-positive terminal cell nuclei, a very substantial shift to the cytoplasm was detected (Fig. 3C-C″), in particular compared with neighboring internal control heterozygous cells (Fig. 3C, white carets). This suggested that activity of the TOR-vATPase axis drives Archease to the nucleus, or is required for its maintenance there. Genetic epistasis tests of Archease (disjointed) loss of function with TSC1 (jolly green giant) loss of function, or with over-expressed Rheb, are consistent with Archease acting downstream of TOR (Fig. S1).

Archease acts as a co-factor for the tRNA ligase RtcB (Popow et al., 2014) during one step of tRNA maturation, a multi-step process that begins with intron splicing and 3′ end ligation, followed by 5′ and 3′ end processing (Yoshihisa, 2014). The 3′ ends are removed by the endoribonuclease tRNAseZ (RNaseZ, the Drosophila homolog of human ELAC1/2) (Takaku et al., 2003). Removal of the 5′ and 3′ ends is followed by the addition of CCA, which prepares tRNAs for aminoacylation and nuclear export (Hopper and Shaheen, 2008). To test whether defects associated with loss of dsj could result from its role in tRNA processing, we depleted tRNAseZ from terminal cells and assessed the intercellular junction phenotype. We observed that 68% (n=47) of tRNAseZ-depleted terminal cells displayed stalk cell invasion defects (Fig. 4). Similar results obtained from blocking the nuclear export of RNAs, including tRNAs, by knockdown of NTF2-related export protein 1 (NXT1): terminal cells were recovered with two intercellular junctions (43%) and with three or more intercellular junctions (26%) (n=46).

To test whether the loss of nonconventional splicing of Xbp1 could also phenocopy the terminal cell requirement for Archease, we depleted cells of Xbp1 mRNA by RNAi and also examined trachea of larvae null for Xbp1, but rescued to viability by Xbp1 cDNA expression in the alimentary canal (Huang et al., 2017). Prior studies have established that Xbp1s nonconventional splicing occurs at detectable levels in wild-type larval trachea, brain glia, Malpighian tubules and gut, whereas Xbp1s splicing in other tissues appears to require induction of ER stress (Sone et al., 2013). We found that loss of Xbp1 did not perturb terminal cell development or morphology, consistent with the hypothesis that it is the requirement for Archease in tRNA maturation that results in the disjointed phenotype.

In this and previous work (Francis and Ghabrial, 2015; Burguete et al., 2019), we determined that most mutations that impair tracheal terminal cell apical membrane growth, and, in particular, mutations in components of the TOR pathway, trigger induction of a compensatory growth program in neighboring wild-type tracheal stalk cells. As part of this compensatory program, stalk cells, which form autocellular tubes, are induced to grow, branch, and extend into the terminal cell. These two cells, which normally share a single intercellular junction, instead share two, three or more intercellular junctions. Moreover, these junctions are no longer located at the initial point of cell-cell contact, but instead the stalk autocellular tubes and their surrounding cytoplasm invade the terminal cell and become enveloped by the cytoplasm of the terminal cell. One fascinating and unresolved question is whether this invasion of the terminal cell requires the generation of a signal from the terminal cell to the stalk cell, or whether the altered mechanical properties of the terminal cell are sufficient to trigger compensatory growth.

The relationship between core TOR pathway components, the vATPase and Archease is complex. In our prior work (Francis and Ghabrial, 2015), we determined that vATPase is epistatic to (downstream of) TOR in regulating apical membrane growth, consistent with the results of Gleixner and colleagues (Gleixner et al., 2014). However, the genetic relationships are not quite so straightforward, as vATPase activity also regulates mTOR (Peña-Llopis et al., 2011; Zoncu et al., 2011). There had been no previous indication in the literature that TOR would regulate Archease; however, given the role of TOR in promoting translation, and the role of Archease in tRNA splicing, such regulation has an appealing logic.

We found that mutations in Rheb as well as vATPase components Vha26 and Vha13 had profound effects on Archease subcellular localization. In wild-type tracheal cells, Archease appeared greatly enriched in nuclei – we found this distribution of Archease in Drosophila to be consistent across multiple tissues, including salivary gland epithelia and neurons (Fig. S2). This stands in contrast to published reports of Archease localization in HeLa cells (Jurkin et al., 2014; Duan et al., 2020), where Archease is reported to be largely cytoplasmic. This discrepancy may reflect differences between post-mitotic and actively cycling cells, cells in culture compared with in vivo, or other factors. Importantly, in vATPase and Rheb mutant cells, there was a dramatic redistribution of Archease from the nucleus to the cytoplasm. The subcellular localization of Archease is likely important, as its two best characterized functions, splicing of tRNA introns and splicing of Xbp1 mRNA, are thought to occur in the nucleus and the cytoplasm, respectively. This would suggest that conditions required for induction of the stress response by Xbp1s might correspond with a general decrease in tRNA availability, and, likewise, that robust mTOR-driven cell growth might limit the ability of the cell to generate the Xbp1s isoform. A less-well-characterized role of the tRNA ligase complex and Archease is in the production of tricRNAs – circularized tRNA introns that are stable and abundant, but of unknown function (Schmidt and Matera, 2020). A role for tricRNAs in compensatory growth cannot be ruled out.

In addition to implications for cellular growth and the stress response, regulation of Archease subcellular localization may also be of interest from the perspective of neuronal regeneration. Genetic disruption of the XBP1 arm of the UPR impedes axonal regeneration (Song et al., 2015; Oñate et al., 2016) (although also see Kosmaczewski et al., 2015). This would suggest that cytoplasmic localization of Archease, where it would be required for productive splicing of Xbp1 mRNA, might contribute to regeneration. However, depletion of the mTOR pathway inhibitors PTEN and TSC1 (Tuberous Sclerosis 1/Hamartin) significantly enhances axonal regeneration (Park et al., 2008, 2010; Liu et al., 2010; Song et al., 2012; Yang et al., 2014; Du et al., 2015; Miao et al., 2016), and would be predicted to shift Archease to the nucleus. It may be that under these conditions sufficient Archease is present in the cytoplasm to promote Xbp1 splicing, or, alternatively, it may be that regeneration could be further enhanced by upregulation of Archease expression together with activation of TOR signaling.

Fly stocks used were: btl-Gal4, UAS-DsRED2^NLS, UAS-GFP; FRT^82B UAS-GFP RNAi (Ghabrial et al., 2011); btl-GAL4, UAS-GFP; FRT^2A, FRT^82Bdsj¹⁶⁹/ TM3, Sb, Tub-GAL80 (Ghabrial et al., 2011); btl-GAL4, UAS-GFP; FRT^2A, FRT^82BTSC1^{jollygreengiant} /TM3, Sb, Tub-GAL80 (Ghabrial et al., 2011); FRT^82Bcnj ³⁵⁶/ TM6 (Francis and Ghabrial, 2015); SRF>GAL4 (gift of M. Metzstein, Department of Human Genetics, University of Utah, USA); P{TRiP.HMS03015}attP2 – xbp1 RNAi (Bloomington Drosophila Stock Center); P{TRiP.JF02012}attP2– xbp1 RNAi (Bloomington Drosophila Stock Center); UAS-CG6353 ^HMS02543 (dsj) RNAi (Bloomington Drosophila Stock Center); UAS-Rheb^PA (Bloomington Drosophila Stock Center); UAS-tRNAseZ ^{jhl-1 HMC03826} RNAi (Bloomington Drosophila Stock Center); UAS-NXT1 ^GL00414 RNAi (Bloomington Drosophila Stock Center); TRiP JF02012 (Xbp1 RNAi) (Song et al., 2015) (Bloomington Drosophila Stock Center); and 4E-BP^intron DsRed, Xbp1^ex79; UAS-Xbp1-RA/SM5-TM6B, which were crossed with NP1-GAL4, Xbp1^ex79/CyO,GFP; tho2p>DsRed/TM6B to produce Xbp1-null larvae rescued to viability by BP1>xbp1 (gifts from Don Ryoo, Department of Cell Biology, New York University, USA). Ten dorsal branch terminal cells from each of five animals of the correct genotype were examined.

Third instar larvae of both genders were filleted and fixed in 4% paraformaldehyde (Electron Microscopy Sciences) for 15 min at room temperature. Antibodies used were: rat anti-DE-Cadherin (1:50, Developmental Studies Hybridoma Bank), Rb anti-aPKC (PKCz, 1:200, Santa Cruz Biotechnology), chicken anti-GFP (1:1000, Invitrogen, A10262) and mouse anti-HA tag (1:100, Clone HA7, Sigma-Aldrich, H3663). Larvae were mounted in aqua polymount (Polysciences), and images were acquired on Leica DM5500 and DMI6000 B microscopes. z-stacks were captured and processed with Leica and Fiji (ImageJ) software. Projected z-stacks or single slices are shown, as noted. Statistical significance was determined using the Fisher exact probability test (http://www.vassarstats.net).

Generation of pUASt–HA-dsj: CG6353 complementary DNA FI17516 (Drosophila Genomics Resource Center) was used as a template for PCR primers. A PCR product containing an N-terminal HA tag was subcloned into pUAST (Brand and Perrimon, 1993). The transgene was injected into w¹¹¹⁸ embryos to generate UAS-HA-dsj transgenic flies.

SRF-GAL4- or drm-GAL4-driven expression of UAS-CG6353^HMS02543 RNAi, UAS-tRNAseZ^{jhl-1 HMC03826} RNAi, UAS-NXT1^GL00414 RNAi, and UAS-GFP was used. Flies were kept at 29 and third instar larvae were collected.

The following crosses were carried out: UAS-HA-dsj; FRT^82Bdsj¹⁶⁹/TM6B flies were crossed to btl-GAL4, UAS-RFP/CyO; FRT^82B Tubulin Gal80/TM6B. Mosaic larvae with RFP marked clones were collected for immunostaining.

dsj genomic DNA and control genomic DNA from the parental strain on which dsj mutations were induced were amplified by PCR and sequenced. The following primers were used: CG6353 genomic region; F1 5′-GCATAGATGGTCACACTAAGCGG-3′; R1 5′-AAATCGCTCGGTGTTACCAGC-3′; F2 5′-TCAGTAGGGAGAACTTCCTGCTGC-3′; R2 5′-TGGATTCTGTCAAATGGGAAGG-3′; F3 5′-CAAAAACAACAAGTGTGCCCG-3′; R3 5′-ATTGCCTTCACTTCGGTGCC-3′; F4 5′-TGTGACTGTTCTGTTTCAACCCC-3′.

The following crosses were carried out: (1) UAS-Rheb (homozygous on the 2nd), FRT^82Bdsj/TM6 males were crossed to btl>Gal4, UAS-GFP, FRT^82B UAS-GFP RNAi females to generate clones. Junction phenotypes of GFP marked terminal cells were compared with unmarked terminal cells within the same larvae. (2) UAS-CG6353 ^HMS02543 RNAi (homozygous on 2nd), FRT ^82BTSC1^{jollygreengiant}/TM3 males were crossed to btl>Gal4, UAS-GFP, FRT^82B UAS-GFP RNAi females. Mosaic larvae were collected, filleted, immunostained and imaged.

Some of the text and data in this paper formed a part of Deanne Francis's PhD thesis in the Department of Cell and Developmental Biology at the University of Pennsylvania in 2015. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200908.reviewer-comments.pdf