An unexpected bifurcation in the Pointed transcriptional effector network contributes specificity and robustness to retinal cell fate acquisition

Spatiotemporally specific and robust cell fate transitions are fundamental to the development of appropriately patterned tissues. In the Drosophila retina, receptor tyrosine kinase / mitogen activated protein kinase (MAPK) signaling acts through the transcriptional effector Pointed (Pnt) to direct two distinct rounds of photoreceptor specification. A relay mechanism between two Pnt isoforms, a MAPK responsive form PntP2 and a constitutively active form PntP1, initiates and sustains the transcriptional response. Here, we report an unexpected bifurcation in the Pnt effector network. We show that PntP2 works redundantly with a closely related but previously uncharacterized isoform, PntP3, to activate pntP1 during specification of first round photoreceptors. Intrinsic activity differences between PntP2 and PntP3, combined with positive and negative transcriptional auto- and cross-regulation, buffer first-round fates against conditions of low signaling. In contrast, in a mechanism that may be adaptive to the stronger signaling environment used to specify second round fates, PntP2 uniquely activates pntP1. We propose that differences in expression patterns, transcriptional activities and regulatory interactions between Pnt isoforms together facilitate context-appropriate cell fate specification in different signaling environments.


Introduction 28 29
During development, cells integrate external signals and internal information to coordinate the 30 transition from a multipotent to a differentiated state. A relatively small number of transcription 31 factors acting downstream of an even smaller handful of signal transduction pathways coordinate 32 the gene expression changes that drive cell fate acquisition (Flores et al., 2000;Halfon et al., 33 2000;Voas and Rebay, 2004). Two conditions must be fulfilled to effect these transitions: 34 specificity, whereby cells adopt the correct fate in a precise spatiotemporal manner (Cagan, 2009;35 Guillemot, 2007;Wolpert, 1969); and robustness, whereby cells reliably execute the appropriate 36 program despite genetic and nongenetic variations (Félix and Barkoulas, 2012;Liu et al., 2019). 37 Despite recent progress, the regulatory strategies used by transcription factors to elicit specific 38 and reproducible developmental transitions remain poorly understood. 39 40 Photoreceptor specification during Drosophila retinal development offers an ideal model to study 41 the regulatory mechanisms that confer specificity and robustness to cell fate acquisition. Each of 42 the ~750 ommatidia that comprise the compound eye contains a core cluster of eight 43 photoreceptors, R1-R8. These neurons are specified in a stereotyped spatiotemporal sequence 44 that is initiated repeatedly as the morphogenetic furrow (MF) travels anteriorly across the 45 epithelial field (Wolff and Ready, 1991). Photoreceptor specification occurs in two rounds that 46 are separated temporally by a single synchronized cell division known as the second mitotic 47 wave (SMW) (Ready and Hanson, 1976;Tomlinson and Ready, 1987). During the first round, 48 R8, the founder cell of each ommatidium, emerges from the MF's wake, followed by the R2/R5 49 and R3/R4 pairs. During the second round, specification of photoreceptors R1/R6 and finally R7 50 completes the cluster. Recruitment of non-neuronal support cells to the ommatidia follows 51 immediately, starting with the four lens-secreting cone cells. 52 53 Specification of all photoreceptors except R8 requires inductive signaling by the receptor 54 tyrosine kinase (RTK) / Ras / mitogen-activated protein kinases (MAPK) pathway via the 55 transcriptional effector Pointed (Pnt), the Drosophila homologue of the mammalian ETS family 56 activators ETS1 and ETS2 (Freeman, 1996;Scholz et al., 1993). Multipotent retinal progenitors 57 must therefore translate this common signal into specific photoreceptor fates. Numerous studies 58 expression patterns raised the possibility of both distinct and overlapping functional 238 requirements for PntP2 and PntP3 in the two rounds of photoreceptor specification (Fig 2A). 239

Redundant and non-redundant requirements for PntP2 and PntP3 during two distinct 240 rounds of photoreceptor specification 241
Prior studies of pnt function during retinal development concluded that pntP1 and pntP2 are 242 required non-redundantly to specify photoreceptors R1-R7 (O'Neill et al., 1994), providing the 243 foundation for the sequential activation model (Shwartz et al., 2013;and Fig 1C). However the 244 pntP2 allele used in the studies, pnt ∆78 (O'Neill et al. 1994), was generated by imprecise excision 245 of a P-element inserted into the first SAM-encoding exon, and so also disrupts pntP3 (Fig. 1A). 246 This means that pnt ∆78 phenotypes, in the eye loss of R1-R7, reflect the combined loss of pntP2 247 and pntP3. 248 To reveal the individual requirements for the two isoforms we generated pnt p2 and pnt p3 specific 249 mutants using CRISPR/Cas9 genome editing to insert stop codon(s) right after the start codon of 250 each isoform (Fig. 1A). To confirm the effectiveness of the molecular strategy, we also 251 engineered a pnt p2p3 double mutant allele. As reported for pnt ∆78 (Morimoto et al., 1996;O'Neill 252 et al., 1994), homozygous pnt p2p3 adults were never recovered, indicating that the combined 253 function of the two isoforms is essential for viability. In contrast, homozygous pnt p3 animals 254 were fully viable while homozygous pnt p2 animals occasionally survived. From a cross between 255 balanced heterozygous pnt p2 parents, only 35 homozygous pnt p2 animals were found from a total 256 of 2058 progeny scored at the 3 rd larval instar stage; this 1.7% survival rate to 3 rd instar is 257 significantly lower than the 33% expected for full viability. The differences in survival of the 258 isoform specific mutants suggested both redundant and non-redundant requirements for PntP2 259 and PntP3 during development, with PntP2 playing the major role and PntP3 a more auxiliary 260

one. 261
Focusing on photoreceptor specification, homozygous pnt p2p3 clones were missing all 262 photoreceptors except R8 (Fig. 3A-A''); this phenotype is consistent with published analysis of 263 pnt ∆78 mutant clones (O'Neill et al., 1994;Shwartz et al., 2013). We reasoned that if the function 264 of both PntP2 and PntP3 is required for photoreceptor specification, then neither single mutant 265 should recapitulate the double mutant phenotype. If so, we predicted the requirement for PntP3 266 should manifest in the first round fates where it is strongly expressed but not in second round 267 fates where its levels are low ( Fig. 2A, B and S3B). Alternatively, if PntP3 does not contribute 268 activity essential to photoreceptor specification, then the pnt p2 and pnt p2p3 mutants should show 269 identical loss of R1-R7 phenotypes. 270 To distinguish between these two possibilities, we first assessed photoreceptor loss in adult eyes, 271 using F-actin to highlight the number and spatial arrangement of the rhabdomeres. In a wild type 272 ommatidium, the larger rhabdomeres of R1-R6 are arrayed in a trapezoidal-shaped ring with the 273 smaller rhabdomeres of R7 in its center ( 3D, red arrow). The more modest photoreceptor loss seen in pnt p2 single mutants (Fig. 3D) 279 relative to pnt p2p3 double mutants (Fig. 3A) indicates a functional requirement for PntP3. 280 As an independent test of this conclusion, we crossed our new pnt alleles to flies carrying a Sev-281 Yan ACT transgene, a genetic background in which constitutive activity of the RTK antagonist Yan 282 blocks specification of the photoreceptor fates in which it is expressed (Rebay et al., 1995). This 283 background had been shown previously to be sensitive to pnt dose such that pnt heterozygosity 284 dominantly enhanced the Sev-Yan ACT photoreceptor loss phenotype (Rebay et al., 2000). 285 Removal of one copy of either pnt p2 or pnt p3 dominantly enhanced the Sev-Yan ACT rough eye 286 phenotype and loss of photoreceptors, while loss of either both copies of pnt p3 or one copy each 287 of pnt p2 and pnt p3 enhanced even further ( Fig. S4A-G) . Quantification of photoreceptor numbers 288 showed that homozygous loss of pnt p3 and heterozygous loss of pnt p2p3 were phenotypically 289 equivalent with respect to Sev-Yan ACT enhancement (Fig. S4G). Thus, PntP3 contributes to 290 photoreceptor specification, with redundancy between PntP2 and PntP3 buffering completely 291 against PntP3 loss and partially against PntP2 loss. 292 In contrast, and consistent with expression pattern-based predictions, loss of pnt p2 resulted in loss 293 of cell fates recruited during the second round of specification. First, only a few Pros positive 294 cells remained in the posterior of discs from homozygous pnt p2 animals; a similar posterior 295 scattering of Cut-positive cells suggested a complete failure to specify R7 photoreceptors and 296 most cone cells ( Fig. 3I and Fig. S4H, I). Second, normal expression of Sal and reduction of Svp 297 expression to only two, rather than four cells per ommatidia, indicated correct specification of 298 photoreceptors R3/R4 and a failure to specify R1/R6 (Fig. 3J-L). Third, and confirming no other 299 consistent photoreceptor specification defects, examination of Sens and Elav patterns in pnt p2 300 mosaic discs indicated normal recruitment of photoreceptors R8/R2/R5 (Fig. 3M, L). Thus the 301 complete loss of R1-R7 fates that occurs in pnt p2p3 double mutant ommatidia (Fig. 3A) reflects 302 the combined loss of redundant inputs to R2-R5 first found fates plus the PntP2-specific input to 303 R1, R6 and R7 second round fates. 304

Synergistic and unique functions of Pnt isoforms during wing patterning 305
PntP2 has been implicated in EGFR-mediated regulation of wing disc patterning (Paul et al., 306 2013). EGFR signaling initially specifies the dorsal compartment and then as the disc grows, 307 continued strong signaling specifies the notum, restricting development of the wing proper to the 308 rest of the epithelium (Campbell, 2000;Pallavi, 2003;Paul et al., 2013;Zecca and Struhl, 2002). 309 Reflecting these roles, partially duplicated wing pouches were reported in hypomorphic pntP2-310 specific allelic combinations (Scholz et al., 1993). As predicted by this prior work, homozygous 311 pnt p2 null mutant adult escapers had duplicated and malformed wings (Fig. 4B, green line marks 312 the wing tissue); examination of 3 rd instar wing discs revealed duplication of the wing pouch and 313 reduction of notum tissue ( Fig. 4A, C, yellow arrow). In contrast, pnt p3 null mutant wings and 314 wing discs were normally patterned ( Fig. 4D and not shown). 315 To explore whether redundancy with PntP2 might mask the role of PntP3 in the wing, we first 316 compared their expression patterns in the 3 rd instar wing disc. As in the eye, both overlapping 317 and non-overlapping expression domains were detected (Fig. 4E). Thus both isoforms were 318 expressed in a cluster of cells in the posterior compartment of the notum (Fig. 4E', E'', yellow 319 arrow), cells elsewhere in the notum expressed primarily pntP2 (Fig. 4E', red arrow), cell 320 clusters in the dorsal hinge region expressed primarily GFP-PntP3 ( Fig. 4E'', green arrow), and 321 neither showed strong expression in the wing pouch. 322 comparing the phenotypes of pnt p2 versus pnt p2p3 somatic mosaics difficult, we instead assessed 324 the consequences of reducing pntP3 dose in a sensitized genetic background in which animals 325 were doubly heterozygous for null alleles of egfr and rolled(rl). The wings of the double 326 heterozygotes were wild type, indicating that EGFR signaling remained adequate to support 327 normal development (Fig. 4F). Confirming the background was indeed sensitized, removal of 328 one copy of pntP2 produced defects in the distal wing margin and creases along the longitudinal 329 axis, albeit at low penetrance ( Fig 4G, K). Although heterozygosity for pntP3 was not sufficient 330 to produce a phenotype in the egfr/rl background (Fig. 4H, K), when both copies were removed, 331 80% of the adults showed distal wing margin defects and longitudinal creases (Fig. 4I, K). 332 Margin defects and creases also occurred in egfr/rl; pnt p2 /pnt p3 quadruple heterozygotes (Fig. 4J); 333 the 82% penetrance of these phenotypes relative to the low penetrance in the single 334 heterozygotes ( Fig 4K) suggested strong synergy between pntP2 and pntP3. Given that neither 335 isoform had detectable expression in the third instar wing pouch, these adult phenotypes must 336 reflect loss of expression at a different stage. Further work will be required to identify when and 337 where PntP2 and PntP3 are coexpressed in the wing pouch in order to understand better these 338

phenotypes. 339
PntP2 and PntP3 provide robustness through redundant activation of pntP1 transcription 340 341 A central tenet of the current model of pnt function during photoreceptor specification is that 342 PntP2 activates pntP1 transcription (Shwartz et al. 2013;Fig. 1C) . Given the partial genetic 343 redundancy between PntP2 and PntP3, we asked whether PntP3 also contributes to this activation. 344 To start, we used reverse transcription quantitative polymerase chain reaction (RT-qPCR) to 345 measure pntP1 transcript levels in pnt p2 and pnt p3 mutant tissues. In both mutants, modest, but 346 not significant decreases in pntP1 transcripts were measured in eye discs (p = 0.1; Fig. 5A) and 347 no changes were detected in wing discs (Fig. S5A). This suggests either redundancy between 348 PntP2 and PntP3 with respect to activating pntP1 transcription, inadequate sensitivity in the RT-349 qPCR assay, or that PntP2 and PntP3 are not the primary activators of pntP1. 350

14
We were concerned that by grinding up whole tissue we were destroying spatial information and 352 therefore missing locally significant changes in pntP1 levels. Also, the animals lacking both 353 PntP2 and PntP3 do not survive to 3 rd instar, precluding RT-qPCR analysis of the double mutant. 354 Thus we turned to fluorescence in situ hybridization (FISH) to ask whether the two isoforms 355 work redundantly to active pntP1 expression. To start we examined pntP1 transcription, using 356 probes that target its isoform-specific exons. The FISH showed an expression pattern consistent 357 with that of the pntP1 enhancer trap allele (Scholz et al., 1993;Shwartz et al., 2013). Specifically, 358 we detected peak pntP1 transcription in a periodic pattern at the MF, lower levels of expression 359 extending to the SMW region and then decreased levels in the posterior of the disc (Fig. 5B, D). PntP2 and PntP3 redundantly activate pntP1. 364

Context specific auto-and cross-regulation of pntP2 transcription 365
Having established the functional redundancy of PntP2 and PntP3 with respect to induction of 366 pntP1, we next investigated how the system tunes these two parallel inputs to achieve the desired 367 output. In particular we wondered whether cross-regulatory feedback might coordinate and 368 optimize PntP2/PntP3 expression levels, and ultimately their activity. To test this possibility, we 369 used RT-qPCR to measure changes in pntP2 and pntP3 transcript levels in eye imaginal discs 370 dissected from pnt p2 and pnt p3 homozygous mutant 3 rd instar larvae. 371 372 Two findings emerged. Most striking, and unexpectedly, the experiment uncovered negative 373 auto-regulation for both isoforms (Fig. 6A). Thus, pntP2 transcript levels were significantly 374 increased in pnt p2 mutant tissue (p < 0.01) and pntP3 transcripts were significantly increased in 375 pnt p3 mutant tissue (p < 0.05). Given the surprising nature of this result, we repeated the 376 experiment using wing imaginal discs, and again found significant increases in transcript levels 377 in the respective mutant (Fig. 6B). This suggests that both isoforms negatively regulate their own 378 transcription, either directly or indirectly. 379 380 Second, evidence of cross-regulation emerged from the eye disc experiments (Fig. 6A), with a 381 significant increase in pntP3 transcript levels measured in pnt p2 mutant tissue (p < 0.05). A 382 suggestive, but not statistically significant, increase in pntP2 transcript levels was noted in pnt p3 383 mutant tissue (p = 0.18), hinting at bidirectional inhibitory cross-regulation. Cross-regulatory 384 interactions were not detected in the wing disc (Fig. 6B). 385

386
The coexpression of PntP2 and PntP3 in the anterior half of the disc where first round 387 photoreceptor fates are specified predicted that the regulatory interactions uncovered by RT-388 qPCR were occurring in this context. We therefore turned to FISH to corroborate the negative 389 auto-regulation and to assess further the possibility of cross-regulatory interactions. We found 390 that pntP2 transcription initiated at the MF, peaked in the region of the second mitotic wave 391 (SMW), and then continued at a more moderate level across the posterior half of the disc (Fig.  392 7A, C). This pattern was consistent with that reported by the enhancer trap pnt 1277 although the 393 prolonged perdurance of beta-galactosidase likely over-reports pntP2 levels in the posterior half 394 of the disc (Fig. 2). Unfortunately our FISH protocol was not able to detect pntP3, presumably 395 because its specific exon is too short for adequate numbers of probes (see Methods). Because PntP3 expression is strongest anteriorly (Fig. 2), we wondered whether the increase in 408 pntP2 transcripts detected in pnt p2 mutant tissue might depend on cross-regulatory activation by 409 PntP3. To test this we examined pntP2 transcript levels in pnt p2p3 double mutant clones (Fig. 7F). 410 No increase was detected at the MF or in the adjacent region of peak expression. In more 411 posterior pnt p2p3 mutant clones, pntP2 transcript levels were lower than in adjacent wild type 412 clones, exactly as seen in pnt p2 single mutant clones (Fig. 7E). Thus in anterior regions where 413 PntP3 expression is strong, loss of PntP2 results in a PntP3-dependent increase in pntP2 414 transcription whereas in posterior regions where PntP3 expression is normally low, loss of PntP2 415 results in a PntP3-independent reduction in pntP2 transcription. As discussed below (Figure 8) Together our results suggest that essential regulatory responsibilities previously attributed solely 428 to PntP2, are actually distributed between PntP2 and PntP3, and that depending on context, the 429 two work redundantly, uniquely or synergistically. We speculate that the network of auto-and 430 cross-regulatory interactions we have uncovered between the isoforms fine-tunes Pnt 431 transcriptional output to confer specificity and robustness to the developmental transitions it 432 directs. 433

434
Our investigation of the PntP3 isoform has uncovered an unexpected bifurcation in the 435 transcriptional effector network that transduces RTK/MAPK signaling. In doing so, it has also 436 corrected an erroneous assumption regarding the role of the closely related PntP2 isoform. Prior 437 to our study, the accepted model was that MAPK phosphorylation of PntP2, followed by PntP2p-438 mediated induction of pntP1 transcription, provided the essential activating input for RTK-439 dependent transitions ( Fig. 1C; Shwartz et al., 2013). As exemplified by studies in the eye, the 440 genetic cornerstone of this model was that null alleles of either pntP2 or pntP1 produce identical 441 phenotypes, namely a failure to specify photoreceptor R1-R7 fates (O'Neill et al., 1994;Shwartz 442 et al., 2013;Yang and Baker, 2003). However the allele pnt ∆78 (O'Neill et al., 1994), previously 443 misinterpreted as a pntP2-specific null, actually disrupts the exon common to pntP2 and pntP3. 444 Thus the failure to specify R1-R7 fates reflects the compound loss of both PntP2 and PntP3. It 445 should also be noted that an earlier study using hypomorphic truly pntP2-specific alleles 446 concluded correctly that there is an "absolute requirement for pntP2 function in R1, R6 and R7" 447 but did not find a requirement in R2-R5 (Brunner et al., 1994a). is required for R1, R6, R7 fates during the second round of photoreceptor specification and so 458 eyes from isoform-specific pnt p2 null mutants lack these three photoreceptors whereas pnt p3 459 mutant ommatidia have the wild type complement of eight. Because pntP1 transcript levels 460 posterior to the SMW are already quite low in wild type discs, our FISH experiments were 461 unable to detect the presumed reduction in pntP1 in pnt p2 mutant discs. 462 463 As a general developmental strategy, the redundant use of PntP2 and PntP3 may provide an 464 effective buffer against genetic perturbations that reduce RTK signaling. Using R2-R5 465 photoreceptor specification as a specific example, the presence of redundant MAPK effectors in 466 the early stages of ommatidial assembly may maximize overall robustness by minimizing early 467 "mistakes" that would derail the entire process. Supporting this idea, we found that in a 468 genetically sensitized background with reduced MAPK signaling output in R3, R4 precursors 469 (Rebay and Rubin, 1995;Rebay et al., 2000), loss or reduction in dose of either pntP2 or pntP3, 470 which in otherwise wild type discs did not compromise patterning, now resulted in loss of these that enhance RTK pathway output, such as increased Pnt expression or activity, severely disrupt 480 ommatidial assembly and wing patterning (Brunner et al., 1994b;Karim et al., 1996;Prober and 481 Edgar, 2000). Therefore to prevent redundant use of PntP2 and PntP3 from overactivating 482 transcriptional programs, there need to be mechanisms to fine-tune and limit output. Counteracting the negative auto-regulation at pntP2 and pntP3, we also uncovered positive 496 transcriptional cross-regulation whereby PntP3 can activate pntP2. Thus in pnt p2p3 double mutant 497 clones, the increase in pntP2 transcript levels that occurs in pnt p2 single mutants was no longer 498 observed. Again, we favor the simplest model of direct activation of pntP2 by PntP3 (Fig. 8A), 499 but cannot rule out more complicated indirect regulatory relays. Whether the converse cross-500 regulation of pntP3 transcription by PntP2 occurs, and whether PntP2 and/or PntP3 positively 501 auto-regulate their transcription anterior to the SMW remains to be assessed. However the 502 decrease in pntP2 transcript levels measured posterior to the SMW in pnt p2 mutant tissue argues 503 that positive auto-regulation is possible, making it plausible that such regulation could also help 504 fine-tune PntP2/P3 levels and output during specification of first round fates. 505 506 How specific PntP2:PntP3 ratios influence the acquisition of different photoreceptor cell fates 507 will be in an interesting focus for future work. Numerous studies have shown that regulatory 508 networks can either amplify or suppress both the intrinsic noise (i.e. the randomness associated 509 with mRNA/protein expression and degradation) and extrinsic noise (i.e. the variability caused 510 by fluctuations in cellular processes or environment) of protein levels to influence cell fate 511 decisions (Chang et al., 2008;Singh and Hespanha, 2009;Voliotis and Bowsher, 2012). Very 512 speculatively, perhaps the network of auto-repressive and cross-activating interactions between 513 PntP2 and PntP3 also tunes the cell-to-cell variation in Pnt isoform or Pnt target gene expression, 514 thereby influencing the response to inductive signaling. 515

516
Another intriguing feature of the network of transcriptional interactions uncovered in our study is 517 that corresponding to the switch from redundancy between PntP2 and PntP3 to the uniqueness of 518 PntP2, the balance of PntP2 autoregulation shifts from repression during first round fate 519 specification to activation during the second round (Fig. 8A). Fig. 8B offers speculation on how 520 the distinct RTK signaling environments anterior vs. posterior to the SMW, combined with 521 intrinsic differences in PntP2 vs. PntP3 activity, could produce this shift. Briefly, R1-R7 fates all 522 rely on EGFR signaling while the R1, R6, R7 photoreceptors specified during the second-round 523 experience additional RTK signaling through Sevenless (Sev); studies focused on R7 524 specification have highlighted the requirement for both EGFR and Sev (Basler and Hafen, 1989;525 Stark et al., 1976;Tomlinson et al., 2019). Use of the same Ras/MAPK/Pnt pathway means that 526 EGFR and Sev-initiated signals can be considered interchangeable (Fortini et al., 1992;Freeman, 527 1996), with lower pathway activity required in the first round and higher activity needed in the Because both PntP2 and PntP3 are direct MAPK substrates whose transactivation potential is 531 increased by phosphorylation, their combined transcriptional output will be sensitive to the 532 abundance of activated MAPK. Under conditions of lower pathway signaling and when both 533 isoforms are co-expressed, as occurs anterior to the SMW, competition for the limited pool of 534 activated MAPK will lead to domination by the unphosphorylated, less active forms. The 535 presence of PntP3, whose unphosphorylated form has equivalent activity to PntP2p, and whose 536 phosphorylated form has twice the activity of PntP2p ( Figure 1D), may be important to make 537 sure pathway output remains above a certain threshold in situations with lower levels of 538 signaling. Although at first glance this might predict that the system would not tolerate loss of 539 PntP3, because loss of PntP3 also reduces MAPK substrate competition, this would shift the 540 distribution of PntP2 protein toward the phosphorylated more active form, thereby ensuring a 541 robust transcriptional response. between the phosphorylated and unphosphorylated forms bias the competition. We suggest such 551 biased competition will be essential to achieving limited activation, or even repression, of target 552 genes such as pntP2, while allowing strong induction of others, such as pntP1, in the same cell. 553 Based on a large-scale interactome study that reported closely related isoform pairs often have 554 distinct protein-protein interaction patterns (Yang et al., 2016), it is possible that association with 555 distinct cofactors also contributes to Pnt isoform enhancer occupancy bias. Much more 556 complicated patterns of competition and cooperativity in which different Pnt isoforms and 557 species co-occupy enhancers with each other and with heterologous transcription factors will 558 undoubtedly contribute to the target gene-specific regulation needed to induce different cell types. 559

560
The substrate competition-based model also readily explains the transcriptional shifts that may 561 occur in the individual pnt p2 and pnt p3 mutants. If one removes either PntP2 or PntP3, then 562 overall competition for activated MAPK is eased, resulting in domination by the phosphorylated 563 form of the remaining protein to boost transcriptional output. This would derepress targets like 564 pntP2, as detected in our experiments, while activation of targets like pntP1 would continue at 565 21 physiologically functional levels. This same scenario plays out in an even stronger form in the 566 wild type disc during specification of second round photoreceptor fates, where the combination 567 of only PntP2 plus twice the RTK pathway input would result in phosphorylation of an even 568 greater proportion of total PntP2 protein (Fig. 8B). Because PntP2 appears to have intrinsically 569 weaker transaction potential than PntP3, ensuring full phosphorylation in situations where it is 570 the sole MAPK effector may be critical to activating the transcriptional program. 571 Our study adds to the growing appreciation of the enormous and exquisitely sensitive regulatory 572 potential available to developing tissues through the combinatorial expression and use of 573 different protein isoforms, and also offers insights beyond the Drosophila arena. The human 574 homologs ETS1 and ETS2 show intriguing structural and functional parallels to Drosophila 575 PntP2 and PntP3 (Wasylyk et al., 1997;Watson et al., 1988). In particular, ETS1 and ETS2 have 576  (A) A schematic, not drawn to scale, of the ~55kb pnt locus, showing the 5' and 3' UTRs, exons and introns of pntP1, pntP2 and pntP3. All three isoforms splice into common 3' exons encoding the ETS DNA binding domain (yellow box). pntP2 and pntP3 also share three internal exons encoding the SAM and PLTP MAPK phosphorylation site (blue boxes). Unique N-terminal exons encode isoform-specific sequences. Approximate insertion sites of key P-element-derived alleles are shown: the white+, lacZ enhancer trap insertions pnt 1277 and pnt HS20 respectively report pntP2 and pntP1 expression (Scholz et al., 1993;Shwartz et al., 2013); the excision allele pnt ∆78 disrupts the SAM-encoding exon common to pntP2 and pntP3 (O'Neill et al., 1994).
Green boxes labeled ATG-GFP signify genomic BAC transgenes in which PntP3 was Nterminally GFP tagged. Red boxes labeled ATG-TAA represent the CRISPR-generated null alleles of pnt p2 and pnt p3 in which stop codons were introduced immediately after the ATG; both alleles also carry exonic deletions (see Methods); pnt p2p3 carries identical stop codon insertions and deletions. (D) PntP3 has stronger activity but similar MAPK responsiveness relative to PntP2 in transcription assays using a reporter with 6 tandem high-affinity ETS sites (O'Neill et al., 1994), show. For each sample, activity was normalized to reporter alone control (see Methods). Error bars are S.D. of three independent experiments. P-values were calculated using two tailed pairwise Student T-tests between the samples indicated.
(E-G) lz-GAL4-driven overexpression of UAS-pntP2 (F) and UAS-pntP3 (G) disrupts external eye morphology, pigmentation and size relative to driver alone control (E). Scale bar: 50 µm.    (E) β-gal expression from the pnt 1277 (red) is seen throughout the notum (red arrow) and overlaps GFP-PntP3 (green) in the posterior compartment (yellow arrow). PntP3 is also expressed in the dorsal hinge region (green arrow). No expression of either isoform was detected in the pouch.
Scale bar: 50 µm.  (B, C) pntP1 transcripts were detected in a periodic pattern at the MF and then at a low uniform level across the wild type eye field (B). No obvious changes were detected in pnt p2 discs (C).
(D, E) pntP1 transcript patterns were comparable between wild type (D) and pnt p3 (E).   (A-B) Maximum projection images of pntP2 FISH in representative wild type and pnt p2 3 rd instar eye imaginal discs, oriented posterior to the right. Orange arrowheads mark the MF, red arrows mark the peak of pntP2 expression and blue arrows mark the start of lower expression in the posterior half of the disc; the three can be mapped to correspondingly colored marks in Fig.   2A based on the pixel distances. In pnt p2 discs (B) relative to wild type (A), an increased and broader peak of pntP2 transcripts was detected in and immediately posterior to the MF while a decrease was seen in the posterior half of the disc. Scale bar: 5 µm.
(C-D) Quantification of pntP2 FISH in wildtype (C) and pnt p2 mutant (D) from maximum projections of 6 independent discs of each genotype. In wild type pntP2 levels begin to rise anterior to the MF (orange arrowhead), peak (red arrow) and decrease to a steady state in the posterior half (blue arrow). In pnt p2 discs pntP2 levels were higher than normal in the anterior half (left of blue arrow) but lower in the posterior (right of blue arrow). Each dot plots the product of the fluorescent intensity and the size of an individual pntP2 FISH focus, representing the relative amount of pntP2 transcript (y-axis on the left); one focus was detected per nucleus, with individual foci varying in both size and brightness. The line connects the moving average of the sum of all foci within one-pixel windows along the x-axis (y-axis on the right). Further details in materials and methods.
(E) Homozygous pnt p2 clones in a 3 rd instar eye disc, positively marked with GFP (green). Clone boundary is circled with green line (E'). pntP2 levels in the mutant clones were increased relative to levels in neighboring wild type tissue in the anterior region (orange arrowhead and red arrow) but appeared decreased in more posterior clones (blue arrow). Examination of 8 clones in 7 discs from 3 independent experiments showed consistent changes. Images are partial projections for representative clone regions, scale bar: 5 µm.
(F) pntP2 FISH in homozygous pnt p2p3 clones, positively marked with GFP (green). pntP2 levels in the mutant clones were indistinguishable from wild type in the anterior regions (orange arrowhead and red arrow) but appeared decreased in more posterior clones (blue arrow).
Examination of 9 clones in 6 discs from 2 independent experiments showed consistent changes.
Images are partial projections for representative clone regions, Scale bar: 5 µm.  MAPK substrate competition between PntP2 and PntP3 keeps the ratio of phosphorylated to unphosphorylated protein low, reducing overall transactivation potential in the system. At the molecular level, if the enhancer of pntP2 is biased towards unphosphorylated PntP2, this would limit access to the more active forms (PntP2p, PntP3 and PntP3p), effectively auto-repressing.
The pool of more active forms would also be more likely to activate pntP1 enhancer, even more so if it had a bias toward the phosphorylated forms. In contrast, regulation during the second round is simpler, with stronger MAPK activation from EGFR plus Sev signaling plus reduced substrate competition from PntP3 increasing the ratio of phosphorylated to unphosphorylated PntP2 to ensure robust activation of its own transcription and of pntP1.