Translation in E. coli is usually initiated at an AUG codon, which encodes the amino acid methionine. In some cases, however, the start codon is GUG, which normally encodes valine. If GUG is used as the start codon, is a tRNA charged with methionine used or one charged with valine, and does the use of GUG as a start codon affect the efficiency of translation?
The NCBI translation table translates all alternative start sites as methionines. To my understanding, all translation is initiated by the fMet-tRNA. I don't know if there are any exceptions to this rule.
Regarding translation efficiency, I only found a 1985 paper in PNAS (Reddy et al, PNAS 82:5656-60), in which they compared the translation efficiency of adenilate cyclase's own UUG start vs GUG or AUG, obtaining a translation ratio 1:2:6 UUG:GUG:AUG, suggesting that AUG is the most efficient one, followed by GUG. Also, Romero and Garcia, FEMS microbiology letters 84:325-330 (1991) compared the efficiency of AUG vs AUC, AUA and AUU, showing a much lower efficiency for those codons, but they did not compare it to GUG.