A clean null allele of amx fully recapitulates previously reported maternal-effect phenotypes

In a previous study, Michellod et al. reported the generation of a null allele of amx (amx^m) by generating flies that are homozygous for a deletion that removes amx and several other genes [Df(1)FF8] and bringing back a genomic rescue construct for Dsor1, an essential gene that lies within this locus [30]. The authors reported that zygotic amx^m [actual genotype: Df(1)FF8; P(Dsor1^+S)] mutant flies have reduced eye size and nicked wing margin as adults, which are phenotypes that had not been reported in the classic amx¹ allele. Because the amx¹ allele is caused by a 5bp deletion that may still produce a product that encodes a portion of the N’-extracellular domain (amx is a single exon gene in Drosophila, hence will not be subjected to nonsense mediated decay), the authors concluded that amx¹ is a hypomorphic allele.

Since Df(1)FF8 has not been molecularly characterized and it was uncertain whether all of the phenotypes attributed to amx^m are due to loss of amx alone, we decided to generate a clean null allele of amx (hereafter referred to as amx^Δ) using CRISPR/Cas9 technology [31,32] (Fig 1). We knocked-in a dominant wing color marker (y^wing2+) to replace the coding sequence of amx using HDR [33] (Fig 1A). Insertion of this cassette was screened by the presence of the visible marker (dark wings) in a yellow mutant background and the targeting event was molecularly confirmed via Sanger sequencing. We also confirmed the loss of amx transcript by RT-PCR (S2A Fig). Similar to amx¹, homozygous amx^Δ females exhibit sterility and all embryos produced by these animals exhibit a neurogenic phenotype (Fig 1C and 1D). Both female fertility and neurogenic phenotypes of their progeny can be suppressed by introducing a 3.3 kb genomic rescue construct containing the amx locus (Fig 1A, 1C and 1E) [13]. Human TM2D3 expressed using the same regulatory elements has ~50% activity of fly Amx (Fig 1C and 1G), consistent with what we previously observed using amx¹ hemizygous females [13]. We did not observe any morphological defects in the eye and wing of homozygous and hemizygous amx^Δ flies at all temperatures tested (between 18–29°C). We also looked at other tissues that are often affected when Notch signaling is defective including the nota and legs [34,35] but we did not observe any morphological phenotypes in these tissues either. In summary, amx^Δ is the first clean loss-of-function (LoF) allele of amx generated by CRISPR/Cas9, and phenotypically resembles the classic amx¹ allele rather than the amx^m allele.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. A clean null allele of TM2D3 fly ortholog almondex (amx^Δ) behaves like the classic amx¹ allele.

(A) Schematic diagram of almondex (amx) locus, amx^Δ allele and genomic rescue constructs used in this study. (B) Predicted 2D-structure of Drosophila Amx protein. SS = signal sequence for membrane localization. Transmembrane 2 (TM2) domain is boxed in red. Hexagon denotes where 3xHA epitope is located in 3xHA::Amx protein. (C) Egg hatching assay shows that genomic rescue constructs can suppress embryonic lethality (amx^Δ n = 857. amx^Δ + amx n = 139. amx^Δ + 3xHA::amx n = 1673. amx^Δ + TM2D3 n = 257). Error bars show SEM. One-way ANOVA followed by Dunnett test. **** = p-value≤0.0001. (D-G) Embryonic nervous tissue (neuronal nuclei, Elav, green) of developing embryos. Embryos from homozygous amx^Δ females (D) exhibit a neurogenic phenotype. This phenotype can be suppressed by wild-type amx (E), 3xHA::amx (F), or human TM2D3 (G) genomic rescue constructs. Scale bars = 100 μm in (D-G).

https://doi.org/10.1371/journal.pgen.1009962.g001

Null alleles of other TM2D genes exhibit the maternal-effect neurogenic phenotype similar to amx

Using a similar strategy that we used to generate amx^Δ [33], we generated null alleles for CG11103 (TM2D2 ortholog, located on the X-chromosome) and CG10795 (TM2D1 ortholog, located on the 2^nd chromosome), two uncharacterized genes without official names (Fig 2). This time, we inserted a dominant body color marker (y^body+) into the endogenous locus of CG11103 or CG10795 to individually knockout these genes (Fig 2B and 2F), and we phenotypically and molecularly characterized these alleles in a similar manner. We confirmed the loss of each transcript by RT-PCR (S2A Fig), confirming that the new lines we generated are true null alleles.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Null alleles of fly orthologs of TM2D2 (CG11103/amrt) and TM2D1 (CG10795/bisc) phenotypically mimics the loss of TM2D3 (amx).

(A) Protein alignment of human and Drosophila TM2D proteins. The TM2 domain (boxed in red) is composed of two transmembrane domains (TMD) and an intracellular DRF motif (denoted by black bars for TM2D3). (B) Schematic of amaretto (amrt) locus, amrt null allele (amrt^Δ), and amrt genomic rescue plasmid construct generated for this study. (C) Embryos from homozygous amrt^Δ females exhibit neurogenic phenotype, which can be suppressed by the amrt genomic rescue construct (D). (E) Egg hatching assay showing that amrt genomic rescue construct suppresses 68% of embryo lethality (amrt^Δ n = 954. amrt^Δ + amrt n = 1490). (F) Schematic of biscotti (bisc) locus, bisc null allele (bisc^Δ), and bisc::GFP genomic rescue fosmid construct generated for this study. (G) Embryos from homozygous bisc^Δ females exhibit neurogenic phenotype (n = 379) which can be suppressed by bisc::GFP genomic rescue construct (H). (I) Egg hatching assay showing that bisc::GFP suppresses 30% of embryonic lethality (bisc^Δ n = 379. bisc^Δ + bisc::GFP n = 436). t-test, **** = p-value<0.0001. Scale bars = 100 μm in (C, D, G, H).

https://doi.org/10.1371/journal.pgen.1009962.g002

Both males and females that are hemizygous or homozygous null for CG11103 are viable and these animals do not exhibit any morphological defects. Similar to amx, CG11103 mutant females are fully sterile, and all embryos laid by these mothers exhibit neurogenic phenotypes (Fig 2C). Importantly, these phenotypes are rescued by a 1.5 kb genomic construct containing the CG11103 locus (Fig 2B, 2D and 2E). Given the phenotypic and molecular similarities with amx, we gave CG11103 the name amaretto (amrt), after the sweet Italian liqueur traditionally flavored with almonds. The knockout allele for this gene is referred to as amrt^Δ hereafter.

Like amx and amrt, mutants that are homozygous null for CG10795 appear morphologically normal, but all females are sterile and their embryos exhibit a neurogenic phenotype (Fig 2G). We were able to suppress these phenotypes using a ~40kb fosmid transgene in which the CG10795 protein is C-terminaly tagged with multiple epitopes including GFP (Fig 2F, 2H and 2I) [36]. Given the phenotypic similarity to amx and amrt mutants, we named this gene biscotti (bisc), after the Italian biscuits traditionally made with almonds. The knockout allele for this gene is referred to as bisc^Δ hereafter and the fosmid-based GFP-tagged genomic rescue transgene is referred to as bisc::GFP.

Triple knockout of TM2D genes is phenotypically similar to single gene knockouts

The embryonic neurogenic defect is a rare phenotype that is almost exclusively associated with genes that affect the Notch signaling pathway [20]. Given the importance of Notch signaling in most stages of development preceding adulthood [7], it was peculiar that all three TM2D gene mutants do not exhibit any obvious developmental morphological defects related to Notch signaling while all three alleles exhibit a robust maternal-effect neurogenic phenotype. To test whether this may be due to redundancy between the three genes, we generated a fly strain that lacks all three TM2D genes (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Triple null mutant for all three TM2D fly genes is phenotypically similar to single null mutants.

(A) Crossing scheme used to generate TM2D triple null mutant flies. (B) A normal embryonic nervous system highlighted by neuronal nuclei marker Elav (green). (C-D) amx^Δ amrt^Δ double mutants (C) and amx^Δ amrt^Δ bisc^Δ triple mutants (D) exhibit a neurogenic phenotype. (E-K) TM2D double and triple null mutants exhibit no overt morphological phenotypes. Head structures of mutants (F, G) appear normal compared to y w control (E). Mutant wings (I, J) and thorax (L, M) also appear normal compared to control (H, K). Scale bars = 100 μm in (B-D).

https://doi.org/10.1371/journal.pgen.1009962.g003

Because amx and amrt are both located on the X-chromosome (located in cytological regions 8D2 and 12C4, respectively, which are 19cM away), we recombined the two null alleles by following the dominant wing color and body color markers knocked into each locus (Fig 3A). The amx, amrt double null flies (amx^Δ amrt^Δ) in a yellow mutant background exhibit wild-type color wings (y^wing2+ marker of amx^Δ) and bodies (y^body+ marker of amrt^Δ). Absence of both amx and amrt transcripts in this line was verified by RT-PCR (S2B Fig). These double null adult flies also appeared morphologically normal similar to the single null animals (Fig 3F, 3I and 3L), suggesting that these two genes do not play redundant roles during the development of imaginal tissues. Consistent with single null animals, the amx, amrt double null females were fully sterile, exhibiting a maternal-effect neurogenic phenotype (Fig 3C).

We next combined the bisc^Δ allele on the second chromosome with the amx^Δ amrt^Δ mutant X-chromosome to generate a triple null mutant line (amx^Δ amrt^Δ; bisc^Δ) (Fig 3A). Absence of transcripts of all three TM2D genes in this line was verified by RT-PCR (S2B Fig). This genetic manipulation still did not produce any adult animals with obvious Notch signaling related external morphological defects (Fig 3G, 3J and 3M), suggesting that these three genes also do not play redundant roles in these contexts. Consistent with the single and double null lines, the triple null mutant females are also completely sterile and their progeny show maternal-effect neurogenic defects (Fig 3D). We did not observe any differences in the severity of the neurogenic phenotype in the embryos from the single gene knockouts and the triple knockout mothers (Figs 1D, 2C, 2G and 3D), likely because the neurogenic defect is already strong in single gene knockout animals, precluding any additive or synergistic effects.

According to large-scale transcriptomic studies, mRNA of amx, amrt and bisc are expressed relatively ubiquitously with some enrichment found in the female ovary [37,38]. Within the ovary, these three genes are likely to be expressed in similar cell types including germline cells and follicle cells according to a large single cell RNA-sequence database that have recently been released on the bioRxiv preprint server [39]. To further test whether amx, amrt and bisc do not have overlapping functions, we assessed whether genomic rescue transgenes for amrt or bisc can rescue amx mutants. If TM2D family proteins perform similar molecular tasks, increased expression of one gene is expected to suppress the LoF phenotype of another gene. Homozygous amx^Δ female flies with amrt or bisc::GFP genomic rescue transgenes (used to rescue amrt^Δ and bisc^Δ in Fig 2, respetively) are fully sterile, demonstrating that amrt or bisc are not able to substitute for the loss of amx. Together with our earlier finding that single TM2D gene knockout flies are phenotypically indistinguishable (Figs 1 and 2), and that no further phenotype or exacerbation is seen in the triple null mutant line (Fig 3), we conclude that these three genes are uniquely maternally required during embryonic neurogenesis and likely function together.

A truncated but not the full length form of Amx acts as a potent inhibitor of Notch signaling

Knockout experiments revealed that each of the three TM2D genes are individually maternally necessary for proper Notch signaling during embryogenesis, while being dispensable for other developmental contexts that are known to depend on Notch. To further study the function of TM2D proteins in vivo, we tested whether overexpression of Amx is sufficient to modulate Notch signaling (Fig 4). In addition to generating transgenes that allow the expression of a full-length Amx protein with or without an N’-3xHA tag (placed immediately after the predicted signal sequence, 3xHA::Amx^FL) (Fig 4A) using the GAL4/UAS binary expression system [40], we generated a version of the N’-3xHA tagged construct that lacks the majority of the non-conserved N’-terminal extracellular domain (3xHA::Amx^ΔECD) (Fig 4D) to specifically test the function of the highly conserved TM2 domain through overexpression experiments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Amx that only possesses the highly conserved TM2 domain is a potent inhibitor of Notch signaling.

(A, D) Schematic of proteins generated from UAS-3xHA::amx^FL and UAS-3xHA::amx^ΔECD transgenes. A 3xHA epitope (grey hexagon) was inserted after a predicted signal sequence (SS, blue box). The majority of the extracellular domain (ECD, light brown box) was removed to generate 3xHA::amx^ΔECD, consisting mostly of the TM2 domain (red box) tagged with the N-terminal 3xHA epitope. (B) Overexpression of 3xHA::Amx^FL with pannier(pnr)-GAL4 has no effect on notum morphology. The absence of bristles along the midline is due to a very mild dorsal closure defect caused by the pnr-GAL4 insertion. (C) Expression of 3xHA::Amx^FL in the posterior wing using engrailed (en)-GAL4 has no effect of the morphology of wings of adults raised at 29°C. (E) pnr-GAL4 driven overexpression of truncated Amx causes an increase in the number of micro- and macrochaete, indicative of loss of Notch mediated lateral inhibition. (F) en-GAL4 driven overexpression of 3xHA::Amx^ΔECD causes notching of the posterior wing margin. (G-H) Immunostaining of wing imaginal discs expressing full-length or truncated 3xHA::Amx. (G) 3xHA::Amx^FL expression in the posterior imaginal wing disc using en-GAL4 has no effect on NRE (Notch response element)-GFP expression, a synthetic in vivo Notch signaling reporter (G’, rainbow) and on Cut (G”, white) expression, a downstream target of Notch activation in this context. The domain expressing GAL4 is marked by RFP (red). (H) Expression of 3xHA::Amx^ΔECD decreases NRE-GFP (rainbow) expression (H’) and reduces Cut expression (H”). Scale bars = 50 μm in (G-H).

https://doi.org/10.1371/journal.pgen.1009962.g004

We first overexpressed these transgenes in the developing dorsal thorax using pannier-GAL4 (pnr-GAL4). While we observed that Amx^FL does not cause any morphological defects (Fig 4B), we found that Amx^ΔECD causes an increase in the number of mechanosensory bristles (Fig 4E), suggesting an effect on Notch mediated lateral inhibition. We then expressed the two proteins in the developing posterior compartment within wing imaginal discs using engrailed-GAL4 (en-GAL4). We observed notching of the posterior wing margin when Amx^ΔECD was expressed (Fig 4F), while no such defect was seen upon expression of Amx^FL (Fig 4C). To determine whether the wing notching caused by Amx^ΔECD overexpression was indeed due to loss of Notch signaling, we visualized Notch activation using NRE-GFP (Notch Responsive Element-Green Fluorescent Protein), a synthetic in vivo Notch signaling reporter [41], as well as immunostaining of Cut, encoded by an endogenous downstream target gene of Notch activation in this context [42]. While overexpression of Amx^FL did not affect NRE-GFP and Cut expression (Fig 4G–4G”), Amx^ΔECD caused a reduction in both NRE-GFP and Cut expression within the en-GAL4 expression domain (Fig 4H–4H”). In summary, overexpression of full-length Amx did not affect developmental events related to Notch signaling in the thorax and wing, while overexpression of a truncated form that only carries the conserved TM2 domain and its short C’-tail inhibited Notch signaling in several developmental contexts.

Truncated Amx inhibits Notch signaling at the γ-secretase mediated receptor cleavage step

To determine how Amx^ΔECD inhibits Notch signaling when ectopically overexpressed, we performed similar epistasis experiments as [25], but with improved genetic tools (Fig 5). First, we generated several new UAS constructs expressing different forms of Notch and inserted them into the identical genomic location on the 2^nd chromosome using site specific ϕC31-mediated transgenesis to avoid positional effects [43,44] (Fig 5A). In addition to transgenes that allow expression of N^ICD, N^EXT and full-length Notch (N^FL), we also generated a ligand-independent form of UAS-Notch that still depends on ADAM10 and γ-secretase by deleting several epidermal growth factor-like repeats (EGF) of the extracellular domain that contains the ligand binding domain and Lin-12/Notch Repeats (LNR) within the negative regulatory region (N^{ΔEGF1-18.LNR}) based on an earlier construct designed by [45]. When we overexpressed N^ICD, N^EXT or N^{ΔEGF1-18.LNR} in the developing wing pouch using nubbin-GAL4 (nub-GAL4), we observed increased Cut expression throughout the wing pouch, indicating ectopic Notch activation (Fig 5B–5E and 5I). Over-expression of N^FL only showed a mild increase in Cut expression in limited regions of the wing pouch (S3A Fig), likely due to its ligand-dependence. These observations are consistent with previous reports using UAS-Notch transgenic lines generated using random (P-element mediated) transgenesis technology [46].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Genetic epistasis experiments place truncated Amx at the γ-secretase cleavage step of Notch activation.

(A) Schematics and characteristics of Notch proteins made from each UAS-Notch transgenes that were generated for this study. Full-length Notch (N^FL) requires ligand binding and processing by Kuzbanian (Kuz) and γ-secretase (GS) for activation. Notch with EGF repeats and LNR domains removed (N^{ΔEGF1-18.LNR}) is not dependent on ligands but are dependent on both Kuz and GS for activation. Notch with an extracellular truncation (N^EXT) is dependent only on GS for activation. The Notch intracellular domain (N^ICD) is constitutively active. (B) Cut (green) along the dorsoventral midline within the wing pouch [marked by nub-GAL4 UAS-mCherry (red), outlined by a white dash oval) is induced by Notch signaling. LacZ is expressed as a neutral protein. (C-E) Expression of Notch constructs leads to increase in Cut expression, quantified in (I). (F) Co-overexpression of 3xHA::Amx^ΔECD has no effect of N^ICD mediated increases in Cut expression (F’, I). (G-H) 3xHA::Amx^ΔECD expression suppresses the effects of N^{ΔEGF1-18.LNR} and N^EXT on Cut expression (G’, H’, I). t-test. * = p<0.05. **** = p<0.0001. Error bars show SEM. (J-L) Overexpression of 3xHA::Amx^ΔECD causes an increase of Notch protein levels (K’) compared to overexpression of a neutral protein, LacZ (β-galactosidase) (J’). Knockdown of psn mediated by shRNA also results in mild increase Notch levels (L’), mimicking the effect of 3xHA::Amx^ΔECD. DAPI marks nuclei of wing disc cells whereas mCherry labels the dpp-GAL4 expression domain, the boundary of which is indicated by white dash lines, in J-L. Scale bars = 50 μm in (B-H). Scale bars = 5 μm in (J-L).

https://doi.org/10.1371/journal.pgen.1009962.g005

Next, we recombined nub-GAL4 and UAS-3xHA::amx^ΔECD onto the same chromosome. The resultant genetic recombinants constitutively overexpresses Amx^ΔECD in the wing pouch, with heterozygous adults exhibiting wing notching and homozygous animals completely lacking wings in adulthood (S4A and S4B Fig). We then crossed these nub-GAL4, UAS-3xHA::amx^ΔECD flies to different UAS-Notch lines to determine if Amx^ΔECD can modulate the ectopic Cut expression phenotype caused by Notch overexpression. We found that co-overexpression of Amx^ΔECD significantly suppresses the induction of Cut expression caused by N^{ΔEGF1-18.LNR} or N^EXT (Fig 5G–5I), but had no effect on the same phenotype caused by N^ICD (Fig 5F and 5I). This indicates that amx^ΔECD genetically acts at the γ-secretase-mediated cleavage step of Notch activation, consistent with previous epistasis experiments performed in embryos using amx¹ [25].

To further understand how Amx^ΔECD inhibits Notch signaling upon overexpression, we assessed the distribution of the Notch receptor through immunostaining using a monoclonal antibody that recognizes both cleaved and uncleaved forms of Notch [47]. Upon overexpression with decapentaplegic-GAL4 (dpp-GAL4), which is expressed in a limited domain within the wing pouch, we observed that Amx^ΔECD causes wing notching (S4D Fig) that is accompanied by a dramatic upregulation of Notch receptor (Figs 5K–5K’ and S5B). Wing notching was not observed when Amx^FL was overexpressed (S4C Fig), consistent with the results obtained using en-GAL4 and nub-GAL4. To test whether a similar phenotype is seen upon loss of γ-secretase function, we knocked down Presenilin (Psn, which is orthologous to human PSEN1 and PSEN2) in the wing disc and assessed its effect on Notch protein levels. When we performed RNAi against Psn using a UAS-RNAi line that had been validated in a previous study [48] with dpp-GAL4, we found that knockdown of Psn shows an accumulation of Notch similar to Amx^ΔECD over-expression but to a lesser extent (Figs 5L–5L’ and S5C). Next, to determine if the Notch receptor was accumulating at the cell surface, we performed a detergent-free immunostaining using an antibody that recognizes the extracellular domain of the Notch receptor [49]. We found that the level of Notch present at the cell surface was increased upon overexpression of Amx^ΔECD or Psn knockdown, however to a lesser extent compared to the total Notch staining (S5D–S5F’ Fig). Therefore, Notch receptors seem to accmumulate both at the cell surface and in intracellular compartments in these contexts, suggesting a defect in protein trafficking or turnover. To also test whether this Notch accumulation phenotype occurs when Notch cleavage is altered by another mechanism, we tested whether loss of ADAM10 also causes this defect. When we generated mutant clones of a null allele of kuz (kuz^e29-4) [50] using MARCM (Mosaic analysis with a repressible cell marker) [51], we did not observe any alterations in Notch levels (S6 Fig), indicating that the increase in Notch levels seen upon over-expression of Amx^ΔECD and Psn knockdown is rather specific.

To obtain additional insights into why overexpression of Amx^ΔECD inhibits Notch signaling whereas Amx^FL does not, we examined the expression and subcellular localization of the two proteins. We found that Amx^ΔECD is expressed at significantly lower levels (~10%) compared to Amx^FL (S7A and S7B Fig), suggesting that the lack of phenotype induced by Amx^FL overexpression was not due to lower expression of this protein compared to Amx^ΔECD. Furthremore, we found that while Amx^FL exhibited a memebranous localization pattern when expressed in the developing wing disc (S7C Fig), Amx^ΔECD was found in a more punctate pattern (S7D Fig). These differences in protein levels and localization may contribute to their different ability to modulate Notch signaling.

In summary, while over-expression of the full-length Amx protein did not cause any obvious defects, we serendipitously found that a truncated form of this protein that contains the small region that is highly conserved among all TM2D proteins can act as a potent inhibitor of Notch signaling when overexpressed in a several tissues including the wing disc, which is a tissue that endogenously expresses all three TM2D genes (S7E Fig). We also found through epistasis experiments that Amx^ΔECD acts at the S3 cleavage step of Notch activation, suggesting that it likely regulates γ-secretase or factors/processes that facilitate this cleavage. This is further supported by our findings that overexpression of Amx^ΔECD and knock-down of Psn, but not loss of kuz, both lead to accumulation of Notch both at the cell surface as well as intracellularly.

Null mutants of TM2D family genes have shortened lifespan

AD is an adult-onset age-dependent disease that worsens over time. While TM2D3^p.P155L has been associated with LOAD [13] and TM2D3^p.P69L has been recently reported in a proband with EOAD or frontotemporal dementia [22], there is no functional data that directly links this gene to an age-dependent neurological phenotype in any species. In Drosophila, knockdown of genes encoding subunits of the γ-secretase complex, Psn and Nicastrin (Nct), specifically in adult neurons causes a shortening of lifespan phenotype which is accompanied by an age-dependent motor function decline [48]. Considering our results suggesting that Amx is involved in a process that relates to γ-secretase (Fig 5), we asked whether loss of amx causes an age-dependent phenotype in flies (Fig 6). To asssess the effect on lifespan, we compared the longevity of amx^Δ hemizygous male flies with control flies that also carry a genomic rescue construct (amx^Δ + amx) to neutralize the effect of genetic backgroud [13]. We selected males for our analysis because amx loss does not affect male fertility, allowing us to ignore any changes in lifespan that may be caused by alterations in fecundity [52]. In contrast to the rescued control animals that exhibit a median lifespan of 51 days, the median lifespan of amx null flies is significantly shorter at 27 days (Fig 6A, p = 1.0x10^-11). We next tested whether human TM2D3 can substitute for the loss of amx in this context by introducing the humanized genomic rescue construct [13] into the amx null mutant background (amx^Δ + hTM2D3). These humanized TM2D3 animals had a median lifespan of 33 days (Fig 6A). Although this effect seems to be statically significant (p = 9.5x10^-10) and suggests that human TM2D3 may have some activity in this context, we cannot rule out the possibility that other genetic factors may be contributing to this effect considering the weakness of the phenotypic rescue.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Loss of amx causes shortening of lifespan and age-dependent neurophysiological defects.

(A) Lifespan assay shows amx^Δ animals (black, n = 247) have reduced lifespan compared to amx^Δ + amx controls (orange, n = 224). amx^Δ + human TM2D3 flies (green, n = 234) have significantly longer lifespan than amx^Δ animals, but shorter than control. Animals were reared at 25°C; Log-rank test (Mantel-Cox), **** = p<0.0001. (B) Restoring Amx expression specifically in neuronal cells rescues the reduced lifespan in amx null animals. LacZ overexpression was used as a control. Animals were reared at 25°C; Log-rank test (Mantel-Cox), **** = p<0.0001. (C) Western blot of amx^Δ; 3xHA::amx brains shows that 3xHA::Amx (predicted 35 kDa size) is expressed in the adult nervous system (arrow). Protein isolate from five brains was loaded per lane and Actin was probed as a loading control. (D) Schematization of the giant fiber electrophysiological recordings. Stimulating electrodes are inserted into the brain and recording electrodes record responses from the TTM and DLM muscles. (E-F) TTM muscles of 5d old amx^Δ mutants (black) have a response similar to amx^Δ + amx controls (orange) while DLM muscles have small but significant decrease in response probability. 3xHA::amx (blue) flies also perform as well as controls. (G-H) TTM and DLM response in 25d old amx^Δ mutants is significantly reduced. DLM response of 25d old amx^Δ + human TM2D3 (green) flies is reduced compared to controls (I). Multiple unpaired t-tests with Holm-Šídák correction for multiple comparisons. * = p<0.05. ** = p≤0.01. *** = p<0.001, **** = p<0.0001. Error bars show SEM. Additional data can be found in S12–S14 Figs.

https://doi.org/10.1371/journal.pgen.1009962.g006

Next, considering that loss of amx is phenotypically indistinguishable from the loss of amrt and bisc, we assessed whether loss of amrt or bisc causes a similar lifespan shortening phenotype in adult flies. By comparing the lifespan of amrt^Δ or bisc^Δ to their respective genomic rescue controls, we found that both mutants are significantly short lived (S8A–S8B Fig). Median lifespan of amrt^Δ was 48 days compared to 63 days in rescued control flies (p = 6.0x10^-11), and median lifespan of bisc^Δ was 32 days compared to 48 days in rescued animals (p = 3.0x10^-12). The difference in the median lifespan of rescued animals for the three mutants could be due to differences in their genetic backgrounds. In summary, flies that lack amx, amrt or bisc are short lived, suggesting that TM2D genes contribute to maintaining a normal lifespan.

Amx is expressed in the adult brain and required in neurons to maintain a normal lifespan

To further study the role of TM2D genes in aging animals, we decided to perform additional experiments on amx null mutant flies, considering the link between its human ortholog to AD. Shortened lifespan seen in amx^Δ animals could be due to a number of reasons, including defects in the nervous system or other organs functionally affected by the loss of amx. Based on publicly available microarray [37], RNA sequencing (RNA-seq) [38] and single cell RNA-sequencing (scRNA-seq) [53] datasets, amx mRNA has been detected in the adult nervous system at low levels in a subset of neurons and glia cells (S9 Fig). To determine whether Amx protein can be detected in the adult nervous system, we generated an N’-tagged amx genomic rescue construct in which a 3xHA epitope is inserted immediately after the signal sequence of the Amx protein (Fig 1A and 1B), similar to the UAS-3xHA::amx transgene discussed earlier (Fig 4A). This tagged genomic rescue construct is able to rescue the female sterility and the maternal-effect neurogenic phenotype of amx^Δ (Fig 1C and 1F), indicating the epitope tag does not have a major effect on Amx function. Next, we verified expression of 3xHA::Amx in the female ovary through immunofluorescent staining and western blot (S10 Fig). Based on immunostaining, we detected the highest signal of 3xHA::Amx in nurse cells of the ovary and observed that it localizes to the cell membrane as well as in intracellular puncta (S10A and S10B Fig). Based on western blot, we identified 3xHA::Amx at the predicted molecular weight [34.82 kDa (31.35 kDa for Amx, 3.47 kDa for the 3xHA tag)] (S10C Fig). Finally, we assessed the expression of 3xHA::Amx in the adult brain via immunostaining and western blot. While we did not observe a strong signal beyond background fluorescence based on immunofluorescence staining using an anti-HA antibody (S11 Fig), we detected 3xHA::Amx expression via western blot using brain extracts (Fig 6C). Therefore, in addition to amx mRNA that has been detected through multiple high-throughput transriptomic studies, Amx protein is also expressed in the adult brain, albeit at relatively low levels.

To determine a if amx is required neuronally, we reintroduced wild-type Amx specifically in neurons using nSyb-GAL4 [54] in amx null animals. We found that the lifespan defect of amx^Δ can be significantly suppressed by this manipulation (p<1.0x10^-16, Fig 6B). In conclusion, Amx is expressed in the adult nervous system at relatively low levels, and required in neurons to maintain a normal lifespan.

amx null mutants show progressive motor and electrophysiological defects

Considering its neuronal requirement, we next tested whether loss of amx causes age-dependent behavioral defects. Since neuronal knockdown of γ-secretase genes in flies was reported to cause age-dependent climbing defects [48], we assessed whether amx null animals also show this phenotype. We found that amx^Δ mutants show a climbing defect as early as 5 days post-eclosion, which progresses in severity as time goes on (S8C Fig), suggesting that this gene is not only required to maintain normal lifespand but also for healthspan [55].

To directly assess whether amx is required for maintenance of neural function, we performed electrophysiology on amx^Δ animals and controls. The giant fiber system is a neuronal circuitry that is required for rapid escape responses in insects and has been used as a model circuitry to assess neuronal function in a quantitative manner in adult flies [56]. This pathway can be activated through direct stimulation of the brain and outputs of the circuit can be monitored by recording the responses from tergotrochanteral (TTM) and dorsal longitudinal (DLM) muscles (Fig 6D). For a given time point, we applied multiple stimulations at different frequencies (20, 50, and 100 Hz; Figs 6E–6H and S12–S14), and measured how well these muscles respond to each stimulation. Healthy neurons can follow the stimulations showing a ‘probability of response’ that is close to 1.0, but neurons that are unhealthy show decreased probability of response, which indicates failure of the signal to travel from the brain to the muscles [57,58].

At 5 days post-eclosion, amx^Δ flies have a minor but significant failure rate at later stimulations at 100 Hz in the DLM muscle compared to control animals (rescued with a wild-type amx [amx^Δ + amx] or N’-3xHA tagged amx [amx^Δ + 3xHA::amx]; Fig 6F). Additionally, amx^Δ animals show only slight, early stimulation failure rates in the TTM at 50 and 100 Hz at 5 days (Figs 6E and S12D). At 15 days post-eclosion, amx^Δ animals still only show mild failure rate in the TTM at 100 Hz (S13E Fig) but begin to show a significant increase in failure rate compared to control animal recordings from the DLM at 50 and 100 Hz (S13D and S13F Fig). At 25 days post-eclosion, the defects in the DLM increase in severity for 50 and 100 Hz compared to rescued controls (Figs 6H and S14D). At this time point, the TTM also begins to show a significant failure rate in amx^Δ animals at 100Hz compared to rescued animals but not at lower frequencies (Figs 6G, S14A and S14C). Interestingly, unlike in the lifespan assay in which the human TM2D3 did not dramatically suppress the amx^Δ mutant phenotype (Fig 6A) and in the fertility/neurogenesis assay in which the human TM2D3 showed ~50% activity of the fly Amx protein (Fig 1C), we found that human TM2D3 is able to significantly suppress the electrophysiological defects close to the level of fly Amx (Figs 6E–6H and S12–S14). In summary, loss of amx causes an age-dependent decline in neuronal function, and this defect can be fully rescued by both fly and human TM2D3, indicating an evolutionarily conserved role of this gene in healthy aging.

Protein alignment and phylogenetic analysis of TM2D family genes

For informatic analysis, we downloaded the FASTA files corresponding to the following proteins from Uniprot (https://www.uniprot.org/). Human (Homo sapiens) TM2D1 (Q9BX74), TM2D2 (Q9BX73), TM2D3 (Q9BRN9); Mouse (Mus musculus) Tm2d1 (Q99MB3), Tm2d2 (Q8R0I4), Tm2d3 (Q8BJ83); Frog (Xenopus tropicalis) tm2d1 (A0A6I8QKF8), tm2d2 (Q5M8E3), tm2d3 (Q07FZ2); Zebrafish (Danio rerio) tm2d1 (A5PLH4), tm2d2 (Q6DHN3), tm2d3 (A5PLF5); Fly (Drosophila melanogaster) CG10795 (Q9W2H1), CG11103 (Q9VY86), amx (Q9U4H5); Worm (Caenorhabditis elegans) Y66D12A.21 (Q95PJ8), C02F5.13 (P61228), C41D11.9 (Q95QZ5). Protein alignment was performed using Clustal Omega using default parameters (http://www.clustal.org/omega/). Phylogenetic analysis was performed using MAFFT (multiple alignment using fast Fourier transform) version 7 (https://mafft.cbrc.jp/alignment/server/) through UPGMA (unweighted pair group method with arithmetic mean) and Neighbor-Joining methods using default parameters.

Drosophila strains and fly husbandry

Drosophila melanogaster stocks used in this study are listed in S1 Table. Some strains were generated in house for this study (see below), and others were obtained from Bloomington Drosophila Stock Center (BDSC) and other sources. Flies were kept on standard media and maintained at room temperature (21–23°C). Crosses were performed at 25°C in an incubator unless otherwise stated. Genotypes of the flies shown in each figure panel are listed in S2 Table.

Generation of amx overexpression transgenes

To generate the untagged UAS-amx transgene, we PCR amplified the open reading frame (ORF) of amx derived from the cDNA clone GH02974 (Genbank ID: AF181623, Drosophila Genomics Resource Center, Indiana University) with NotI and XbaI restriction sites on the 5’ and 3’ end, respectively, and subcloned this fragment into pUASTattB [93]. The cDNA templates used to generate the UAS constructs to overexpress 3xHA epitope tagged full-length (UAS-3xHA::amx^FL) and truncated (UAS-3xHA::amx^ΔECD) Amx were synthesized by Genewiz. The amx ORF used for these constructs is identical to the GH02974 clone, which is also identical to the coding sequence of the genomic rescue construct we reported in [13]. A 3xHA epitope sequence (N’-YPYDVPDYAGYPYDVPDYAGSYPYDVPDYA-C’) was inserted after the signal sequence (SS) as predicted by SignalP 4.1 (http://www.cbs.dtu.dk/services/SignalP-4.1) and Phobius (http://phobius.sbc.su.se). A 968 bp fragment corresponding to 3xHA tagged full-length Amx was synthesized with NotI and XbaI restriction sites on the 5’ and 3’ end respectively. Furthermore, a 464 bp fragment corresponding to 3xHA::Amx^ΔECD was designed based on the full-length 3xHA::Amx construct, but with the majority of the endogenous Amx N’-end removed leaving the SS, epitope tag, and TM2 domains intact. We subcloned the synthesized fragments into pUASTattB [93] via restriction enzyme cloning utilizing the NotI and XbaI cut sites. We then injected the final plasmids into VK37 (PBac{y[+]-attP}VK00037) attP docking site [43] and injected animals were crossed to y w and screened for the presence of the white⁺ marker encoded in the pUASTattB backbone in a white^- background. Positive animals were then crossed to SM6a to create a balanced stock. One line for each transgenic construct was chosen for this study.

Generation of a N’-tagged amx genomic rescue transgene

The 3xHA::amx genomic rescue transgene was designed and constructed based on the non-tagged genomic rescue construct we reported in [13]. A ~3.3kb region including the full amx gene has the ability to fully rescue the maternal effect neurogenic phenotype of amx¹ [13] and amx^Δ (this study). We inserted a 3xHA sequence into the pattB-amx genomic rescue construct described in [13] via NEBuilder HiFi DNA Assembly. 3xHA was inserted after the predicted signal sequence (N’-ATMRLQRQCIVVNMRSAIVLIMIFVLTGIRNSET-C’) to tag Amx at its N’. pUASTattB-3xHA::amx^FL described above was used as a template to amplify and add appropriate homology arms to the SS-3xHA::Amx DNA sequence with the primers 5’-CCCCGCTCTATCTGACCAAAGCCACCATGAGGCTCCAACGAC-3’ and 5’-AAAACTAAACTAAGAACGGACTACTATATGTAAAGTGAGCCATCCGC-3’ using Q5 High-Fidelity 2X Master Mix (M0492S, NEB). These and other Oligo DNA were synthesized by Sigma Aldrich. The section of pattB-amx plasmid containing the amx regulatory elements was linearized by PCR using Q5 polymerase and primers 5’-CGTTGGAGCCTCATGGTGGCTTTGGTCAGATAGAGCG-3’ and 5’-GCTCACTTTACATATAGTAGTCCGTTCTTAGTTTAGTTTTACAGGGGT-3’, which add appropriate homology arms to allow for assembly with the 3xHA::Amx fragment. The construct was assembled following the protocol described by NEBuilder HiFi DNA Assembly Master Mix (NEB Catalog #E2621S) and confirmed by Sanger sequencing. The validated construct was injected into embryos expressing ϕC31 integrase with a 2^nd chromosome attP docking site (VK37) [43]. We also injected a empty pattB plasmid into the same docking site to generate a control like (pattB[w+]). Transgenic flies were isolated based on eye color (w⁺ encoded by the mini-white gene in pattB vector) and balanced over SM6a.

Generation of the amrt^Δ mutant

The amrt (CG11103) null allele was created based on a method we described in [33]. We selected gRNA (guide RNA) target sites using CRISPR Optimal Target Finder (http://targetfinder.flycrispr.neuro.brown.edu/). sgRNAs (single gRNA) expressing plasmid was generated using the pCFD3-dU6:3 gRNA vector (Plasmid ID: #49410, Addgene) as described in [94]. Oligo DNA to generate the upstream sgRNA plasmid were 5’-GTCGCGCTGCGTGCCTGTATCGCT-3’ and 5’-AAACAGCGATACAGGCACGCAGCG-3’. Oligo DNA to generate the downstream sgRNA plasmid were 5’-GTCGTCCTCGGGCAGCAAATTGT-C’ and 5’-AAACACAATTTGCTGCCCGAGGA-3’. Donor plasmid containing the yellow^body+ marker (y^body+) to be integrated into the amrt locus via HDR was generated through NEBuilder HiFi DNA Assembly. Primers amplifying and adding homology arms to a pBH vector [95] backbone were 5’-GATGCTGTTAGACTAACGGTGTATATCTAGAGCCGTCCCGTCAAG-3’ and 5’-CCGCAAGCAATGGCCAAACTGGGTCCTCGAGTCGACGTTG-3’. Primers for amplifying the ybody⁺ insert and adding homology arms from P{ybody+} plasmid [33] were 5’-CAGACAACACGATGCCCAAGCGGATCGCTTGATGTTGTTTTGTTTTG-3’ and 5’-TGCCGTCCTCGGGCAGCAAATGGAAGGAACCTGCAGGTCAACG-3’. All plasmids were verified by Sanger sequencing. y, w, iso#6(X); attP2{nos-Cas9} [33] embryos were injected with 25 ng/ul concentration of each sgRNA plasmid mixed with 150 ng/ul of the y^body+ donor plasmid. Resulting adults were crossed to y w animals and offspring screened for the presence of the y^body+ marker (dark body color instead of yellow body). Positive animals were crossed to the FM7c balancer to establish the lines and were molecularly genotyped (see below).

Generation of the bisc^Δ mutant

The bisc (CG10795) null allele was generated using the same strategy discussed above that was used to generate the amrt null allele. Oligo DNAs to generate the upstream sgRNA plasmid were 5’-GTCGTATGAGGGACCATGTACAT-3’ and 5’-AAACATGTACATGGTCCCTCATA-3’. Oligo DNA to generate the downstream sgRNA plasmid were 5’-GTCGAGGCGGTGGTGGTGTTCTGT-3’ and 5’-AAACACAGAACACCACCACCGCCT-3’. Primers amplifying and adding homology arms to a pBH vector backbone were 5’-GTAGACACACGGCATAGATGGTATATCTAGAGCCGTCCCGTC-3’ and 5’-GAAGAAGTTGACAATGTGTTGGGTCCTCGAGTCGACGTTG-3’. Primers for amplifying the ybody⁺ insert and adding homology arms from p{ybody+} plasmid were

5’-AATTGTATGAGGGACCATGTAGGATCGCTTGATGTTGTTTTG-3’ and 5’-GTTGACCTGCAGGTTCCTGTTGGCGGGAGCTCTTTCTC-3’. All plasmids were verified by Sanger sequencing. y, w; iso#2(2); attP2{nos-Cas9} embryos were injected with 25 ng/ul concentration of each sgRNA plasmid mixed with 150 ng/ul of the y^body+ donor plasmid. Resulting adults were crossed to y, w animals and offspring screened for the presence of the y^body+ marker. Positive animals were crossed to the SM6a balancer to establish the lines and were molecularly genotyped (see below).

Detection of TM2D mRNA expression through RT-PCR

The presence or absence of mRNA corresponding to TM2D genes were determined using RT-PCR (Reverse Transcription-Polymerase Chain Reaction). Whole body RNA was isolated from adult animals through standard TRIzol/chloroform RNA extraction protocol. We prepared cDNA with iScript Reverse Transcription Supermix, (#1708840, BioRad). PCR was done using Q5 polymerase (M0492S, NEB). Primers used to detect the presence of amx cDNA were 5’-TCCCCGCTCTATCTGACCAA-3’ and 5’-GCTCTGTTGCCACATTTCCG-3’. Primers to detect the presence of amrt cDNA were 5’-CTACGGACTACTGGCGTTCC-3’ and 5’-CCCGTTTGACCGAGACAGAA-3’. Primers to detect the presence of bisc cDNA were 5’-CCCCGCGAACTGCAATAAAC-3’ and 5’-CACAACCTGCAGGGCTATCA-3’. Primers targeting TM2D cDNA were annealed at 68⁰ C, extended for 30s at 72⁰ C for 30 cycles. Primers for control gene rp49 were 5’-TCTGCATGAGCAGGACCTC-3’ and 5’-CGGTTACGGATCGAACAAG-3’ [96]; annealed at 64⁰ C and extended 30s for 30 cycles. The same protocol was used to assess the mRNA expression of TM2D genes in wing imaginal discs.

Generation of an amrt genomic rescue construct

Genomic DNA was isolated from y, w, iso#6; +/+; attP2{nos-Cas9} animals using PureLink Genomic DNA Mini Kit (Cat no. K1820-01, Invitrogen). Genomic region fully containing the amrt (CG11103) locus and neighboring sequences was amplified by PCR using the primers 5’-TATATACTCGAGcgcgaaactttcgatttcc-3’ and 5’-TATATAGAATTCatcgaatgtagagatgggc-3’ (small letters indicate annealing region) which added EcoRI and XhoI restriction sites for follow-up cloning into the pattB vector [93]. The pattB-amrt plasmid was injected into VK37 [43] docking site on the second chromosome. Eclosed animals were crossed to y, w and screened for the w⁺ marker in the subsequent generation. Positive animals were then crossed to SM6a to create a balanced stock.

Generation of biscotti::GFP fosmid transgenic line

A ~40kb genomic fosmid construct in which bisc is C’-tagged with GFP and other epitopes [36] (bisc::GFP, FlyFos021003, Drosophila TransgeneOme Resource ID: CBGtg9060C1139D) was obtained from Source BioScience. Bacterial colonies were provided to GenetiVision Corp. for DNA preparation, injection into VK33 (PBac{y[+]-attP}VK00033) on the third chromosome [43], selection, and balancing of fly lines using the TM3, Sb balancer. Three independent lines were generated and all three behaved in a similar manner. One line was chosen for the experiments performed in this study.

Creation of Notch overexpression transgenic lines

All transgenic constructs were generated by Gateway (Thermo Fisher Scientific) cloning into the pUASg-HA.attB plasmid [93]. First, we generated Gateway compatible plasmid that contains the full-length Notch open reading frame. We subcloned the full-length Notch (N^FL) open reading frame into the pDONR223 plasmid based on a cDNA clone provided by Dr. Spyros Artavanis-Tsakonas [97], which was mediated by a Gateway reaction using BP clonase II (Thermo Fisher Scientific, #11789100) following a PCR reaction and addition of attB sites to the amplicon [98,99]. Truncated Notch constructs were generated by Q5 site-directed mutagenesis (NEB) with the following primers:

N^EXT: 5’-GCGGCCAAACATCAGCTG-3’ and 5’-GTGCATTTTGTTAATCCAAAAACAAATCC-3’.

N^ICD: 5’-GTCTTGAGTACGCAAAGAAAG-3’ and 5’-CATGGTGAAGCCTGCTTT-3’.

N^{ΔEGF1-18.LNR} (two-step mutagenesis): 5’-CTGAGCGATGTGGACGAGTGCGCATCGAAT-3’ and 5’-CAACGCGGTATCAGTTCC-3’ followed by mutagenesis with 5’-AACAAGACCCAGTCACCG-3’ and 5’-CATGGCACGTTGTTGCTC-3’.

All constructs were fully sequenced (Sanger), and cloned into pUASg-HA.attB via LR clonase II (Thermo Fisher Scientific, #11791020). All expression constructs were inserted into VK37 integration site [43]. Two transgenic lines were established for each and one line each was used for this study.

Embryo collection, staining and imaging

Embryo collection, staining and imaging was performed as previously described [13]. In brief, virgin females that are homozygous for each or all TM2D gene mutations, with or without genomic rescue constructs, were crossed to males flies of the same genotype or Canton-S males, and allowed to mate in a vial for 24 hrs. Flies were then transferred to a bottle with a grape juice plate supplemented with active yeast and allowed to lay eggs overnight. Embryos were then gently collected using a paint brush and their chorions were removed by 1.5 minute incubation in 66% bleach. Dechorionated embryos were then washed with water and fixed in 4% paraformaldehyde/PBS(phosphate buffered saline)/n-heptane solution for 30 minutes at room temperature. Fixed embryos were washed and stored in 100% methanol at -20°C until use. Fixed embryos were rehydrated and rinsed with 0.03% Triton-X in PBS (PBST). A primary antibody to label neuronal nuclei [anti-Elav, rat monoclonal (7E8A10) [100], 1:200, Developmental Studies Hybridoma Bank (DSHB)] was applied in a solution of PBST/5% normal donkey serum (NDS)/0.1% NaN₃ overnight at 4°C. Embryos were further washed with PBST upon removal of the primary antibody, and a secondary antibody (donkey anti-rat-Alexa488, 1:500; Jackson ImmunoResearch #712-545-153) was applied for 1 hour at room temperature. Stained embryos were washed in PBST and mounted onto glass slides using Vectashield with DAPI (4’,6-diamidino-2-phenylindole, Vector labs). Embryos were imaged using Ti2E Spinning Disc confocal microscope (Nikon) and images analyzed using NIS software (Nikon).

Egg hatching assay

Egg hatching assay was performed as previously described [13]. Embryos were collected and dechorionated as described above. Dechorionated embryos were suspended in PBS and placed in 12-well cell culture dishes. The dishes were then placed in a 25°C incubator for 24 hours. The ratio of hatched to unhatched embryos after the 24 hour period was recorded for each genotype and used to calculate hatching rate (%). This was repeated at least 3 times and statistical analysis and graph generation was performed using GraphPad Prism 9.0 software. We performed one-way ANOVA followed by Dunnett test or t-test. **** = p-value ≤0.0001.

Imaging of adult flies

Heads and wings from adult flies were removed from the body using fine dissection scissors and imaged directly using a MZ16 microscope (Leica) with attached Microfire camera (Optronics) using ImagePro Plus 5.0 acquisition software (Media Cybernetics). Extended focus function was used to obtain deep focus images out of Z-stack images. Imaging of the dorsal thoraxes (nota) were described in [101] with slight modifications. In brief, legs, head, and abdomen were removed from thoraxes with fine dissection scissors. Then, the dissected thoraxes were then placed in 10% KOH at 95°C. for 10 min to dissolve soft tissue. Thoraxes were then further trimmed with scissors prior to imaging and mounted on a glass slide with spacers using 75% glycerol/25% ethanol solution. Photos were taken using microscope system and imaging software as above.

Immunostaining and imaging of Drosophila wing imaginal discs, ovaries or brains

Wing discs from wandering larvae or ovaries and brain from adult flies were dissected in 1x PBS and fixed for 30 minutes in 4% paraformaldehyde in PBS. Tissues were then washed with 0.2% PBST for total Notch staining or with cell-culture grade PBS (Thermo Fisher Scientific, #10010023) for extracellular Notch staining. All antibodies used in this study are listed in S1 Table. Primary antibodies [mouse anti-Notch intracellular domain (NICD) (1:50; DSHB, C17.9C6), mouse anti-Cut (1:100; DSHB, 2B10), rat anti-HA (1:100, Sigma-Aldrich, 11867423001), or mouse anti-NECD (1:100; DSHB, C458.2H)] were applied in 0.2% PBST or PBS with 5% NDS/0.1% NaN₃ overnight at 4°C. Tissue was then washed with 0.2% PBST or PBS (3 times, 15 minutes each) and secondary antibodies/stains [donkey anti-rat IgG-Cy3 (1:500; Jackson ImmunoResearch #712-165-153), donkey anti-mouse IgG-Alexa-647 (1:500; Jackson ImmunoResearch #715-605-151), donkey anti-mouse Alexa-488 (1:500; Jackson ImmunoResearch #715-545-151), and/or Alexa-488 Phalloidin (1:1000; ThermoFisher A12379)] were applied in 0.2% PBST+NDS or PBS+NDS for 2 hours at room temperature. Tissues were further washed with 0.2% PBST or PBS and mounted in Vectashield with DAPI (Vector labs). Images were taken with LSM 710 Confocal Microscope (Zeiss).

Notch epistasis assay

UAS-Notch lines were crossed to either nub-GAL4, UAS-CD8::mCherry (control) or nub-GAL4, UAS-amx^ΔECD flies. Wing discs from 3^rd instar larvae were dissected out, fixed washed and stained for Cut as described earlier. Cut intensity within the wing pouch was quantified using ImageJ (https://imagej.nih.gov/ij/). Graph generation and statistical analysis was performed with GraphPad Prism software version 9.0. One way t-test was used to compare experiments. * = p≤0.05. **** = p≤0.0001.

MARCM analysis

The following fly lines were used for MARCM [51] analysis of kuz mutant clones:

hsFLP; tub-Gal80^ts, FRT40A/CyO; tub-GAL4, UAS-GFP/TM6b, Tb [102]

kuz^e29-4, FRT40A/CyO [50]

Flies were crossed and maintained at 25°C. First-instar larvae (36–48 hours after egg laying) were heat shocked for 40 minutes twice a day at 37°C. Wing discs were dissected in 1×PBS from wandering larvae (108–120 hours after egg laying). Discs were then fixed in 4% paraformaldehyde in PBS for 20 minutes at room temperature. Fixed discs were washed with 0.2% PBST. A mouse anti-Notch primary antibody (1:40; DSHB, C17.9C6, raised against the intracellular domain) was applied in 0.2% PBST with 5% NDS/0.1% NaN₃ overnight at 4°C. Discs were further washed with 0.2% PBST after primary antibody staining. A donkey anti-mouse-Cy3 secondary antibody (1:500; Jackson ImmunoResearch #715-165-151) in 0.2% PBST was applied for 2 hours at room temperature. Stained discs were washed with 0.2% PBST and mounted with Vectashield with DAPI (Vector labs). Fluorescence Images were taken by LSM 710 Confocal Microscope (Zeiss).

Western blot of HA tagged Amx

To determine whether 3xHA::Amx is expressed in adult brains or ovaries, we performed western blot. Adult brains from y, w; VK37{pattB-3xHA::amx} flies were dissected out as described in [103]. For ovaries, we mated the y, w; VK37{pattB-3xHA::amx} females flies to male flies of the same genotype while supplying plenty of active yeast to stimulate oogenesis for two days prior to protein isolation. Ovaries were then dissected out from the abdomen in cold (4°C) PBS. Dissected brains and ovaries were rinsed with cold PBS and placed immediately in cold 8M urea lysis buffer (8M urea, 10% glycerol, 0.5% SDS, 5% β-mercaptoethanol) with Halt Protease Inhibitor Cocktail 100X (Thermo Scientific, #78430) added before lysis. We homogenized the brains or ovaries were via pestle in 15 uL of 8M urea lysis buffer. Homogenate was incubated for 30 minutes on ice, and 2x Laemmli Sample Buffer (Bio-Rad, #1610737) was added prior to gel loading. Best results were obtained when avoiding heating/boiling protein sample. Homogenate was loaded directly into Mini-PROTEAN TGX 4–20% Precast Gels (Bio-Rad, #4561094). SDS-PAGE (Sodium Dodecyl Sulphate–PolyAcrylamide Gel Electrophoresis) was run for 30 minutes at 120V in Tris/Glycine/SDS buffer (BioRad, #1610732). Protein was then transferred onto PVDF membrane using Bio-Rad TransBlot Turbo system using the low molecular weight protocol. The membrane was blocked with 5% skim milk in 0.5% Tween-20/Tris-Buffered Saline (TBST) for 1 hour at room temperature and then washed 3 times for 5 minutes with TBST. A primary antibody (rat anti-HA, 1:1,000, Sigma, 3F10) was diluted in 5% fetal bovine serum/TBST and the membrane was incubated overnight at 4°C. The membrane was again washed with TBST and a secondary antibody (donkey anti-rat HRP, 1:5,000, Jackson ImmunoResearch, #712-035-150) was applied in 5% milk/TBST for 2 hours at room temperature. The membrane was washed again and imaged using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific, 34096) and ChemiDoc imaging system (Bio-Rad) using default settings. The same protocol was used to compare the levels of overexpressed 3xHA::Amx^FL or 3xHA::Amx^ΔECD from adult brains dissected from elav-GAL4/+; UAS-3xHA::Amx^{FL or ΔECD} animals.

Longevity assay

To determine whether loss of amx, amrt or bisc causes lifespan defects, we compared the longevity of male flies that lack each gene (amx^Δ, amrt^Δ or bisc^Δ) to respective control flies in which the null allele has been rescued with genomic constructs (VK37{amx}, VK37{amrt} or VK33{bisc::GFP}). For amx, we also tested whether N’-3xHA tagged Amx (y, w, amx^Δ; VK37 {3xHA::amx}) or human TM2D3 expressed under the control of fly amx regulatory elements (y, w, amx^Δ; VK37 {TM2D3}) can also rescue the mutant phenotype. Flies were reared and collected as described in [104]. Ten animals were housed together in a single vial and flies were flipped to a new vial with fresh food every 2–3 days. Vials were kept in a 25°C incubator with a 12 hr light/dark cycle. Dead flies were recorded after every vial flip. Generation of the Kaplan-Meier curve and statistical analysis was performed with GraphPad Prism 9.0. We applied log-rank test (Mantel-Cox), **** = p≤0.0001.

Climbing assay

8–10 flies that were age matched were placed in an empty vial and flies were tapped down to the bottom. The number of flies that successfully climbed past a 7 cm line from the bottom of the vial within 15 seconds were recorded. This was repeated 3 times for each group of flies. The graph was generated using GraphPad Prism 9.0.

Electrophysiological recordings of the giant fiber system

Electrophysiological recordings of the giant fiber system were performed with a protocol modified from [105]. Flies were first anesthetized on ice then transferred to a petri dish filled with soft dental wax; wings and legs were mounted in wax, ventral side down, using forceps. Five electrolytically sharpened tungsten electrodes were used: two for stimulating the giant fiber, one as a reference electrode, and two for recording from the TTM and DLM, respectively. To activate the giant fiber, two sharp tungsten electrodes were inserted into each eye and voltage stimulation was applied at different frequencies ranging from 0.5 Hz to 100 Hz. DLM and TTM responses were measured through the two electrodes implanted in the DLM and TTM.

For each adult fly, prior to applying high frequency stimulation on the giant fiber, low frequency stimulations at 0.5 Hz were applied after placing the two recording electrodes in TTM and DLM to ensure that the electrodes are recording from the proper muscles (the latency of responses for TTM: 0.8 ms and for DLM: 1.2 ms [105]. For the actual experiments, high frequency train stimulations of 20 pulses were delivered to the giant fiber at 20, 50 and 100 Hz in random order. Ten times repetitive stimulations were applied for each particular frequency train, interspersed with few minutes rests between two trains of stimuli (for 20 Hz, 1 minute resting; 50 Hz 2 minutes and 100 Hz 3 minutes). 0.5 Hz stimulations were used again after high frequency stimulation to confirm that electrodes were still in the proper muscle. The aforementioned process was considered as one biological sample. Stimuli of the crossing electrodes were fixed at a duration of 10 microseconds at 10–13 V of amplitudes through a stimulus isolation unit (Digitimer Ltd, model DS2A) and the frequency of train stimuli was controlled by LabChart Pro-8 acquisition software (ADInstruments). A microelectrode amplifier (A-M system, Model 1800) was used for all recordings. PowerLab 4/35 (ADInstruments) was used for data acquisition. The probability of responses for one biological sample, under particular frequency of giant fiber stimulation, due to a particular stimulus, was calculated from the proportion of successful responses (out of 10) for both TTM and DLM pathways. The difference of ‘probability of responses’ between control and experimental samples (p-value) for each stimuli were calculated by multiple unpaired t-tests with Holm-Šídák correction for multiple comparisons using Graph Pad Prism 9.0. n.s. = not significant. * = p≤0.05. ** = p≤0.01. *** = p≤0.001. **** = p≤0.0001.