It is easy to see how a system of differently shaped and/or differently positioned black shutters placed over a window in a dark room could create an almost endless variety of patterns of light areas in the unshuttered parts of the window. Apparently an analogous scheme accounts for some of the diversity of white or yellow bands and bars observed in Heliconius wings as illustrated by hindwing patterns in broods derived from genetic crosses of cydno galanthus and pachinus (see Figure 4 upper right). Furthermore, in crosses of Costa Rican H. cydno and melpomene the gene for dorsal HW shutter of cydno is epistatic to HW yellow window of melpomene (see Figure 4 lower right) with the exception that a scattering of yellow scales sometimes occurs in the window region of heterozygous males, a phenotype seen in presumptive natural zones of introgression between these species as illustrated in Brower (1996a Figure 5A).

Dorsal melanic scales of most H. melpomene races are flat black while those of cydno are typically a reflective blue/black. On the Pacific slopes of Colombia and Ecuador melpomene HW windows are totally shuttered dorsally (but express ventrally), and they display the cydno traits of reflective dorsal scales and white HW borders, traits critical for mimetic tracking of co-occurring H. erato. Meanwhile, cydno races in this region show a number of reciprocal influences of introgression from melpomene, manifested on the dorsal HW by variation in width of the white marginal band ( Figure 7 top: i, j, k). This is a reflection of underlying segregation of melpomene genes and cydno genes for small vs. large window size respectively, and can be obtained in the laboratory crosses of the Costa Rican races of these species ( Figure 7 top: h). Even though the HW window may not be visible if covered by a shutter, its size limits the area of the wing available for other "program windows" such as remnants of the nymphaline ground plan (Nijhout et al. 1990) around the margin. Thus when melpomene's small HW window is placed in a cydno background marginal white lines or bands are unveiled. Ecuadorian H. cydno also show occasional segregation of melpomene HW yellow windows, most often seen in heterozygous expression (Kapan and Gilbert in prep, Gilbert and Kapan in prep).

Dorsal FW: Uncovering the concept of HW windows and shutters led me to reevaluate my interpretation of genetic variation in area of the FW white patch of H. cydno. All the patterns seen in the F2 of cydno x pachinus can be attributed to approximately two shutter gene loci with different alleles at each characterizing the revealed parts of the underlying FW windows of cydno and pachinus as discussed in Nijhout et al (1990). Why do such close relatives look so different? The answer is that several trivial changes in scale color and shutter position create a major shift in pattern phenotype: (1) pachinus FW type I scales are recessive yellow, not white as in cydno (Gilbert et al 1988) and (2) the pachinus shutter sits centrally in the wing's FW window, allowing yellow (Type I) scales to be expressed both distal and proximal to the shutter as shutter in the middle of a real window allows light to enter above and below, while that of cydno is like a shutter pulled up against the wall above the window. A sample of one brood ( Figure 4 upper right) provides a dorsal view of segregation of most of the major FW and HW shutter phenotypes.

In a darkened room it is not possible to perceive the boundary of wall and window where a shutter is attached. Analogously, in cydno galanthus, it is not possible to perceive this boundary directly because both shutter and wall consist of melanic (type II ) scales. Presence of this shutter was initially deduced from the pattern of shutter positions in cydno x pachinus hybrids. For example in the F1 of cydno x pachinus the hybrid shutter shifts proximally (roughly averaging the position of respective parental shutters), revealing the distal edge of the white window ( Figure 6e). This phenotype, produced by introgression of shutter alleles from pachinus into cydno galanthus, is not uncommonly seen in the Atlantic forests east of Costa Rica's Meseta Central (across which gene flow occurs) but is not found in cydno populations away from that zone (personal observations).

Serendipitously, the hybridization of Costa Rican melpomene with both pure cydno, pure pachinus, and their various hybrid offspring verified the homology of the cydno group FW shutter system. Central American melpomene is fixed for a realizator gene that replaces type II (melanic) scales with type III (xanthommatin) scales exclusively in the FW shutter region. Crossing a male heterozygous for this red gene (an F1 cydno x melpomene) with a cydno x pachinus F2 hybrid heterozygous for different settings of the FW black shutters, showed a 1:1 segregation of fully expressed red shutters in the FW ( Figure 4 lower center) indicating epistasis in single dose or complete dominance. Of greater interest is that the melpomene red gene is in essence a FW shutter "marker," highlighting the shutter's boundaries and hence the window/screen boundary in cases, like H. cydno galanthus, in which the melanic FW shutter precisely abuts the window's distal boundary with surrounding melanic pattern elements. Based on its apparent early determination (see Figure 3b), I hypothesize that this boundary represents a true compartment boundary in its strict sense (see Figure 3 legend).

The expression of the melpomene FW red gene, which I isolated in the cydno genome through a process of reiterative backcrossing fertile male hybrids displaying red FW to pure race females of cydno or pachinus, confirms for cydno (compare Figure 1h vs. 1i) and for pachinus (compare Figure 1d vs. 1j) (1) that the red gene of Central American melpomene acts to precisely "mark" the FW shutter system (i.e., that clone in which shutter realizators are switched on) in these species and (2) that the FW shutter system of melpomene is homologous to that of the cydno group. Because in pure melpomene rosina no white or yellow scales show either distally or proximally to the (red) shutter, it, by definition, must completely obscure melpomene's FW window. This predicts that in crosses involving melpomene (e.g., Amazonian races) lacking red on the distal FW (or cydno for that matter) and those of Central America displaying solid red bands, all black distal FW phenotypes should segregate in F2 broods (i.e., individuals simultaneously homozygous for red null alleles and pure melpomene rosina window and shutter genotypes). Such phenotypes occur both in interracial crosses within melpomene (see Sheppard et al. 1985 ) or, more interestingly, in interspecific crosses of Costa Rican populations of melpomene and cydno/pachinus (see Figure 4 upper left d). I have also obtained individuals virtually identical in both HW and FW to Colombian H. c. weymeri form gustavi (see Figure 7 lower left b) from synthetic hybrid zones restricted to melpomene rosina and cydno galanthus (not shown). These experiments indicate that to achieve the all-black state in the distal half of the FW as seen in the Colombian H. c. weymeri f. gustavi one needs only to infuse both the melpomene window and shutter system into an otherwise pure cydno genome where the gene for red does not exist.

An important complication for interpreting hybrid phenotypes must be mentioned at this point: wing shape and size, along with relative shape, size, and position of yellow/white windows vary between species and races. If the window/screen boundary does represent a compartment boundary as hypothesized, then melpomene x cydno broods have an interesting developmental conflict in heterozygotes across the zone of the distal FW window between their respective small vs. large window "settings" of that boundary. This conflict is apparently manifested by the feathering of this otherwise sharp boundary, as seen for example in cydno x melpomene F1 (big x small window) and many of the offspring of the backcross to melpomene (Figure 4 lower right). I believe that the genetics of variation in developmental boundaries has not been considered by Drosophila workers because without patterns of diverse scale types as one has in Heliconius, it is much more difficult to detect boundaries on wings. Moreover, at levels more central to organismal function than peripheral, physiologically neutral wing pattern, identity of genetically determined boundaries on the action of homeotic genes might be required for hybrid viability in the first place. So even if compartments are real, their interspecific variation is not likely to be an interesting issue for understanding Drosophila

To compare properly the genes which add shutters to the yellow or white windows of the MCS hybrid zones.clade it is necessary to view their expression on the same window. This can be done for example by isolating melpomene shutter genes in pachinus background and highlighting them with melpomene red to assure homology. The proximal boundary of the FW distal window appears to be perpendicular to the wingıs long axis at the top of the discal cell in all the MCS clade members (see the sharp line at this location in cydno x melpomene F1 individuals and in broods that they parent ( Figure 4 lower right, 4 lower center, 4 upper left). However, the position of window's distal boundary varies. Figure 4 lower left shows pure melpomene (top) and pure pachinus (bottom). The two middle specimens resulted from selecting for (1) melpomene shutters and red color on pachinus window ( Figure 4b) and (2) pachinus shutters converted to melpomene red on pachinus window ( Figure 4c). Along a transect from the top of the discal cell to the top of the cydno window, melpomene, and pachinus distal FW shutter boundaries occur at approximately 1/2 and 3/4 the distance to cydno's position on the boundary (compare Figure 4 lower left b, c, and Figure 1i). As FW shutter position is set by approximately two loci (Nijhout et al 1990), it would appear that cydno is fixed for "distal" shutter alleles (dd, dd) at each locus while melpomene is fixed for proximal alleles (pp, pp) at the same loci. Under this hypothesis the pachinus shutter genotype is predicted to be (pp, dd). Thus when we can visualize MCS FW shutters displayed on the same window it appears the pachinus FW shutter phenotype represents a cydno x melpomene hybrid for that trait! This hypothesis is supported by H. cydno gadquae from the Tachira, San Juan de Colon, area of Venezuela. This race shows a pachinus-like position of the FW shutter in combination with strong evidence of melpomene influence on the ventral HW forceps motif (see below). A survey of Venezuelan Heliconius clearly places the race gadquae in an active zone of cydno-melpomene introgression (Brown and Fernandez 1984, figures 165-177). Moreover, I have created virtually the same phenotype (not figured) from synthetic hybrid zones of Costa Rican races of these species (unpublished data).

From these and other initial analyses of dorsal FW pattern genetics, carried out independently with various races of H. cydno and H. melpomene by M. Linares and myself, we must conclude that many of the more distinctive forms, races, or species of the H. cydno complex arise from introgression with H. melpomene (and vice versa). In addition to the distinctive shutter position which creates two FW bands in pachinus, two other traits of the pachinus FW may reflect the influence of melpomene. These are (1) the recessive yellow scales of the window (although not expressed because of epistatic realizator/shutter genes, the type I scale genotype for melpomene FW window is in fact yellow and this phenotype segregates 1:1 in the [cydno x melpomene F1] x melpomene backcross shown in Figure 4 lower right, and (2) the melanic (type II) scales of pachinus are more dull black like melpomene and Müllerian partner H. hewitsoni than reflective like conspecific cydno galanthus.

In addition to the previously mentioned Colombian cydno weymeri form gustavi which tracks Müllerian partner H. erato chestertonii (Linares 1997a), I attribute another important example in cydno variation to introgression from melpomene. The polymorphic variation seen in H. cydno FW Pacific slope forests of Ecuador (Kapan 1998) exactly matches the types of variation which can be produced in synthetic hybrid zones of melpomene and cydno from Costa Rica (Gilbert and Kapan in prep). This intra-population variation, which includes yellow vs. white FW windows and various settings of window sizes and shutter positions, allows for the polymorphic mimicry of H. sapho (white) and H. eleuchia (yellow) by the same cydno population ( Figure 7 top i, j). This population has also evolved recessive FW yellow scales which, because they would not express as rare mutants in heterozygous form, are more likely to have been derived from introgression than by mutation. Moreover, the absence of linkages of collaborating mimetic loci and their appropriate alleles into Papilla dardanus-like super genes (Kapan 1998) is consistent with a recent introgressive origin of this variation. Other polymorphic cydno races in South America, such as H. c. hermogenes, show similar evidence for multiple influences by melpomene. Ventral FW: The interesting case of red FW melpomene was already mentioned above in explaining my computer window metaphor. The oft-described "pale" red of the ventral FW surface of some melpomene races results from viewing dorsal red scales through the translucent wing membrane and (type I) white (pigmentless) scales in the ventral homologue of the FW distal window. To my knowledge, this D/V switch in scale type is an exclusively melpomene trait. Note that even cydno races with white FW windows displayed both dorsally and ventrally possess an epistatic gene for expression of type III scales (red/brown) on the ventral FW such that all crosses of the cydno group with red banded melpomene races convert the ventral FW window from white to brown or red scales (e.g., see Figure 2c). This case and the ventral HW discussed below provide ample evidence of independent regulation of realizator genes between dorsal and ventral surfaces. However, whether the switch is the type of selector gene envisioned in Garcia-Bellido's compartment model is not shown.

The semi-independence of the ventral and dorsal FW pattern means that the precise overlay of homologous pattern elements is itself under (canalizing) selection and one frequent trait of interracial and interspecific hybrids (or even strongly inbred lines) is the disruption of this D/V coordination such that boundaries do not line up (e.g., Figure 5e lower left) or a patch on one surface will appear smaller than on the opposing surface (e.g., Brower 1996, figures 8A vs.8B). This phenomenon can occur on either FW or HW but I notice it most frequently on the FW. It should be useful in detection of natural cases of interspecific introgression.

Ventral HW: This final wing surface has been the most complex to interpret in H. cydno and was ignored in previous attempts to reconcile Heliconius genetics with the nymphaline ground plan (Nijhout et al 1990) but has been studied in Colombian cydno by Linares (1989, 1996, 1997a). In Costa Rica the ventral HW of pachinus and cydno share large windows, which although polygenically different in size and shape, are sufficiently similar to allow the realizator genes for shutter traits of the two races to be visualized in essentially the same window framework. Fortunately for unraveling the details, several other major ventral HW differences between these taxa appear to consist of presence vs. null alleles or +/- states of unlinked shutter loci, some of which are also expressed on the DHW (e.g., cydno shutter) and some of which simultaneously act on the FW with respect to homologous wing-vein landmarks (e.g., the pachinus shutter).

a. the pachinus shutter: The pachinus HW is basically that of melpomene rosina except that (1) the yellow window is much larger and (2) the top half of that window is replaced by a shutter just beyond (proximal to) the discal cell. Simultaneously, this gene (or one tightly linked to it) shutters the proximal half of the FW from the end of the FW discal cell. (Compare the pachinus x cydno F1, Figure 6 to a mating pair of parents, Figure 5c and d).

b. the galanthus shutter: The shutter which totally obscures the DHW window characterizes cydno galanthus and is also present ventrally although the interaction of two other realizators in the same area obscure its presence. Fortunately, because these other loci are unlinked and are represented only by null alleles (at least with respect to major effects) in pachinus, it is possible to recover homozygotes and heterozygotes for the galanthus shutter and view its effects on the ventral HW. A shutter-free individual showing a pachinus window and the "brown line" of cydno (Figure 5a) is shown mating with a heterozygote for galanthus ( Figure 5b). Note that the reflectance shift of melanic scales heterozygous for this gene allow one to detect the window boundary.

c. red/brown realizator for shutter type (II vs. III): Just as in the case of the FW shutter system in melpomene , the "shutter color gene" for type III scales (red/brown) converts the cydno galanthus shutter region of the HW window from melanic to brown scales. This phenotype is easily obtained in laboratory crosses and Figure 5e shows a natural version of the same phenotype collected in central Costa Rica early this century and described by William Schaus (1913).

d. "forceps" shutter of cydno: This realizator gene converts most of the brown (galanthus) shutter back to black scales but leaves an arching line of these brown scales visible parallel to the upper ventral HW margin in pure cydno galanthus. A heterozygote phenotype for this pattern gene is seen in Figure 5g and in homozygous form wild type cydno galanthus ventral HW ( Figure 5c). I call this the "forceps shutter" because in combination with the non-homologous brown line which arches parallel to the lower ventral HW margin, it creates the forceps-like brown pattern on this wing surface. It is the homozygous "null" at this locus, due to gene flow from pachinus, that produces the phenotype shown in Figure 5e.

Note that one influence of introgression from yellow HW window races of melpomene into cydno is to replace the large window with much smaller versions depending on the degree of melpomene influence. As the HW window's size decreases, the brown line moves to keep its relationship to the edge of the window, and the sides of the forceps motif straighten and close to near touching. Such cydno phenotypes (e.g., Figure 7 top e and f) are common in areas of Venezuela, Colombia and Ecuador where other evidences of introgression from yellow HW window melpomene are abundant. As the brown line pulls away from the lower margin to follow the smaller melpomene window, broad streaks of white scales move in behind creating, for example, the marginal white bands seen on the Pacific side of Ecuador (see discussion above). Remarkably, this phenomenon can be replicated in synthetic hybrid zones involving these species using only the pattern tool box options available in Costa Rica (e.g., Figure 7, top panel, f, h). .

While it is melpomene's influence that widens a HW marginal window (as seen in Figure 7, top panel, e, f, and h), it is cydno galanthus that provides the previously mentioned white pattern elements in an area that is solid melanic in melpomene rosina and, incidentally, in pachinus. In the latter population, submarginal white dots on the ventral HW are best viewed as evidence of gene flow from H. c. galanthus. In northwestern South America, introgression from cydno supplies the dorsal HW shutter, reflective blue sheen on the melanic scales, and the white borders that allow H. melpomene vulcanus and H. m. cythera to track their respective Möllerian partner races of H. erato , venus and cyrbia (both of which display their yellow bands on the ventral, but not dorsal HW).

It is appropriate to mention here that under the MCS tool box hypothesis, Amazonian races of melpomene showing yellow FW, narrow orange HW window, and "rays" on both surfaces of the HW (see Figure 7 top) would be more closely related to cydno galanthus than to melpomene rosina with respect to the essential pattern elements employed. Indeed, crosses of those melpomene races with Costa Rican cydno show that a principal element of melpomene that is not immediately recognizable as homologous, the system of HW rays, may be a "streaked out" version of the brown line according to the mechanism just detailed above ( Figure 7 top: compare a, d ,m). And of course, in melpomene, the brown-shuttered HW window and the accompanying system of rays (= cydno brown line) appear dorsally as well as ventrally. In synthetic hybrid zones of Amazonian melpomene crossed Costa Rican cydno, melpomene contributes the null (-) allele for the realizator that keeps the dorsal HW surface black in cydno galanthus. This allele, when homozygous in a cydno background, permits display of cydno ventral HW brown forceps phenotypes dorsally as well (personal obs., not shown).

e. basal red: H. pachinus is alone among races of cydno in lacking ventral HW brown and in possessing red basally like its Müllerian partner, H. hewitsoni. The Müllerian partner of H. cydno galanthus, H. sapho, also sports basal red, yet cydno has not followed suite. I believe this is a consequence of the ease of mimicking the white forewing of sapho on the one hand and the difficulty of tracking the red pattern given fact that neither melpomene x cydno F1 nor backcrosses to cydno produce homozygous recessive melpomene basal red pattern. While it is possible to synthesize better sapho mimicry in the lab via reiterative backcrosses and production of "pseudo F2" broods, the crosses required to produce recessive basal red would be highly unlikely in nature without isolation from backcrosses to standard cydno and/or much stronger mimetic selection on the ventral HW. In contrast, the hypothesized cydno x melpomene hybrid forerunners of H. pachinus faced jumping to an adaptive peak of black, yellow, and red (H. hewitsoni colors). It is worth noting that the latter two scale colors are recessive to their respective white and brown counterparts. Therefore, in any small, and isolated hybrid populations which would have allowed the occurrence and fixation of FW yellow homozygotes by genetic drift, the co-occurrence of the red basal pattern on the ventral HW could be expected. Moreover, mimicking the HW yellow band of H. hewitsoni required removal of all those nasty galanthus shutters (as just reviewed) and with them the genes epistatic or dominant to the basal red.

f. shutter interaction and novelty: Without going into much further detail it is worth repeating that peeling away these layers of interacting realizator genes, particularly on the ventral HW, would not have been possible without the serendipity of their independent assortment between pachinus and cydno and the communal sharing of the same tool box for pattern generation across the MCS clade. By the same token, it is independence of these unlinked realizator loci and the cell autonomous nature of their effects across bold patches of wing surface that provide the creative power of the MCS tool box. Virtually all departures from the common pattern themes within the cydno clade (e.g., H. heurippa, H. timareta, H. pachinus, H. cydno weymeri f gustavi) can be explained by introgression of toolbox elements from melpomene. Reciprocally, several melpomene races show clear influence of cydno in its adaptive tracking of H. erato. In addition to vulcanus and cythera, already mentioned, I regard the white FW patches and black HW shutters of m. plessini and the all black HW of m. flagrans possible examples of cydno influence on melpomene pattern evolution.