We analyzed the transcriptome of Escherichia coli K-12 by strand-specific RNA sequencing at single-nucleotide resolution during logarithmic- growth and upon entry into stationary phase under carbon, nitrogen, and phosphate starvation conditions. To generate high-resolution transcriptome maps, we developed a quantitative method for first annotating and then calculating the three features that define an operon: the promoter, terminator, and deep RNA sequence read coverage to connect the two transcript ends. Based upon the annotation of transcription features we were able to calculate relative promoter activities, terminator efficiencies, and transcription unit activities for 2,122 promoters, 1,774 terminators, and 1,510 operons, respectively. Our analyses revealed an unprecedented view of E. coli operon architecture. A large proportion (36%) of operons are complex with internal promoters or terminators that generate multiple transcription units. We found that 276 of 370 convergent operons terminate inefficiently, generating complementary 3’ transcript ends which overlap on average by 286 nucleotides, and 136 of 388 divergent operons have promoters arranged such that their 5’ ends overlap on average by 168 nucleotides. We found 89 antisense transcripts of 397-nucleotide average length, 7 unannotated transcripts within intergenic regions, and 18 sense transcripts that completely overlap operons on the opposite strand. Of 519 overlapping transcripts, 75% correspond to sequences that are highly conserved in E. coli (>50 genomes). Additionally, we sought to identify and characterize RpoS-dependent operons, genes and promoters under carbon, phosphate and nitrogen starvation. RpoS-dependency was identified using DEseq software. Following differential expression analysis by DEseq, only transcription units, genes and promoters that were statistically significant (p-value ≤ 0.05) and demonstrated a 4-fold or greater change in expression were classified. As a result of our analysis 315 operons, 317 genes, and 278 promoters were classified as being RpoS-dependent. It was observed that RpoS-dependency was most impactful when the culture was starved for carbon, accounting for two-times more differentially regulated transcription units than nitrogen or phosphate starvation. Significant differences in the structure of RpoS-dependent transcripts were observed when compared to RpoS-independent transcripts. It was determined that most RpoS-dependent operons are monocistronic and are approximately half the size of RpoS-independent operons. Analysis of the -10 regions of the 278 putative RpoS-dependent promoters determined that the most abundant nucleotide sequence was CTACGCTTAA, a significant deviation from the consensus motif (CTATAATTAA). We hypothesize that the presence of guanine and cytosine nucleotides (CGC) at base locations -8 through -10 results in the preferential binding of RpoS to these promoter regions, whereas the vegetative sigma factor RpoD would not bind. Additionally, four new RpoS-dependent transcripts were identified within the intergenic regions of the E. coli genome. These results and conclusions describe RpoS-dependency at the operon, gene, and promoter levels, and elucidate the “core” of the RpoS regulon under three different starvation conditions.