Opsin evolution: annotation tricks
Using the Opsin Classifier to compare introns across Lophotrochozoa and Vertebrates
The snapshot of a colored alignment makes clear how to compare intron position and phase in melanopsins from widely diverged species. The magenta coloring in the query sequence from Schistostoma shows 3 exon ends followed by numbering 0,1,2 to indicate phase (basepairs extending beyond the last coding triplet). The Schmidtea ortholog is colored similarly.
Next, the query sequence is run against the full set of opsins in the classifier. The alignment to stickleback is important because the genome project there has reliably established where and how the exon breaks occur. Using the Schistosoma exon termini in a web browser text search locates the homologous peptides in stickleback. These are colored red in the alignment and also where located in the intronated sequence. It turns out exon breaks and phasing exactly match those of Schistosoma (for the 3 exons considered but not all).
However the peptides transfered homologously to tick and bee (and vertebrate ciliary opsins, not shown) fail to correspond to exon breaks. This establishes some specificity -- the matching introns are not some generic property of all GPCR nor all rhodopsin-class receptors
Using the Opsin Classifier to annotate a new opsin in hemichordate
Below is a step-by-step example showing how the Opsin Classifier is used within an overall annotation strategy for recovering intronated opsin homologs in an unstudied species, here the hemichordate Saccoglossus kowalevskii. It's best to follow along actively replicating the steps on actual live web tools.
Here we expect to find minimally an opsin that underlies epidermal photoreception. The best inital queries (for detecting diverged sequences) are likely fellow Ambulacraria (ie sea urchin) opsins. The best initial target is Trace Archives Saccloglossus 'other' because these are transcripts which give longer blastn matches than dispersed exons. However this draws a blank, possibly because cell types with opsin transcripts are exceedingly rare. Hence we move to the "wgs" division.
A flawed but long sea urchin dna query, XM_778236, a GNOMON pipeline product labelled "similar to Go-coupled rhodopsin" elicits a significant hit provided the "Somewhat similar sequences (blastn)" trace query option has been selected. This option is critical for queries across diverged species or diverged opsin classes.
The key feature of this match is length. The rule of thumb is anything exceeding 40 bp will prove informative; here we have 71 bp. The percent identity is quite respectable at 76% given the time elapsed since sea urchin and acornworm divergence -- and this may be even be a match of paralogs. Note however the two gaps. The second can be written off as a slight gap misplacement by blastn. The first cannot. It represents an apparent inactivating frameshift (change in reading frame). However traces contain many artefactual indels and most likely this is sequencing error. We cannot rule out a recent processed pseudogene at this point -- indeed they give better initial hits than a multi-exonic gene.
The trace blast graphic shows five stacked hits. I selected one of these because its length was good at 917 bp and sufficient material was left over on both flanks, that is 754 bp upstream and 94 bp downstream.
>gnl|ti|1695985935 Length=917 Score = 41.0 bits Expect = 9.0 Identities = 54/71 (76%), Gaps = 3/71 (4%) Strand=Plus/Minus
Query 474 CCTT-TTCTGGACTATCACACCGTTCTTTGGATGGAGCAGCTACAC-CTACGAACCATTTGGCACGTCGTG 542
|||| |||||| | || | ||| | ||||||||||||||||| | ||| |||||| || || |||||
Sbjct 163 CCTTATTCTGGTCGATGATGCCGCTGTTTGGATGGAGCAGCTATGCGCTA-GAACCAGAAGGTACATCGTG 94
Next, the retrieved full length trace is back-blastxed against the Opsin Classifier collection. This gives three immediate benefits: properly translating the trace irrespective of its frameshifts (which surface as exon breaks with breaks too short for an intron in the numbering), finding the best available gene model (the initial query choice may have beeen sub-optimal), and extending match length over what Trace Archive blastn could do.
Here we see that the top 15 matches for trace ti|1695985935 are consistently opsins previously classified as peropsins and neuropsins. There appear to be two disjoint segments to the blastx match to PERa_braFlo Branchiostoma floridae but the numbering shows they really reflect a single extended match encompassing a frameshift. Here the trace was on the minus strand which means larger numbers are earlier in the coding sequence. The two fragments can then be joined into a single polypeptide, likely with a one amino acid glitch at the frameshift join. Blast often extends its matches too far into bogus territory and this must be trimmed.
PERa_braFlo Branchiostoma floridae (amphioxus) ?? ... -1 182 3.6e-28 2
PERa_braBel Branchiostoma belcheri (amphioxus) ?? ... -1 174 5.7e-27 2
PER_xenTro Xenopus tropicalis (frog) ?? 0.2.0.2.1... -1 195 4.8e-26 2
NEUR_calMil Callorhinchus milii (elephantfish) ?? ... -2 151 1.9e-24 2
PER_gasAcu Gasterosteus aculeatus (stickleback) ?... -1 174 4.2e-23 2
...
Score = 167 (58.8 bits), Expect = 3.6e-28, Sum P(2) = 3.6e-28 Identities = 28/60 (46%), Positives = 41/60 (68%), Frame = -2
Query: 769 GLTIFGMSLSCVSSFAGRWLFGKFGCYFHGFAGMLFGLGSIGNLTVISIDRYIITCKRNL 590
G+ IFG S SS WLFG GC ++GF GM FG+ +IG LT +++DRY++ C+++L
Sbjct: 83 GICIFGYPFSGASSLRSHWLFGGVGCQWYGFNGMFFGMANIGLLTCVAVDRYLVICRQDL 142
Score = 182 (64.1 bits), Expect = 3.6e-28, Sum P(2) = 3.6e-28 Identities = 35/82 (42%), Positives = 50/82 (60%), Frame = -1
Query: 266 YVYSCNQNFNYKLHLFTEWSYRHYYALLAVAWSNALFWSMMPLFGWSSYALEPEGTSCTIDWMNNDNQYISYVSCVTVTCFI 21
Y+ C Q+ K++ Y Y + A+ W A FW+ +PL GW+ Y+LEP GT+CTI+W ND+ YISYV+ +CFI
Sbjct: 134 YLVICRQDLVDKVN------YNTYGVMAALGWLFAAFWAALPLVGWAEYSLEPSGTACTINWQKNDSLYISYVT----SCFI 205
The two fragments can then be joined into a single polypeptide, likely with a one amino acid glitch at the frameshift join. This 133 residue peptide is blastped against Opsin Classifier. Here the Expect = 5.4e-33 is highly encouraging but the two small insertions raise questions for an opsin. Perhaps extraneous residues have been incorporated into our gene model, currently a fragment corresponding to residues 83-197 of the 361 residue peropsin in amphioxus.
Switching to a six-frame translation view in a second browser tab, we see GLTI is preceded by a stop codons without good opportunities for a GT-AG splice site that does not sacrifice regions of apparent alignment. This could reflect additional frameshifts but without any guidance from blastp output, we have no idea what frame might contain further good upstream sequence. Next we look for exon boundary guidance from the best gene model amphioxus PERa_braFlo. Here we see GLTI is towards the end of an exon. If we look to GRWL as the valid start of the alignment, it lacks a supporting phase 0 intron of PERa_braFlo
Downstream there is a GT splice option for a phase 2 intron at YISYV (which would eliminate the insert) or an option for adding IRSKTDTTFVDT followed by a phase 0 splice start (and a stop codon). However that extension has no support in any known opsin. Further, there is no support from PERa_braFlo exon boundaries.
The information in this trace has not yet been exhausted because we can re-blastn against the trace archives, perhaps finding a trace that stayed on task. Indeed this picks up two high-identity trace with clear multiple exons, ti|1723199539 and ti|1705099698, supporting the notion that the original trace reflects a processed pseudogene. By recursively using the Opsin Classifier with further rounds of trace archive searching (to tile out ends), we eventually arrive at a 246 residue fragmentary gene model with consistent top matches to peropsins and neuropsins, sharing two appropriate introns with correct phasing but also two fused introns with respect to later deuterostomes. This fragment is a good candidate for ortholog as these genes are single copy in all studied species.
The fragmentary gene is best further studied after release of acornworm contig assembly.
>PER_sacKol Saccoglossus kowalevskii Expect = 2.0e-49 PERa_braFlo Identities = 97/246 (39%) IIYYFFLLSTGLTIFGMSLSCVSSF GRWLFGKFGCYFHGFAGMLFGLGSIGNLTVISIDRYIITCKRSL 1 2 WSYRHYYALLAVAWSNALFWSMMPLFGWSSYALEPEGTSCTIDWMNNDNQYISYVSCVTVTCFILPCAVMTYDYLAAYMKMVKAGYTLSEETEKPNND 0 0 MCIALVAAFLLSWFPSATVFLWAAFGNPGNIPLSFTGVADAFTKIPAVFNPVIYVALNPEFRKYFGKTIGCRRKRKKPIAVRLNGSEQNVENTI* 0 PERa_braFlo Branchiostoma floridae (amphioxus) ?? 0.2... 484 2.0e-49 1 PERa_braBel Branchiostoma belcheri (amphioxus) ?? 0.2... 472 3.8e-48 1 PER_monDom Monodelphis domestica (opossum) ?? 0.2.0.... 443 4.5e-45 1 PER_xenTro Xenopus tropicalis (frog) ?? 0.2.0.2.1.0.... 442 5.7e-45 1 PER_gasAcu Gasterosteus aculeatus (stickleback) ?? 0... 436 2.5e-44 1 PER_homSap Homo sapiens (human) ?? 0.2.0.2.1.0.1 ind... 435 3.2e-44 1 PER_galGal Gallus gallus (chicken) ?? 0.2.0.2.1.0.1 ... 418 2.0e-42 1 NEUR_gasAcu Gasterosteus aculeatus (stickleback) ?? 0... 304 3.9e-41 2 NEUR_ornAna Ornithorhynchus anatinus (platypus) ?? 0.... 396 4.3e-40 1 NEUR_monDom Monodelphis domestica (opossum) ?? 0.2.2.... 390 1.9e-39 1 NEUR_homSap Homo sapiens (human) ?? 0.2.2.2.0.1 indel... 383 1.0e-38 1 NEUR_galGal Gallus gallus (chicken) ?? 0.2.2.2.0.1 in... 383 1.0e-38 1 NEUR_xenTro Xenopus tropicalis (frog) ?? 0.2.2.2.0.1 ... 370 2.4e-37 1 NEUR_anoCar Anolis carolinensis (lizard) ?? 0.2.2.2.0... 365 8.3e-37 1
