21Development of eye tissue is initiated by a conserved set of transcripton factors 22 development, but may act to buffer Glass expression level. Moreover, we assessed 80 the requirement and functionality of the three Glass protein isoforms by CRISPR-81 mediated genome editing introducing deletions into the glass locus resulting in the 82 loss of one or two of the three isoforms. We analyzed these isoform mutants for the 83 morphology of their eyes, the expression of photoreceptor markers that depend on 84Glass function, photoreceptor activity, and light preference behavior. We found that 85 the short Glass PB isoform is not able to confer normal eye development and 86 function resulting in a glass mutant phenotype, while the Glass PA isoform alone is 87 fully functional. Our results suggest that the expression of glass is tightly regulated as 88 the development of a functional tissue is surprisingly robust resulting in no detectable 89 change in the physiological response or alteration in photoattraction behaviour upon 90 deletion of the smORF. Similarly, only one of the isoforms is critical for eye 91 development, suggesting that the other isoforms may function in a similar way to 92 control levels of the functional protein. Since sequence comparison to closely related 93 species show conservation of these features, such mechanisms may function to fine-94 tune gene expression. 95 96 RESULTS 97 98 An overlapping upstream open reading frame inhibits the expression of a glass 99 reporter 100 In the developing eye, expression of glass is initated at the morphogenetic furrow in 101 the eye-imaginal disc of third instar larvae and is detectable in the nuclei of all cells 102 posterior to the morphogenetic furrow [7]. The same expression pattern is obtained 103 with a reporter construct containing a 5.2 kb DNA fragment upstream of glass [8], 104spaning from -4190 bp to the AUG at +960 (Fig. 1A) [8]. Suprisingly, using this 5.2 kb 105 upstream genomic sequence to drive a GFP reporter we observed GFP expression 106 was barely detectable in the eye imaginal discs (Fig. 1B, B''). By increasing the gain 107 at the confocal microscope, we were able to detect a weak GFP signal posterior to 108 the morphogenetic furrow, barely above background level ( Fig. 1B''', B''''). A closer 109 inspection of our reporter construct revealed the presence of two potential start 110 codons in the 5'UTR of glass, that were also present in the GFP reporter construct, 111 one at position +889 relative to the predicted transcription start, the other at position 112 +955. Translation from the first start codon, if functional, may compete with the GFP 113