Gαq splice variants mediate phototransduction, rhodopsin synthesis, and retinal integrity in Drosophila

Heterotrimeric G proteins mediate a variety of signaling processes by coupling G protein–coupled receptors to intracellular effector molecules. In Drosophila, the Gαq gene encodes several Gαq splice variants, with the Gαq1 isoform protein playing a major role in fly phototransduction. However, Gαq1 null mutant flies still exhibit a residual light response, indicating that other Gαq splice variants or additional Gq α subunits are involved in phototransduction. Here, we isolated a mutant fly with no detectable light responses, decreased rhodopsin (Rh) levels, and rapid retinal degeneration. Using electrophysiological and genetic studies, biochemical assays, immunoblotting, real-time RT-PCR, and EM analysis, we found that mutations in the Gαq gene disrupt light responses and demonstrate that the Gαq3 isoform protein is responsible for the residual light response in Gαq1 null mutants. Moreover, we report that Gαq3 mediates rhodopsin synthesis. Depletion of all Gαq splice variants led to rapid light-dependent retinal degeneration, due to the formation stable Rh1-arrestin 2 (Arr2) complexes. Our findings clarify essential roles of several different Gαq splice variants in phototransduction and retinal integrity in Drosophila and reveal that Gαq3 functions in rhodopsin synthesis.

Heterotrimeric G proteins and G protein-coupled receptors play pivotal roles in mediating a variety of extracellular signals to intracellular signaling pathways, such as hormones, neurotransmitters, peptides, and sensory stimuli (1,2). In the Drosophila visual system, light stimulation activates the major rhodopsin (Rh1) to form metarhodopsin, which in turn activates heterotrimeric G proteins and norpA gene-encoded phospholipase C (PLC␤) 4 (3). Activated PLC catalyzes phosphatidylinositol 4,5-bisphosphate to generate diacylglycerol (DAG) and inositol 1,4,5-triphosphate (IP3) (4). IP3 induces the release of Ca 2ϩ from intracellular Ca 2ϩ stores, whereas both DAG and IP3 may trigger extracellular Ca 2ϩ influx by opening transient receptor potential (Trp) and transient receptor potential-like (TrpL) channels on the cell membrane (5)(6)(7)(8). The G␣q gene encodes several G␣q splice variants, among which the G␣q-RD variant generates the G␣q1 isoform protein, and other splice variants generate the G␣q3 isoform protein (9). Although both strong alleles of norpA and trpl;trp double mutants show completely abolished photoresponses (4,10,11), the G␣q1 isoform null mutant allele (G␣q 961 ) still displays a residual light response (12). These data indicate that other G␣q splice variants, or the Gq ␣ subunits encoded by additional genes, contribute to the residual light responses in G␣q1 null mutants.
Intracellular Ca 2ϩ homeostasis controlled by Gq signaling is also essential for photoreceptor cell survival (13). Mutations in phototransduction cascade components, such as those in trp and norpA, prevent normal light-induced Ca 2ϩ influx, resulting in stable Rh1/Arr2 complex formation and severe rapid light-dependent retinal degeneration (14,15). Disruption of stable Rh1/Arr2 complexes by genetic removal of Arr2 or suppression of Rh1 endocytosis can suppress the retinal degeneration either in norpA or trp mutant flies (15,16). Rh1/Arr2 complex formation is thought to attribute to impaired Ca 2ϩ influx-activated CaM kinase II, which usually phosphorylates Arr2 to release Arr2 from Rh1 (17,18). However, neither G␣q 1 nor G␣q 961 mutants undergo rapid retinal degeneration (12,19), exhibiting only slight retinal degeneration after keeping them in 12-h light/12-h dark cycles for 21 days (12). The disagreeing retinal degeneration phenotype between G␣q and norpA mutant is therefore unclear.
Here, we isolate a mutant fly with no detectable light responses and reveal that mutations in the G␣q gene cause the defective light responses. We demonstrate that G␣q3 is responsible for the residual light response in G␣q1 null mutants and show that depletion of all G␣q splice variants results in rapid light-dependent retinal degeneration due to formation of stable Rh1/Arr2 complexes. In addition, we reveal that G␣q3 plays essential roles in Rh1 synthesis. Our study clarify the essential role of different G␣q splice variants in fly phototransduction, retinal degeneration, and rhodopsin synthesis.

Isolation of a fly mutant with no detectable responses to light stimulation
To characterize the components of the phototransduction machinery, we obtained a collection of transgenic transposon   1A). Using the inverse PCR technique, we identified that the p[GawB] element was inserted into the 18C3 chromosomal region located on the X chromosome. To eliminate extra mutations in the genetic background, we backcrossed the mutant with the WT w 1118 strain (based on the ERG phenotype) for eight generations and refer to the out-crossed mutant as nlr (no detectable light response) mutant. Unexpectedly, nlr mutants did not contain any p[GawB] element insertions, indicating the abolished ERG response in the nlr mutants is due to mutations in the genetic background, and not p[GawB] element insertion.
The significantly reduced ERG response in nlr mutants could be due to a defective rhabdomere structure or reflect deficits in the phototransduction cascade. To distinguish between these possibilities, we first performed an electron microscopy (EM) study to examine the rhabdomere structure of newly enclosed adult flies. However, EM images did not reveal any morphological defects in nlr mutant rhabdomeres (Fig. 1B). Intracellular recording found that light stimulation was unable to evoke any detectable responses in nlr mutant photoreceptors (Fig. 1C). These data indicated that the defective light response in nlr mutants was due to abnormalities in the phototransduction cascade. We performed Western blot analysis to examine the protein levels of components and regulators in the phototransduction cascade, including major rhodopsin (Rh1), G␣q, PLC, TRP, INAD, and Arr2. Interestingly, protein levels for all G␣q isoforms recognized by anti-G␣q-N antibodies were significantly reduced (100 Ϯ 2.4 versus 1.4 Ϯ 0.9%, p Ͻ 0.0001, t test; Fig. 1D-E). Meanwhile, a partial reduction of Rh1 (100 Ϯ 12.2 versus 38.3 Ϯ 5.5%, p ϭ 0.0014, t test) was found in nlr mutants, whereas the other visual molecules examined were comparable with WT flies (Fig. 1D-E). These results suggest that the defective light responses in nlr mutants might be due to the absence of G␣q protein.

Mutations in G␣q gene are responsible for defective light responses in nlr mutants
To identify the mutations in nlr flies that are responsible for the defective light-response phenotype, we first mapped the mutations to the 49B5-49B12 chromosomal region based on the ERG phenotype covered by the deficient chromosome Df(2R)Exel7121 (missing 49B5 to 49B12, Fig. 2, A and B). This result further supports the conclusion that the abnormal ERG phenotype in nlr mutants is not due to p[GawB] element insertion. Next, we further narrowed the mutation to the 49B8-49B10 region based on the ERG phenotype covered by the deficient chromosome Df(2R)G␣q1.3, which removes G␣q, CG30054, CG17760, muskelin, and part of the CG33792 genes (Fig. 2, A and B). Next, we generated clones of Df(2R)G␣q1.3covered gene nulls in the retina through ey-FLP-induced FRT recombination in the deficiency line, Df(2R)G␣q1. 3. We found that this fly lacked G␣q protein and showed an abolished light response (Fig. 2, C and D), indicating that the defective light

G␣q splice variants regulate visual function
response observed in nlr mutants is contributed to by the mutations located in the deficient chromosome Df(2R)G␣q1.3-covered region.
Among the genes covered by the deficient chromosome Df(2R)G␣q1.3, the G␣q gene plays an essential role in Drosophila phototransduction (12,20). To test whether G␣q gene mutations contribute to the significantly reduced light responses in nlr mutants, we obtained a G␣q 221c mutant allele that removes a 359-bp fragment around the translation start site of all G␣q splice variants (21). Because G␣q 221c homozygous mutations are lethal, we generated G␣q 221c null mutant clones in the retina through ey-FLP-induced FRT recombination. This fly was also absent of G␣q protein (100 Ϯ 16.4 versus 2.1 Ϯ 0.7%, p ϭ 0.0005, t test) and displayed an abolished light response (Fig. 2, E and F). Next, we recombined nlr with G␣q 221c mutant flies and found that nlr/G␣q 221c flies showed no detectable responses to light stimulus and an absence of G␣q protein (Fig. 2, G and H). These data demonstrate that mutations in the G␣q gene are responsible for the defective light responses in nlr mutants.

Identification of new G␣q1 isoform mutation in nlr mutant
The G␣q gene encodes several G␣q splice variants (Fig. 3A), and the splice variant G␣q1 (also named as G␣q-PD, AAM68631) has been shown to play a major role in Drosophila phototransduction (12,20). Thus, we wondered whether nlr mutants contain any mutations in the G␣q gene. Indeed, subsequent DNA sequencing revealed a mutation (5501 T/A ) in exon 7 of the G␣q gene in nlr mutant flies (Fig. 3A), which is within the G␣q1 isoform but not included in other G␣q splice variants. This mutation corresponds to a missense mutation (303 V/D ) in the GTPase domain of G␣q1 (amino acids 247-359).
To examine whether this mutation (5501 T/A ) disrupts G␣q1 function, we combined the nlr mutant allele with a G␣q hypomorphic allele (G␣q 1 ), which changes a Gly to Ala in a splice acceptor site causing the use of a cryptic splice site 9 nucleotides downstream and an in-frame deletion of 3 codons encompassing amino acid residues 154 -156 (20). Meanwhile, we combined the nlr mutant allele with a G␣q1 isoform null mutant allele (G␣q 961 ), which contains a mutation (961 C/T ) in exon 4 and causes a nonsense mutation (Arg-117 to stop codon) of G␣q1 exclusively (12). Western blotting showed that G␣q protein levels were significantly reduced in both nlr/G␣q 1 and nlr/ G␣q 961 flies (Fig. 3B). ERG recording further revealed that both nlr/G␣q 1 and nlr/G␣q 961 flies exhibited significantly reduced light responses to saturated light stimulation, simi- To further validate that the G␣q V303D mutation in G␣q1 largely contributed to the abolished light response in nlr mutants, we obtained p[HS::G␣q1] transgenic flies and performed rescue experiments. Convincingly, G␣q1 expression largely restored the abolished ERG responses in nlr mutants (Fig. 3, D and E). Taken together, these data demonstrate that the 5501 T/A G␣q gene mutation largely contributed to the abolished light response in nlr mutants and excludes the possibility that the abolished light responses in nlr mutants were due to the dominant suppression of G␣q V303D mutant protein.

G␣q3 is responsible for residual light responses in G␣q1 isoform null mutants
Given that mutations in the G␣q gene are responsible for the abolished light responses in nlr mutants, and the G␣q gene encodes several G␣q splice variants, we suspected that other G␣q splice variants might contribute to the residual light responses in G␣q1 null mutant flies. To test this hypothesis, we generated G␣q 1 /G␣q 221c and G␣q 961 /G␣q 221c flies, and found that although these flies responded to light stimulus, ERG amplitudes were smaller than that of either G␣q 1 (Fig. 4A (Fig. 3A). These data indicate that the residual light response observed in G␣q1 null mutants was contributed to by G␣q3.
To examine G␣q3 protein levels in nlr mutants, we conducted Western blot analysis using anti-G␣q-C antibodies that can recognize G␣q3 specifically. Protein levels of G␣q3 were significantly reduced in nlr mutants ( Fig. 4B; 100 Ϯ 21.8 versus 31.9 Ϯ 18.7%, p ϭ 0.015, t test), suggesting additional mutations in nlr mutants affect the expression of G␣q3 isoforms. Next, we conducted DNA sequencing in the whole G␣q gene region. We failed to identify additional mutations in G␣q3 isoform-coding regions of nlr mutant flies. However, in the promotor region of the G␣q gene in nlr mutants, we identified an 11-bp sequence insertion (GTTTTTCTAAC) at the Ϫ534 to Ϫ524 position that was absent in w 1118 flies, as well as a mutation (2199 C/A ) in an 11-bp sequence (2193-2203, CTAATTCGATT) conserved in the promoter region of several photoreceptor cell-specific genes (22,23) (Fig. 4C). Moreover, qRT-PCR analysis validated that the mRNAs of G␣q3 isoforms, as well as G␣q1, were significant reduced in nlr mutants ( Fig. 4D; G␣q1: 100 Ϯ 1.0 versus 33.7 Ϯ 2.4%, p Ͻ 0.0001, t test; G␣q3: 100 Ϯ 3.3 versus 35.8 Ϯ 7.0%, p Ͻ 0.0001, t test), demonstrating that these mutations in nlr mutants affect G␣q gene transcription.

G␣q splice variants regulate visual function
To further confirm that G␣q3 mediates phototransduction, we generated p[UAS::G␣q3] transgenic flies and performed rescue experiments. Consistently, G␣q3 expression in nlr mutants generated clear ERG responses (0.13 Ϯ 0.2 versus 5.3 Ϯ 1.8 mV, p Ͻ 0.0001, t test), and the amplitude of ERG responses was comparable with that in either G␣q 1 or G␣q 961 mutant flies (Fig. 4, E and F). Taken together, these data demonstrate that G␣q3 also mediates phototransduction, and further indicates that the residual light response observed in G␣q1 null mutants is contributed to by G␣q3.

Depletion of all G␣q splice variants results in severe retinal degeneration
In Drosophila, mutations in genes that prevent normal Ca 2ϩ influx after light stimulation, such as those in trp and norpA, usually cause stable Rh1/Arr2 complex formation and light-dependent retinal degeneration (13)(14)(15). In contrast, neither 6-day-old G␣q 1 nor G␣q 961 mutants show obvious retinal degeneration, and G␣q 961 mutants show only slight accumulation of stable Rh1/Arr2 complexes after light exposure (12,19). The residual light responses in either G␣q 1 or G␣q 961 mutants might be able to trigger sufficient Ca 2ϩ influx to activate CaM kinase II, which subsequently phosphorylates Arr2 to release Arr2 from Rh1 (17,18). If true, stable Rh1/Arr2 complexes should accumulate with severe retinal degeneration after the depletion of all G␣q splice variants. Indeed, EM images revealed obvious retinal degeneration in 7-day-old nlr and Df(2R)G␣q1.3-covered gene null mutants raised under regular 12-h light/12-h dark cycles, but not in G␣q 1 mutants raised in the same conditions (Fig. 5A).
Next, we examined whether retinal degeneration in nlr mutants was due to the accumulation of stable Rh1/Arr2 complexes. With 480-nm blue light exposure, Rh1 is photoconverted to metarhodopsin and induces binding with Arr2. Metarhodopsin can be photoconverted to inactivated rhodopsin by an additional 580-nm orange light exposure, which leads to the release of Arr2 (15). In WT flies, blue light exposure caused about 74% binding between Arr2 and Rh1, and ϳ49% release of Arr2 from Rh1 following exposure to orange light (Fig. 5, B and C). In contrast, blue light exposure triggered approximately 88% binding between Arr2 and Rh1, and less than 32% release of Arr2 from Rh1 following exposure to orange light in nlr mutants (Fig. 5, B and C). These data imply that depleting all G␣q splice variants results in stable Rh1/Arr2 complexes accumulation.
To provide further support that stable Rh1/Arr2 complex formation triggers retinal degeneration in nlr mutants, we conducted an EM study in 7-day-old dark-reared nlr mutants. Interestingly, 7-day-old dark-reared nlr mutants displayed intact rhabdomere structures (Fig. 5D). Furthermore, genetic removal of arr2 in the nlr mutant background suppressed retinal degeneration in nlr mutant flies (Fig. 5E). These data demonstrate that depleting all G␣q splice variants stabilizes Rh1/

G␣q splice variants regulate visual function
Arr2 complex formation, triggering severe light-dependent retinal degeneration.
To explore the role of G␣q3 in regulating Rh1 synthesis, we monitored the distribution of Rh1 in developing photoreceptors. During pupal development, nascent Rh1 in WT pupae was gradually loaded into rhabdomeres, and most Rh1 was successfully loaded into the rhabdomeres by fly eclosion (Figs. 6D). In contrast, a large fraction of Rh1 remained in the cytoplasm in nlr mutant pupae (Fig. 6, D and E). These results indicate that proper loading of Rh1 into rhabdomeres requires G␣q3.

G␣q3 isoforms mediate residual light responses in G␣q1 null mutants
In Drosophila photoreceptors, G proteins are essential to activate the phototransduction cascade (20,24). The G␣q gene encodes several G␣q splice variants, and G␣q1 has been shown to function as the predominant G protein in fly phototransduction (12,20). In this study, we identified a mutation (5501 T/A ) in the G␣q gene, which specifically mutates Val to Asp at residue 303 in G␣q1 but not G␣q3 isoforms. Although Val is replaced with Ile at residue 303 in vertebrate G␣q proteins, the hydrophobicity at this position is evolutionally conserved. Structural analyses have shown that the Val-303 region localizes to the interface between G␣ proteins and its downstream effector PLC (25)(26)(27). The change of a hydrophobic residue to a polar one may affect the interaction between these two proteins.

G␣q splice variants regulate visual function
A recent study has shown that the G␣q V303D mutant protein is unable to activate PLC in vivo (28).
Although the 5501 T/A G␣q gene mutation largely contributes to abolished light responses, this mutation is not fully responsible for the abolished light responses in nlr mutants because both nlr/G␣q 1 and nlr/G␣q 961 flies still exhibited a residual light response similar to G␣q 1 and G␣q 961 mutants (12). These data also excluded the possibility that the G␣q V303D mutant protein dominantly suppresses the function of G␣q protein. G␣q1 expression in nlr mutants largely recovers the light response, and further excluded the possibility that abolished light responses in nlr mutants are due to the dominant suppression of G␣q V303D mutant protein.
The G␣q gene encodes several G␣q splice variants, and G␣q 221c mutants disrupt the expression of all G␣q splice variants (21). Our ERG recording revealed that G␣q 221c null mutant clones showed no light responses. Previous whole-cell voltageclamp recordings showed that the photoresponse of G␣q 1 homozygous cells is larger than that of G␣q 1 heterozygous cells (20). These results indicate that other G␣q splice variants might contribute to the residual light response in G␣q1 null mutants. In this study, we demonstrate that G␣q3 contributes to the residual light response in G␣q1 null mutants.
The G␣q gene encodes several G␣q splice variants. Originally, two cDNAs resulting from different G␣q gene splicing were isolated (23). These two cDNAs encode G␣q1 and G␣q2 isoform proteins, respectively. Functional studies demonstrated that G␣q1 mediates the light response, whereas G␣q2 has no effect on phototransduction (20). Subsequently, two additional G␣q splice variants were isolated (9). Now, seven total G␣q splice variants have been annotated in flybase, and these splice variants encode three different isoform proteins, including G␣q1, G␣q3, and G␣q4. In this study, we demonstrated that G␣q3 also mediate phototransduction. Overexpression of G␣q3 in nlr mutants induced detectable light responses but failed to fully restore the light response. Interestingly, the rescue flies exhibited comparable ERG trace amplitude and dynamics with that of G␣q 1 and G␣q 961 flies. These results indicate that different G␣q isoform proteins play different roles in phototransduction.

G␣q mediates retinal degeneration
Mutations in most genes encoding components of the phototransduction cascade result in rapid retinal degeneration, except for G␣q hypomorphic allele G␣q 1 and G␣q1 isoform null mutant allele G␣q 961 (12,13,19,29). Previous studies have shown that both G␣q 1 and G␣q 961 mutants underwent slow light-dependent retinal degeneration due to slow accumulation of stable Rh1/Arr2 complexes (12,29). In these G␣q mutants, the small residual photoresponse may reduce Ca 2ϩ influx, which partially activates CaM kinase II and leads to the slow release of Arr2 from Rh1. In this study, we showed that nlr mutants underwent rapid light-dependent retinal degeneration similar to that observed in norpA mutants (15,30). We showed that disruption of stable Rh1/Arr2 complex formation prevented retinal degeneration in the mutants. Under normal con-

G␣q splice variants regulate visual function
ditions, the interaction between Arr2 and Rh1 is transient, because light-triggered Ca 2ϩ influx may activate CaM kinase II, which subsequently phosphorylates Arr2 to release Arr2 from Rh1 (17,18). In nlr mutants, photoresponses were completely abolished so that the normal rise in Ca 2ϩ after light stimulation was blocked, causing stable Rh1/Arr2 complex formation and retinal degeneration. These observations and explanations are consistent with mutations such as trp and norpA.

G␣q3 isoforms regulate Rh1 synthesis
In this study, we showed the first evidence that G␣q3 regulates Rh1 synthesis. Rh1 is transported to the plasma membrane by vesicular transport mechanisms regulated by a large number of trafficking proteins (31)(32)(33)(34)(35). Previous studies have shown that G␣q homologue CG30054 regulates inositol 1,4,5,-trisphosphate receptor (IP3R) to mediate calcium mobilization from intracellular stores and promote calcium-regulated secretory vesicle exocytosis (36). Given that G␣q3 shows high sequence identity to CG30054, they may regulate Rh1 synthesis through promoting calcium-regulated secretory vesicle exocytosis.  (38). The genotype of WT flies is w 1118 and the mutant alleles used for each gene in this work are G␣q 961 , G␣q 1 , and arr2 5 . To avoid light-dependent retinal degeneration, all flies were reared at 22°C in dark and examined at 1-2 days old.

Generation of transgenic flies
To generate p[UAS::G␣q3] transgenic flies, G␣q3 cDNA was subcloned into the pUAST-attB vector and injected into y 1 ,w67c23;P{CaryattP2} flies. The transgene was subsequently crossed into the nlr mutant background and proteins were specifically expressed in the eye using binary expression systems (39).

Antibodies
The Nrv3 antibody generated by GenScript (Nanjing, China) was raised in rabbits against GST-Nrv3

Inverse PCR
Inverse PCR was performed as previously described (40). Briefly, genomic DNA was first digested with HpaII, and the resulting fragments were circulated using DNA ligase. The circular products were used as templates in PCR amplifications with primers flanking the 5Ј end or 3Ј end of the known p-element sequence. The PCR products were sequenced and aligned to the fly reference genome to locate p-element insertion sites.

Electrophysiological recordings
ERG recordings were performed as previously described (41). Briefly, recording and reference electrodes filled with Ringer's solution were placed on the surfaces of the fly eye and shoulder. Fly eyes were stimulated with 5-s light pulses (4000 Lux). For each genotype and condition, more than eight flies were examined. To quantitate the light response, the amplitude of ERG response was measured and the standard deviation was calculated.
Intracellular recordings were performed as previously described (42). Briefly, a low-resistance glass microelectrode filled with 2 M KCl was placed into a small hole on the compound eye. A reference electrode was filled with Ringer's solution and placed at the retina layer. The microelectrode placed on the eye was gradually inserted into the opening until light-induced membrane depolarization was observed. The signals were amplified and recorded using a Warner IE210 Intracellular Electrometer.

EM
Electron microscopy (EM) was performed as described previously (43). Fly heads were immerged in 2.5% glutaraldehyde, 0.1 M sodium cacodylate (pH 7.2) at 4°C for 12 h. After rinsing with 0.1 M sodium cacodylate three times, fixed fly heads were stained with 1% osmium tetroxide for 1 h at room temperature.

G␣q splice variants regulate visual function
A series of standard ethanol dehydrations were conducted, and fly heads were immersed in two 10-min washes of propylene oxide. Fly heads were then embedded with standard procedures. For EM, 100-nm thin sections were cut and collected on copper support grids. After staining with uranyl acetate, sections were stained with lead citrate. Micrographs were taken at 80 KV on Hitachi-7650.

Arr2-binding assays
Arr2-binding assays were carried out as described previously (16). For each group and condition, 10 adult flies were collected and adapted in complete darkness for 2 h. After exposure with 60 s of pure blue light (480 Ϯ 10 nm), fly heads were isolated by liquid nitrogen and homogenized in the dark. After centrifugation at 14,600 ϫ g for 5 min, the pellet and supernatant were separated for Western blot analysis. Arr2-release assays were performed in the same manner, except that flies were exposed to pure blue light for 60 s, followed by pure orange light exposure for 2 min (580 Ϯ 10 nm). All operations were conducted under very dim red light.

Experimental design and statistical analysis
Genetic studies and electrophysiological recordings on fly eyes were conducted to explore the function of G␣q splice variants and G␣q homology in phototransduction and Rh1 endocytosis. Biochemical studies and genetic studies on fly eyes, etc. were conducted to explore the detailed mechanism.
Western blotting was analyzed by ImageJ software (National Institute of Health, USA), and data from three independent experiments were averaged. Statistical analysis was performed using Prism 6.0 (GraphPad software). All data are presented as mean Ϯ S.D. For quantification of ERG amplitudes, more than eight flies were recorded for each genotype and ERG amplitudes were averaged to obtain a mean. All data are presented as mean Ϯ S.D. Two-tailed Student's t tests were used to compare between genotypes. Statistical significance was set as p Ͻ 0.05 (*), p Ͻ 0.01 (**), p Ͻ 0.001 (***), and no significance (n.s.).

Data availability
All data are contained within the manuscript.