Correction: Subunits of the Drosophila Actin-capping Protein Heterodimer Regulate Each Other at Multiple Levels

due to a typesetting error. The publisher apologizes for this error. Please download the PDF again to view the corrected article. The originally published, uncorrected article and the republished, corrected article are provided here for reference. Copyright: ß 2014 The PLOS ONE Staff. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.


Introduction
The actin cytoskeleton controls numerous processes, including cell shape, mobility, division and intracellular transport. In normal cells, the actin cytoskeleton is tightly controlled to regulate these essential functions; however, it can be subverted by cancer cells and contributes to changes in cell growth, proliferation, stiffness, movement and invasiveness [1,2]. Moreover, alterations in the activity or expression of actin-binding proteins (ABPs) per se, have been linked to cancer initiation and progression [2,3,4,5,6].
Among these actin regulators, the actin Capping Protein (CP) heterodimer, composed of an a and a b subunit, appears to act as a main tumor suppressor module [7,8,9,10]. CP was named based on its ability to bind and cap actin filament barbed ends, inhibiting the addition and loss of actin monomers [11,12,13]. CP has homologs in nearly all eukaryotic cells, including vertebrates, invertebrates, plants, fungi, insects and protozoa [14]. Drosophila and organisms other than vertebrates have single genes encoding capping protein a (cpa) or b (cpb). In contrast, vertebrates contain two genes expressed somatically that encode two a subunits (a1 and a2), and one single gene that produce two b isoforms (b1 and b2) through alternative splicing [15,16,17]. Although the amino acid sequences of the a and b subunits are not more similar to each other than they are to other ABPs, nor they share common sequences with other proteins, they have extremely similar secondary and tertiary structures [18]. When in complex, the heterodimer resembles a mushroom with the C-terminus of each subunit forming tentacles located on the top surface of the heterodimer [19,20]. In vitro analyses of chicken and budding yeast CP revealed that deletions or point mutations in either the a or b tentacles do not affect protein stability but reduce the capping affinity, while a complete removal of both tentacles fully abrogates the actin-binding activity [12,20]. Thus, CP appears to cap F-actin barbed ends via the independent interaction of both tentacles with actin. In vivo, a truncated form of Drosophila cpa deleted of the Cterminal 28 amino acids has no effect on F-actin when expressed alone but promotes F-actin accumulation when co-expressed with full length cpb [21]. Similarly, a chicken b subunit containing a point mutation changing a conserved leucine to arginine at position 262, which caps actin poorly, disrupts the early steps in myofibrillogenesis of cultured myotubes and the sarcomere of mouse heart [22,23,24].
In yeast and Drosophila, removing either cpa or cpb induces Factin accumulation and identical phenotypes [25,26,27]. In the fly, CP is required for proper differentiation of adult bristles, survival of the adult retina, determination of the oocyte and cortical integrity of nurse cells in the egg chamber [27,28,29,30]. In addition, CP has a key role in restricting tissue growth. In the whole wing disc epithelium, CP-dependent F-actin regulation suppresses inappropriate tissue growth by inhibiting the activity of the Yorkie (Yki) oncogene, which mediates Hippo signalling activity [7,9]. This function is conserved, as the a1 subunit is also required to limit the activity of the Yki orthologs YAP and TAZ in mammary epithelial cells [31]. In addition, in the distal Drosophila wing disc epithelium, CP prevents JNK-mediated apoptosis or proliferation and counteracts the oncogenic ability of Src [8,21,32]. Furthermore, underexpression of the human a1 subunit correlates with cancer-related death and causes a significant increase in gastric cancer cell migration and invasion in vitro, whereas its overexpression has the opposite effect [10].
We aimed in this report to understand how both subunits regulate each other in vivo to control F-actin levels and tissue growth. We show that Cpa and Cpb stabilize each other's protein levels and can stimulate the production of each other's mRNA when the level of one of the subunit is reduced. Because overexpressing CP decreases F-actin levels and tissue growth, while expressing forms of CP mutated in their actin-binding domains has opposite effects, we propose that by regulating each other, Cpa and Cpb assure that a pool of functional CP heterodimer is produced in sufficient quantities to restrict tissue growth and therein prevent tumor development but not in excess to sustain proper tissue growth.

Molecular Biology
To generate UAS-cpb L262R , site-directed mutagenesis was performed on the plasmid UAS-cpb, using the QuikChange kit (Stratagene, # 200519). The mutated plasmid was confirmed by sequencing and transgenic flies were generated by standard methods.

Antibody Generation
The rabbit anti-Cpa and rabbit anti-Cpb polyclonal antibodies were generated by Metabion International AG using full length Cpa or Cpb tagged with Histidine.

Immunohistochemistry and quantification
We performed immunocytochemistry using the procedure described in Lee and Treisman [37]. Primary antibodies used were mouse anti-Arm (N2 7A1, Developmental Studies Hybridoma Bank (DSHB); 1:10), rat anti-DE-Cad (1:50, CAD2, DSHB), rabbit anti-Cpa (1:200); rabbit anti-Cpb (1:200); mouse anti-HA (Covance 11 MMS101P; 1:1000) and rabbit anti-Caspase 3 (Cell Signalling #9661; 1:50). Rhodamine conjugated phalloidin (Sigma) was used at a concentration of 0.3 mM. Secondary antibodies were from Jackson Immunoresearch, used at 1:200. Wing discs were mounted in VECTASHIELD Mounting Media (Vector Laboratories, Inc. #H-1000). Fluorescence images were obtained on a Leica SP5 confocal microscope or on a LSM 510 Zeiss confocal microscope. The NIH Image J program was used to perform measurements. Quantifications of the intensity of Caspase 3 signals were performed as described in [21]. Quantifications of the ratio of Phalloidin signal between posterior and anterior wing compartments were performed as described in [7]. To quantify the ratio of Cpa or Cpb signals between the anterior and posterior wing disc compartments, a region of interest (ROI) of 100 per 50 pixels was selected. The sum of the gray values was measured for each ROI, applied to each compartments for each disc on optical cross sections through distal wing disc epithelium comprising the apical surface. To measure wing size, wing were dissected one to two days after eclosion and imaged using the Hamamatsu Orca-ER camera attached to a Zeiss' Stereo Lumar V12 stereoscope. The total area of each wing was outlined and measured using the area measurement function. Statistical significance was calculated using a two-tailed t-test.

Western Blotting
For each genetic background, proteins were extracted from either four wing imaginal discs or four dechorionated embryos using a 2x SDS sample buffer (Sigma #S3401). Samples were frozen in liquid nitrogen, boiled for 5 minutes in 5 ml Sample Buffer 2x, spun at 13,000 g for 1 minute, loaded on a 10% SDS-PAGE gel and transferred to a PVDF membrane (Amersham Hybond-P, GE Healthcare). Proteins were visualized by immunoblotting using rabbit anti-Cpa (1:2500) or rabbit anti-Cpb (1:2500) or mouse anti-HA (Covance 11 MMS101P; 1:1000) or rabbit anti-Histone H3 (Cell Signalling #9715; 1:3000). HRPconjugated donkey anti-mouse or donkey anti-rabbit secondary antibodies were used at 1:5000 (Jackson ImmunoResearch Laboratories, Inc.). Blots were developed using Amersha ECL Plus Western Blotting Detection System (GE Healthcare). Densitometric analysis of signal intensity was performed using the GelQuant.NET software (biochemlabsolutions.com) and normalized with the loading control. Statistical significance was calculated using a Paired t-test.

Isolation of RNA and Real-Time qRT-PCR
Total RNAs were extracted from either 10 first instar larvae or 50 wing imaginal discs for each genetic background. Samples were homogenized in RLT buffer treated with DNase (Qiagen) at 4 degree C and total RNAs were isolated using the RNeasy mini kit (Qiagen) following manufacturer instructions. First Strand cDNA Synthesis Kit for RT-PCR (Roche) was used to produce cDNAs from 1 mg of total RNA. To quantify mRNA levels, qPCRs were carried out on reverse-transcribed total mRNA using intron-exonspecific primers (Table S1), designed using the Primer3 software [38,39], and ensuring that efficiency is at least 90% and restricting primer dimmer formation. Real-time qPCR was performed using PerfeCTa SYBR Green FastMix (Quanta Biosciences) in 384 well skirted PCR microplates (Axygen) sealed with optically clear sealing tape (STARSTEDT) in the Applied Biosystems 7900HT Fast Real-Time PCR System. The relative amount of mRNA for each condition was calculated after normalization to the RpL32 transcript. Statistical significance was calculated using a Paired t-test with significance at P,0.05.

Cpa and Cpb stabilize each other's protein levels and accumulate at Adherens Junctions
To understand how Cpa and Cpb are regulated to restrict growth of Drosophila epithelia, we generated polyclonal antibodies to each CP subunit. In lysates from embryos expressing UAS-mCD8-GFP under the control of the ubiquitous daughterless-Gal4 (da-Gal4) driver, the Cpa (Fig. 1A) and Cpb (Fig. 1B) antibodies revealed a band at around 32 and 31 kDa respectively by Western Blot. These signals were lost in embryonic extracts from homozygous cpa (Fig. 1A) or cpb mutants (Fig. 1B) respectively. Conversely, overexpressing full length cpa, tagged with HA (UAS-HA-cpa + ; Fig. 1A) or cpb (UAS-cpb + ; Fig. 1B) with da-Gal4, enhanced the anti-Cpa or anti-Cpb signals respectively. Similarly, Cpa levels were increased in wing disc lysates overexpressing HAcpa under scalloped-Gal4 control (sd.HA-cpa + ; Fig. 1C), while endogenous Cpb levels were similar to control sd.GFP lysates (Fig. 1D). Forcing cpb expression in this tissue also induced a significant increase in Cpb levels by Western Blot (Fig. 1D) but did not significantly affect endogenous Cpa levels (Fig. 1C). Crosssections through wing disc epithelia expressing UAS-mCD8-GFP in the posterior compartment using the hedgehog-Gal4 (hh-Gal4) driver showed that Cpa (Fig. 1E-E999) and Cpb (Fig. 1F-F999) accumulated at the apical cell membrane and co-localized with components of Adherens Junctions, including the b-Catenin homolog Armadillo (Arm). Co-expressing cpb and mCD8-GFP in this domain strongly enhanced the anti-Cpb signals but did not affect Cpa levels ( Fig. 2D-D0). Conversely, hh.HA-cpa + wing disc epithelia displayed an apical localization of HA-Cpa, like endogenous Cpa (Fig. 1E-E0), but no change in Cpb levels ( Fig. 2E-E0). Thus, the anti-Cpa and Cpb antibodies recognize specifically Cpa and Cpb respectively.
Strikingly, Cpa levels were strongly reduced not only in wing disc extracts expressing double-stranded RNAs (dsRNA) for cpa under sd-Gal4 control (sd.cpa-IR) but also in discs knocked-down for cpb (sd.cpb-IR; Fig. 1C). In the converse experiment, the amount of Cpb was also strongly reduced in both sd.cpb-IR and sd.cpa-IR wing disc extracts (Fig. 1D). Similarly, knocking down cpa (Fig. 1G-G9 and H-H9) or cpb (Fig. 1I-I9 and J-J9) in the posterior wing disc compartment with hh-Gal4 significantly reduced the apical accumulation of both Cpa and Cpb when compared to anterior compartments used as internal controls. Moreover, both Cpa and Cpb levels were also strongly reduced in lysates from first instar larvae homozygote mutant for cpa or cpb (Fig. S1A) and in clones mutant for cpa or cpb (Fig. S1B-B0 to E-E0). To verify that the cpa dsRNA did not affect cpb mRNA and vice versa, we performed quantitative RT-PCR (qRT-PCR) experiments on wing imaginal discs knocked down for cpa or cpb. As expected, sd.cpa-IR or sd.cpb-IR wing discs showed a significant reduction of cpa (Fig. 1K, 2.560.43 folds) or cpb mRNA (Fig. 1L, 2.660.41 folds) levels respectively, relative to control sd.GFP. However, cpa mRNA levels were not significantly affected by a reduction in cpb (Fig. 1K), nor were cpb mRNA levels reduced in wing discs knocked-down for cpa (Fig. 1L). Similarly, a reduction in cpa or cpb levels had no effect on cpb or cpa mRNA levels, respectively, in first instar larvae expressing cpa-IR or cpb-IR under da-Gal4 control ( Fig. S1F and G). Taken together, we conclude that Cpa and Cpb accumulate at apical cell membrane and enhance each other's protein levels.

Cpa and Cpb levels are rate limited to form a functional heterodimer
The Capping Protein a and b subunits form a functional heterodimer, which caps F-actin barbed ends via the interaction of the a and b tentacles with actin ( Fig. 1A and [11,12,13,20]). To confirm that the stabilization of Cpa and Cpb's protein levels by each other promotes the formation of a functional heterodimer, we first tested if co-expressing cpb and HA-cpa would enhance the levels of both subunits by comparing the levels of HA-Cpa and Cpb when overexpressed alone or together, ensuring that each genetic combination contained the same number of UAS transgenes. Indeed, by Western Blot (Fig. 2B, P,0.0092) and in wing disc epithelia ( Fig. 2 compare F-F0 with E-E0), HA levels were strongly enhanced when HA-cpa was co-expressed with cpb. Similarly, the co-expression of HA-cpa and cpb strongly increased Cpb levels compared to wing disc lysates overexpressing cpb alone (Fig. 2C). Overexpressed HA-cpa and cpb appeared to form a functional heterodimer as their co-expression in the posterior wing disc compartment with hh-Gal4 decreased the apical F-actin ratio between both compartments compared to hh.GFP control ( Fig. 3F, P,0.0001). In contrast, overexpressing either HA-cpa or cpb alone has no effect on F-actin levels [21]. We conclude that the levels of endogenous Cpa and Cpb available are rate limited to form a functional heterodimer.
Forms of CP mutated in a or b tentacle counteract the ability of wild type CP to restrict F-actin accumulation Surprisingly, expressing an HA-tagged form of Cpa deleted of the a tentacle (UAS-HA-cpa DABD ) has no significant effect on Factin when expressed alone [21] but triggered apical F-actin accumulation when co-expressed with cpb ( Fig. 3F, P,0.0001 and [21]), indicating that HA-Cpa DABD affects F-actin only in the presence of overexpressed cpb. We therefore tested if the coexpression of cpb would also enhance the levels of HA-Cpa DABD . In contrast to full length HA-Cpa, which accumulated apically ( Fig. 2E-E0), HA-Cpa DABD localized uniformly along the apicalbasal axis in the posterior compartment of hh.HA-cpa DABD wing discs ( Fig. 2G-G0). Strikingly, co-expressing cpb not only enhanced strongly HA-Cpa DABD levels as assessed by Western Blot (Fig. 2B, P,0.0002), but also relocalized HA-Cpa DABD at the apical cell membrane ( Fig. 2H-H0). Thus, forcing Cpb levels enhances the levels of HA-Cpa DABD and promotes its apical localization.
The heterodimer formed between HA-Cpa DABD and Cpb appears to have reduced capping activity and may be recruited to F-actin barbed ends, preventing the binding of wild type CP. If so, we would expect that a form of Cpb truncated of its b tentacle would also promote F-actin accumulation in the presence of endogenous CP. To test this possibility, we expressed a form of cpb mutated in the highly conserved Leucine 262 (UAS-cpb L262R ), which has been proposed to directly interact with actin [12]. While overexpressing full length cpb had no significant effect on F-actin ( Fig. 3 compare B-B0 with A-A9 and F), hh.cpb L262R wing discs accumulated apical F-actin in the posterior compartment ( Fig. 3C-C0 and F, P,0.0001). However, co-expressing full length HA-cpa in these tissues suppressed the apical F-actin accumulation due to the presence Cpb L262R (Fig. 3D-D0 and F, P,0.0001). Thus, forcing Cpa levels tethers the effects of Cpb L262R on F-actin. In contrast, F-actin accumulation was strongly enhanced when cpb L262R was co-expressed with HA-cpa DABD (Fig. 3E-E0 and F, P,0.0001). Moreover, Cpb L262R , like full length Cpb, enhances HA-Cpa DABD levels and triggered its relocalization to the apical cell membrane (Fig. 2I-I0). We conclude that forms of CP with reduced capping activity inhibit wild type CP to restrict F-actin accumulation, most likely by tethering barbed ends, preventing the recruitment of wild type CP.

CP and forms of CP with dominant negative effects on Factin have opposite effects on tissue growth
Decreasing or increasing CP levels has opposite effects on Factin levels ( Fig. 3F and [25]). Because loss of CP induces overgrowth of the wing disc epithelium by promoting Yki activity [7,9], we asked of overexpressing cpa and cpb has an opposite effect on tissue growth. Indeed, overexpressing full length HA-cpa and cpb in the wing primordium using the nubbin-Gal4 (nub-Gal4) driver significantly reduced the size of the adult wing (Fig. 4A, compare nub.GFP control wing in green to nub.cpa + , cpb + wing in magenta and F; P,0.0151), but does not affect cell survival [21]. Thus, tight CP levels are critical to control tissue growth.
To determine if CP controls tissue growth via F-actin regulation, we analyzed the effect of expressing forms of cpa and cpb that have dominant negative effects on F-actin on wing growth. Expressing HA-cpa DABD and cpb ( Fig. 4B and F, P,0.0001) or cpb L262R alone ( Fig. 4C and F, P,0.0001) or combined with HAcpa DABD (Fig. 4E and F, P,0.0001) under nub-Gal4 control, not  only promoted apical F-actin accumulation (Fig. 3), but also enhanced significantly the growth of adult wings. Strikingly, expressing HA-cpa suppressed the overgrowth of nub.cpb L262R wings ( Fig. 4D and F, P,0.0001), indicating that the effect of Cpb L262R on F-actin and tissue growth is dependent on the levels of full length Cpa. Because altering the levels or activity of CP did not affect the density of wing hairs (Fig. 4A9, B9, C9 D9 and E9), which develop from one single cell, the CP-dependent growth defects most likely result from changes in proliferation rate rather than alteration of cell size. We conclude that a CP-dependent reduction of F-actin levels correlates with tissue undergrowth, while a CP-dependent increase in F-actin levels is associated with tissue overgrowth.
The a tentacle is not absolutely required to form a functional heterodimer Because the heterodimer formed between HA-Cpa DABD and Cpb appears to be recruited at F-actin barbed ends, we tested if HA-Cpa DABD can partially compensate for the loss of endogenous Cpa. Expressing cpa-IR under sd-Gal4 control induced the activation of Caspase 3 in numerous cells in the distal wing disc epithelium (Fig. 5A-A9). Apoptosis was almost fully suppressed by overexpressing full length HA-cpa (Fig. 5B-B9 and G; P,0.0001).
Expressing HA-cpa DABD also significantly prevented apoptosis of sd.cpa-IR wing discs, although to a much weaker extent than HAcpa (Fig. 5C-C9 and G; P,0.0005). These effects were not only due to titration of the cpa dsRNAs by the overexpressed cpa constructs as HA-cpa (Fig.5E-E0 and H) or HA-cpa DABD (Fig. 5F-F0 and H; P,0.0048) also rescued apoptosis of clones mutant for a cpa allele. Expressing HA-cpa or HA-cpa DABD in sd.cpa-IR wing discs also partially restored Cpa (Fig. 5I) and Cpb (Fig. 5J) levels, as assessed by Western blot. Quantification of the ratio of Cpb signals between the posterior and anterior compartments of wing discs expressing cpa-IR under hh-Gal4 control showed that knockingdown cpa reduced Cpb levels in the posterior compartment compared to hh.GFP control (Fig. 5K). This decrease in Cpb levels was significantly alleviated by the presence of HA-Cpa DABD (Fig. 5K P,0.0085). We conclude that in the absence of wild type Cpa, Cpa DABD s capable of forming a functional heterodimer with Cpb, which prevents apoptosis.

Cpb compensates for a reduction in cpa by enhancing cpa mRNA levels and vice versa
Interestingly, co-expressing cpb with HA-cpa DABD almost fully suppressed apoptosis of wing discs knocked-down for cpa (Fig. 6 compare B-B9 with A-A9 and D; P,0.0001). This effect could be due to the stabilization and apical relocalization of HA-Cpa DABD when co-expressed with cpb ( Fig. 2H-H0). However, apoptosis of sd.cpa-IR wing discs was also significantly suppressed by overexpressing cpb alone (Fig. 6C-C9 and D; P,0.0001). Conversely, expressing HA-cpa in tissues knocked-down for cpb (sd.cpb-IR) also prevented apoptosis (Fig. 7 compare B-B9 with A-A9 and C; P,0.0001).
To understand the mechanisms by which Cpa and Cpb compensate for each other's function, we tested the effect of overexpressing cpb on Cpa levels in cpa-depleted tissues. As previously observed, by Western Blots, Cpa (Fig. 6F) and Cpb (Fig. 6G) levels were strongly reduced in wing disc extracts knocked-down for cpa. Forcing cpb levels in these tissues enhanced the levels of both Cpa (Fig. 6F and Fig. S2) and Cpb (Fig. 6G and  Fig. S2). We quantified this effect by measuring the ratio of Cpa signals between the posterior and anterior compartments of hh. cpa-IR-expressing wing discs, in the presence or absence of overexpressed cpb. While in control hh.GFP tissues this ratio was 0.95, knocking down cpa reduced this ratio to 1,34 folds ( Fig. 6H; P,0.0001). This effect was significantly alleviated by the overexpression of cpb ( Fig. 6H; P,0.01). In contrast, overexpressing cpb in control hh.GFP wing discs did not affect Cpa levels (Fig. 6H), indicating that Cpb enhances Cpa levels only when cells contain reduced Cpa levels. By Western Blots, HA-cpa also enhanced both Cpa (Fig. 7D) and Cpb (Fig. 7E) levels when expressed in tissues knocked-down for cpb. Thus, Cpa compensates for a reduction in cpb by stimulating the production of Cpb, and vice versa.
Using qRT-PCR, we next analyzed if overexpressing either subunits affects the mRNA levels of the other. After normalization to the RpL32 transcript used as an internal control, we observed that whereas cpa (Fig. 6I, P,0.0027) but not cpb (Fig. 6K) mRNA levels were strongly reduced in wing discs knocked-down for cpa (sd.cpa-IR), forcing cpb levels in these tissues fully restored cpa mRNA to wild type levels ( Fig. 6I; P,0.0003). In contrast, in wing discs that contained endogenous cpa and cpb, overexpressing cpb, which strongly enhanced cpb mRNA levels (Fig. 6L), had no significant effect on cpa mRNA levels (Fig. 6J). Thus, Cpb stimulates the production or stabilization of cpa mRNA only when Cpa levels are reduced. In the converse experiment, overexpressing HA-cpa in sd.cpb-depleted wing discs enhanced the levels of both cpa (Fig. 7F) and cpb ( Fig. 7H; P,0.0018) mRNA. However, in wing discs that contained endogenous cpa and cpb, only cpa mRNA levels were strongly increased ( Fig. 7G and I). The ability of Cpb to suppress apoptosis of cpa-depleted wing discs was due to the increase in cpa mRNA and protein levels as clones mutant for a cpa allele showed similar apoptotic levels in the absence or presence of overexpressing cpb (Fig. 6E). We conclude that Cpa compensates for a reduction in cpb by increasing cpb mRNA levels and vice versa.

Cpa and Cpb regulate each other at multiple levels
Our data argue that in Drosophila, different pools of Cpa and/or Cpb co-exist, and they regulate each other at various levels. One level of regulation involves their reciprocal stabilization of their protein levels. First, in Drosophila, like in yeast, the loss of one CP subunit reduces the protein levels of the other subunit ( [26] and Fig. 1) but does not affect its mRNA levels ( Fig. 1 and Fig. S1). Second, co-expressing cpa and cpb in Drosophila tissues enhances synergistically the levels of both subunits relative to the levels of each subunit overexpressed alone (Fig. 2). Third, large quantities of soluble active chicken CP can be produced in bacteria only when both subunits are co-expressed [40]. Cpa and Cpb may stabilize each other's protein levels via direct protein-protein interactions [19]. The tight interaction between both subunits may prevent the recruitment of E3 ubiquitin ligases that would otherwise target individual CP subunits for degradation by the 26S proteasome. As an heterodimer, CP has been shown to bind to the fast polymerizing ends of actin filaments, preventing further addition of actin monomers [41,42] and to restrict F-actin accumulation in Drosophila tissues [25,27]. In addition, Cpa and Cpb appear to show some function on their own as overexpressing cpb rescues apoptosis of wing discs knocked-down for cpa and vice versa ( Fig. 6 and 7). Overexpression of cpb alone is also sufficient to   Figure 5E and F and blots were processed in parallel (see Figure S2 showing the whole experiment). enhance the retinal defects of flies knocked down for the Cblinteracting protein cindr [43] and to rescue the migration and Factin polarization defects of Drosophila border cells mutant for warts [44]. Because individual chicken CP subunits expressed in bacteria are mainly deposited into insoluble cytoplasmic inclusion bodies but can be renaturated as active heterodimers [45], individual subunit may exist in the cell as pools of insoluble monomers. The molecular mechanism by which individual CP subunit compensates for each other's function remains to be determined. Several observations argue that this mechanism involves the production of the subunit knocked-down by the other subunit via an increase of its mRNA levels ( Fig. 6 and 7). CP has been observed in the nuclei of chicken retinal and kidney epithelial cells in culture, in Madin-Darby canine kidney (MDCK) cells, in Xenopus laevis oocytes and bovine lens epithelial cells in culture [46,47]. Whether Cpa and Cpb influence each other's transcription in the nucleus is an interesting possibility to be tested. The protein-mRNA feedbacks between Cpa and Cpb may guarantee that a pool of functional heterodimer is present to limit F-actin polymerization. However, a CP-dependent negative feedback mechanism must exist that restricts the production of CP in excess, as forcing the expression of one of the subunit in tissues that contain endogenous CP does not enhance the mRNA and protein levels of the other subunit ( Fig. 6 and 7). Because the loss of one subunit has no effect on the mRNA levels of the other subunit ( Fig. 1 and Fig. S1), the CPdependent negative feedback may act by limiting the ability of individual subunits to stimulate the production of each other's mRNAs. Thus, in addition to regulate each other's protein levels, individual CP subunit stimulates each other's mRNA production up to an optimal physiological threshold of functional heterodimers. Further experiments are necessary to elucidate the protein-

Capping activity of the CP heterodimer at actin filament barbed ends
Our observations argue that in vivo the actin-binding domain of Cpa is not absolutely required to form a functional CP heterodimer, as HA-Cpa DABD partially compensates for the loss of endogenous Cpa (Fig. 5). Consistent with our observations, in actin assembly assays, a mutant form of the chicken a subunit that lacks the a tentacle is able to cap F-actin [12]. Nevertheless, the a tentacle may favor the interaction and therefore stabilization of the a subunit by the b. This possibility is consistent with the observation that HA-Cpa DABD is found in the cell at much lower levels than full length HA-Cpa (Fig. 2) despite both transgenes being inserted at the same locus in the fly genome and therefore likely expressed at similar levels [21]. Consistent with this hypothesis, Arginine 259 of the chicken a1 tentacle forms sidechain hydrogen bonds with three residues of the b subunit, all residues being conserved across isoforms and species [19]. Moreover, in vitro, a truncated form of the chicken a1 subunit, consisting only of the C-terminal domain, retains the ability to form a heterodimer [48]. The reduced ability of HA-Cpa DABD to interact with Cpb may explain its inability to fully suppress apoptosis of Cpa-depleted tissues (Fig. 5) and to affect F-actin levels when overexpressed alone [21]. However, several observations indicate that the a and b tentacles also enable full capping activity in vivo. First, in actin assembly assays, the C-terminus of the chicken a1 and b1 subunits are required for high-affinity capping [12]. Second, in the presence of endogenous CP, stabilizing HA-Cpa DABD levels by forcing cpb expression does not reduce F-actin levels, as does overexpressed HA-cpa/cpb, but instead, promotes Factin accumulation (Fig. 3 and [21]). Third, replacing leucine 262 of the chicken b subunit has no effect on protein stability and global structure but decreases the capping affinity significantly [12,20]. Fourth, identical mutations in the b orthologs induces Factin accumulation in Drosophila tissues (Fig. 3) and disrupts the sarcomere of mouse heart [24]. Thus, we propose that the heterodimers formed between HA-Cpa DABD and Cpb or between Cpb L262R and Cpa are recruited to F-actin barbed ends and cap actin filaments less efficiently than wild type CP. The low capping activity of the HA-Cpa DABD /Cpb heterodimer is sufficient to partially compensate for the loss of Cpa. However, in the presence of endogenous CP, the HA-Cpa DABD /Cpb heterodimers compete with wild type Cpa/Cpb heterodimers for binding the barbed ends of F-actin, which can lead to defects in F-actin.
Tight regulation of CP levels is critical to control tissue growth CP appears to act as a gatekeeper, which limits the development of cancer-related processes. Loss of the a subunit promotes Yki/ YAP/TAZ-dependent proliferation in Drosophila epithelia and in human cells [9,31], causes a significantly increase in gastric cancer cell migration and is associated with cancer-related death [10]. In contrast, increasing CP levels has opposite effects: it reduces tissue growth (Fig. 4) and prevents Src-mediated tumour development in Drosophila [21], and significantly restricts gastric cancer cell migration [10]. Several of our observations argue that the function of CP on tissue growth involves its F-actin capping activity. First expressing cpb L262R , which contains a single point mutation affecting the capping activity [23], induces F-actin accumulation (Fig. 3) and wing overgrowth (Fig. 4). Moreover, CP-dependent Factin accumulation correlates with tissue overgrowth, whereas tissue undergrowth is associated with a CP-dependent reduction in F-actin ( Fig. 3 and 4). Consistent with these observations, other actin regulators have been shown to control Yki/YAP/TAZ dependent tissue growth [7,9,31]. Thus, a reduction or an increase of CP levels has deleterious consequences on tissue growth, implying that it must be tightly regulated. This may be achieved in part by the ability of Cpa and Cpb to stimulate or limit the production of each other in conditions of lower or higher CP levels respectively, assuring that a pool of functional CP heterodimer is produced in sufficient quantities in the cell to prevent cancer development but not in excess to sustain proper tissue growth. Table S1 Intron-exon-specific primers used to quantify cpa, cpb and RpL32 mRNA levels by qRT-PCR. (DOCX)