Bombyx mori (Silk worm) and Manduca sexta (Carolina sphinx/tobacco hornworm) were purchased as larvae or cocoons from Educational Science (Tx). Pachydiplax longipennis (blue dasher, dragonfly) and Tibicen sp. (cicada) were caught from wild populations in Charleston, SC. Libellula pulchella (twelve-spotted skimmer dragonfly) were provided by Dr. J. Marden (Penn State University, PA). Acheta domesticus (cricket) were purchased from local pet stores.

Total RNA was purified from whole animals, isolated body parts (legs, heads), and from dissected thorax flight muscles using Trizol (Invitrogen™) as described before12.

Multiple sequence alignments (MSA) of specific projectin regions available for several insects1213 were performed using the CLUSTALW algorithm1718. Degenerated primers were manually designed from nucleotides stretches showing the highest conservation. Primers used in this study are located within the two stretches of Ig domains found in the NH₂-terminal region (Table 1).

RT-PCR reactions were performed as described before19 on the different RNA preparations using various combinations of the primers in Table 1. Annealing for both the RT and PCR reactions were tested in a range of temperatures to optimize each degenerated primer set. Resulting DNA fragments were isolated after agarose gel electrophoresis and subcloned into the pGEM-T shuttle vector (Promega, Inc.). They were then sequenced (Genewiz, Inc.) and the contigs assembled manually from the overlapping clones.

Sequence comparisons were carried out using the CLUSTALW algorithm1718 and the alignments were viewed in Jalview20. Pairwise homology scores generated by CLUSTALW were used to calculate the average degree of conservation for each individual Ig domain. Intrinsic disorder predictions were also carried out using three predictor software, IUPRED 2122, FoldIndex23 , and PONDR-FIT24.

Dissected flight muscles or half thoraces were quick frozen in Optimal Cutting Temperature (OCT) medium at – 20oC and 10 μM slices were sectioned using a Thermo Scientific Microm HM550 cryostat. Alternatively, individual muscle fibers were cut from dissected flight muscles and directly immobilized on subbed microscope glass slides using hand pressure on a coverslip. Sections or squashes were fixed immediately and the slides processed for immunofluorescence microscopy as described previously25. For double labeling, the two primary antibodies from different species were applied together, as were the two animal-specific secondary antibodies. Both Cy5 and fluorescein-tagged secondary antibodies were used for identification. F-actin was stained using phalloidin-FITC or RITC (Sigma) diluted 1:100 together with the secondary antibody. Slides were mounted in mounting medium (Vectorshield, Inc.). All slides were examined by epifluorescence microscopy on an Olympus 1X81 spinning disk confocal microscope. All immunofluorescence experiment preparations treated with secondary antibodies alone showed no background staining. Images were captured and analyzed using Olympus Slidebook software.

We recently completed the annotation of the full Manduca sexta (tobacco hornworm) projectin gene using sequences available from GenBank, with the analysis of its PEVK domain described elsewhere14. Some insect projectin sequences used in this study for comparison have been previously described121314. The phylogenetic relationship and flight muscle physiology of the insects included in this study, as well as the status of their projectin sequences are summarized in Figure 1

Using a collection of degenerated primers covering the NH₂-terminal Ig domains, initial RT-PCR amplification products were generated for two dragonfly species (P. longipennis and L. pulchella), as well as one cicada (Tibicen sp.) and one cricket (A. domesticus). Sequences were completed by designing species-specific primers to fill-in any remaining sequence gaps, as well as the original degenerated primers. These analyses provide us with the corresponding complementary DNA (cDNA) sequences for the two stretches of Ig domains flanking the PEVK region, which we referred to as the N8Ig and N6Ig regions.

The N8Ig and N6Ig sequences were split into their individual domains, and all the domains at a specific position, e.g. all Ig1, were aligned with each other using CLUSTAL-W. The pairwise homology scores generated by the multiple alignment analysis were used to generate an average homology score for each domain. As presented in Figure 2, the average degree of conservation for individual Ig domain varies, with the highest level for Ig1 and Ig2 (80%), falling to only 43% for Ig8 just before the PEVK extensible region. The six Ig domains of N6Ig and the unique FRAM linker sequence between Ig9 and Ig10 show an intermediate level of conservation from 76 to 58% (Figure 2).

The alignments for each of the Ig domains were also compared to the Ig consensus sequence originally defined for the Ig domain of C. elegans twitchin27. When only the amino acids at these consensus positions (black areas in Figure 3A) are examined, the level of conservation for residues of the consensus sequence is similar for all the Ig domains, from 17 to 21 amino acids out of 37 consensus residues. There is also a strong conservation of non-consensus amino acids at both ends of each domain except for Ig8. Most of the variability between domains can be accounted for by a decreased conservation of residues within the central portion of the domain as exemplified by the differences between Ig1 (80%) and Ig3 (65%). Based on X-ray crystallography or NMR studies of both titin and twitchin Ig domains, the central amino acids correspond to loop D-E, helix E, and loop E-F on the crystal structures of various titin and twitchin domains (Figure 3B). This central domain is variable in both sequence and length in different titin Ig domains, but tends to protrude from the central nucleus of the domain282930313233 as shown in Figure 3B.

The analysis of the alignment for the second stretch of domains (the N6Ig) reveals a similar pattern of conservation of consensus and non-consensus residues at each ends of the domain and a more variable central domain (data not shown). The N6Ig domain between Ig 9 and Ig 10 also contains a unique sequence known as FRAM, which is 46 amino acid long and shows 76% overall homology for all the insects included in this study.

The conservation of the consensus residues suggest that the projectin N-terminal Ig domains could adopt the Ig fold as described for several titin Ig domains2829. However several Ig domains of the titin protein have also been extensively studied in relation with their unfolding behavior under stretch, which is capable of generating both secondary and tertiary elasticity (reviewed in 34). Because some of the projectin N8Ig domains may belong to the extensible region of the molecule, we also evaluated the likelihood of unfolding for these NH₂-terminal Ig domains, by predicting the propensity of the amino acid sequences to be natively disordered. We used several predictors of intrinsically disordered domains (IDPs) and present in Figure 4 the analysis for one predictor, Foldindex23 for the first 8 Ig domains of representative insect species. Predictions for IDP using IUPRED2122 and PONDR-FIT24 were also performed and can be found in Supplemental Materials. Even though the detailed patterns are different between species and predictors, there are certain regions that are consistently predicted as disordered across all analyzed insects and all predictors. These regions localize at the borders between some of the Ig domains and typically encompass the short linker sequences present between Ig domains (Red arrows in Figure 4). These linkers are short, ranging from four to eleven amino acids and are typically well conserved between insects. Only two linkers in A. mellifera and one in L. pulchella are predicted to be folded (Green arrows in Figure 4). Figure 4 also offers a similar analysis for the first two Ig domains of titin, named Z1/Z2 and the six amino acid linker sequence found between them. Interestingly, the short linker between the two titin Ig domains is also predicted to be intrinsically disordered, whereas the Ig domains themselves have a highly ordered propensity consistent with the X-ray crystallography data3536.

In D. melanogaster there are two alternative exons19 for the linker between Ig1 and Ig2, noted as linker 1a and 1b in Figure 4 top panel. As shown in the top panel of Figure 4, the two forms behave differently as one variant (1a) is predicted to be more disordered than the other (1b, insert in top panel). This same difference in disordered behavior is also detected by IUPRED and PONDR-FIT predictors. In the other species only one Ig1-Ig2 linker has been found by annotation and corresponds in sequence to linker 1a, the most disordered one from D. melanogaster. The existence in other insects of a second linker cannot be excluded even though in the species for which genomic data are available, the intronic regions where that linker should be found were searched extensively by BLAST, as well as manually. So unless the sequence of the second linker is extremely different from the one in Drosophila, it probably does not exist in the other species.

A similar analysis of the N6Ig region also predicts an extended region to be unfolded; it encompasses the end of Ig9, the totality of the FRAM region and approximately half of Ig10 (see Figure 2 for domain arrangement, data not shown).

The conservation at the amino acid level at the beginning of the protein is reinforced at the gene level where the exon-intron pattern of the first four Ig domains is identical in all insects for which genomic sequences are available except in Lepidopterawhere exon #1 is split into 2 exons. On the other hand, the exon-intron pattern of the next four Ig domains is divergent with multiple events of intron loss or acquisition (Figure 5). When the rest of the projectin gene is considered there is overall very little conservation of the exon-intron pattern as reported before and reinforced by the highly variable number of total exons1213.

This exon-intron pattern theoretically allows for two alternative splicing patterns, which both conserve the open reading frame, and lead to the precise removal of two (Ig #3 and 4; referred to as Δ3-4) or three (Ig# 2, 3 and 4) domains. The actual use of these alternate splicing patterns was tested for all insects in RT-PCR reactions using forward primers from the Ig 1 or Ig2 domains and reverse primer from the Ig5 domain. Predicted differences in the size of the amplified products allow clear identification of the “full” product (containing Ig3 and Ig4 domains) or the alternative splice variant Δ3-4. We also tested whether the alternate splice options were muscle-type specific, by performing the RT-PCR reactions from RNA isolated from leg, head and flight muscles. Evidence of the alternate Δ3-4 variant was obtained for all tested insects and all muscle types, except in M. sexta where the Δ3-4 variant is absent from the flight muscles. The amplified products were sequenced, confirming that the shorter cDNA product corresponds to an in-frame removal of Ig3 and Ig4 (Δ3-4). Representative data for M. sexta and Tibicen sp. are presented in Figure 6 and the complete analysis is summarized in Table 2.

In the Manduca sexta analysis, the Δ3-4 variant is clearly amplified as a 439 bp product in lanes from head, leg and thorax RNA, but is absent from flight muscle RNA. The full product at 1,081 bp is also visible in all but the head lane. However longer RT-PCR products are often underrepresented in reactions where a smaller product is amplified. The Tibicensp analysis indicates the presence of the Δ3-4 variant (252 bp) in all muscle types. Again the full product (894 bp) is underrepresented in this analysis and can only be detected in the thorax lane. RT-PCR reactions using primers within the Ig3 and 4 domains indicate that the full length transcript is present in all muscle tissues, even though its amplification is not always evident when using primers from Ig1 or Ig2 and Ig5 as the shorter alternate form can be preferentially amplified in RT-PCR reactions.

A similar analysis to test for the alternate removal of Ig2+3+4 did not reveal the existence of this alternate option in any of the species tested and any of the muscle types (data not shown). This does not preclude, however, that this variant is used in other muscles such as larval or embryonic tissues.

The highly conserved FRAM domain was also shown to be alternatively spliced at least in some of the insects included in this study. In the dragonfly P. longipennis, an alternate

variant without Ig9 and FRAM (referred to as ΔIg9-FRAM) is found in leg and head muscles but not in flight muscles (Table 2). The ΔIg9-FRAM variant was also detected in all muscle types of D. melanogaster, but could not be amplified in any muscles preparation of M. sextaor A. domesticus. We could not detect it from A. mellifera, but this might be more related to the fact that in A. mellifera flight muscles, the PEVK domain is extremely short13 (~ 100bp), and the amplified product (typically from Ig8 to Ig10) would be rendered even shorter by removing the exon containing the end of PEVK and beginning of Ig9. Data for the ΔIg9-FRAM variant are summarized in Table 2. It might be particularly relevant that the only flight muscles where the ΔIg9-FRAM has been detected are the asynchronous flight muscles of D. melanogaster.

Some of the insects included in this study were also analyzed for projectin sarcomeric localization in flight muscles because in Drosophila melanogaster, projectin localization is different in synchronous muscles compared to asynchronous flight muscles. In synchronous muscles such as leg and Larva body wall, projectin co-localizes with myosin over the A-band, whereas it is associated with the I-Z-I region and the C-filaments in asynchronous flight16. The question is whether in synchronous flight muscles such as these of the dragonfly, cricket and moths, projectin will still be associated with the I-Z-I domain, and be a component of C-filaments or follow its synchronous A band pattern. Several projectin antibodies for which the epitope localization is known were used in order to ascertain the myofibrillar position of the protein (See Materials and Methods for list of antibodies). The localization was established in comparison with actin detected by phalloidin staining, and several other proteins, such as α-actinin and kettin using specific antibodies7826.

In the dragonfly P. longipennis, projectin N-terminus detected with antibodies P5 (Figure 7A) and NT2 (not shown) co-localizes with kettin at the I-Z-I band, whereas a C-terminus epitope (3b11) does not (Figure 7B). This distribution indicates that in dragonfly flight muscles, projectin has a similar topography to the one described in more derived insects such as D. melanogaster and A. mellifera678. Figures 7C-E present images for M. sextaflight muscleswith the colocalization of projectin epitopes and α-actinin, as well as two projectin epitopes (7E). In this latter case the muscles might have been slightly stretched during the preparation and there is a splitting of the signal (white arrows in 7E), probably representing the two projectin filaments from adjacent sarcomeres protruding on both sides of the Z-band. In Figure 7F and Figure 7G the visualization of cricket projectin shows that the projectin epitopes are again associated with the I-Z-I band. The images also reveal a splitting of the signal (white arrows in F and G), which is wider and more commonly observed than in the M. sexta images. This may indicate a stretch of the muscles during preparation or possibly a longer sarcomere allowing for better separation of epitopes.

The pattern in all three insects is consistent with the presence of connecting C-filaments in insect flight muscles with synchronous physiology, and even in species which are considered basal phylogenetically (like dragonfly).