A quantitative comparison of sRNA-based and protein-based gene regulation

Pankaj Mehta^1,3, Sidhartha Goyal^2,3 & Ned S Wingreen¹

Article highlights

Synopsis

It is now clear that small non-coding RNAs (sRNAs) have a crucial function in prokaryotic gene regulation as both positive and negative regulators. sRNAs are involved in many biological functions, including quorum sensing (Fuqua et al, 2001; Lenz et al, 2004), stress response, virulence factor regulation (Gottesman, 2004; Storz et al, 2004, 2005; Majdalani et al, 2005), and the regulation of outer membrane proteins (Guillier et al, 2006; Vogel and Papenfort, 2006). One major class of prokaryotic sRNAs (antisense sRNAs) negatively regulates proteins by destabilizing the target protein's mRNA. These 100 bp antisense sRNAs prevent translation by binding to the target mRNAs in a process mediated by the RNA chaperone Hfq (Gottesman, 2004; Lenz et al, 2004). On binding, both the mRNAs and sRNAs are degraded (Gottesman, 2004), suggesting that prokaryotic sRNAs—unlike their eukaryotic counterparts—act stoichiometrically on their targets.

Although transcription factor (TF)-based regulation is ubiquitous in prokaryotic gene circuits (Ptashne and Gann, 2001), thus far sRNAs have largely been found in circuits responding to strong environmental cues (e.g. extreme nutrient limitation). This leads to a natural question: are transcriptional regulation by TFs and post-transcriptional regulation by sRNAs distinctly well suited for different biological tasks?

To address this question, we report a quantitative comparison of the signaling properties of TF- and sRNA-based gene regulation. In general, a signaling system can be characterized by how it processes different types of inputs. We therefore treat TF- and sRNA-based regulation as noisy signal processing systems (see Figure 1) with an input signal—the average concentration of the TFs controlling RNA transcription rates—and an output signal—the average level of the regulated protein (Ptashne and Gann, 2001)—and calculate engineering properties of the system such as the steady-state behavior, noise properties, frequency-dependent gain (amplification and noise filtering capabilities), and dynamical response to large input signals (Detwiler et al, 2000).

Figure 1

Genetic regulation through sRNAs. Left: small non-coding RNAs (sRNAs) regulate protein expression as part of a larger genetic network with a specific biological task (e.g. quorum sensing in Vibrio bacteria; Lenz et al, 2004). The sRNAs (stem loops) regulate target proteins by destabilizing target protein mRNAs (wavy lines), a stoichiometric process mediated by the RNA chaperone Hfq (hexagons). When the rate of sRNA transcription _s greatly exceeds the rate of mRNA transcription _m, i.e. when, _sm, nearly all the mRNAs are bound by sRNAs and cannot be translated. By contrast, when _ms, there are many more mRNAs than sRNAs, and protein is highly produced. Right: the stochasticity (randomness) of cellular processes results in noise—statistical fluctuations in the molecular numbers. It is helpful to classify the total noise in the output (output noise) into (i) input noise—noise in the input signal from upstream components in the gene circuit, (ii) intrinsic noise—noise from stochasticity inherent in gene regulation through sRNAs, and (iii) extrinsic noise—all other sources of noise impinging on the signal processing system.

Full figure and legend (205K)Figures & Tables index

The signaling capabilities of a genetic network are limited by intrinsic noise—fluctuations in the regulated protein number for a fixed steady-state input due to the stochastic nature of the underlying biochemical reactions. We find that the intrinsic noise of an sRNA-regulated protein differs significantly from that of a transcriptionally regulated protein. Our analysis shows that over a large range of output protein levels, the intrinsic noise for sRNA-based regulation is much larger than for TF-based regulation due to large amplitude 'near-critical' fluctuations stemming from the stoichiometric nature of the sRNA–mRNA interaction. The intrinsic noise for sRNA circuits is also greatly increased by transcriptional bursting. However, even at a high level of transcriptional bursting, sRNA-based regulation is less noisy than TF-based regulation at low protein levels (in the repressed regimen) because a large pool of sRNAs shortens the effective mRNA lifetime and buffers against target mRNA fluctuations (see Figure 2). Thus, in all cases, protein expression can be kept off much more reliably by sRNAs than by TFs. We also find (when the input signal is coupled to the sRNAs) that sRNAs are better filters of high-frequency input noise than TFs as they implement an extra low-pass noise filter when compared with TFs. This is likely to be physiologically relevant as sRNAs are often found in networks that couple to external signals (Majdalani et al, 2005). In such networks, high-frequency noise in the input could arise from noise in external concentrations or from the fast upstream protein modification reactions such as phosphorylation–dephosphorylation of a two-component system. sRNAs also allow cells to respond quickly to large changes in input signal. In particular, sRNAs can quickly turn off negatively regulated genes and quickly turn on positively regulated genes (Shimoni et al, 2007). This ability to filter high-frequency noise without compromising the ability to respond rapidly to input signals is a defining feature of sRNAs. The above characteristics make sRNA-based regulation useful for constructing genetic switches. In contrast, even for moderate levels of transcriptional bursting, sRNA-based regulatory circuits are worse than TFs at transducing small input signals, suggesting that TFs are likely better suited for quantitative adjustment of protein expression.

Figure 2

Steady-state behavior for gene regulation through sRNAs. For the regulated protein, the steady-state mean number exhibits an approximately threshold linear behavior as a function of the mRNA transcription rate _m. The threshold is set by the sRNA transcription rate _s. Protein expression can be classified into three regimens: repressed (_sm), crossover (_sm), and expressing (_sm). In the repressed regimen, the average protein number is low. By contrast, the protein number increases almost linearly with _m in the expressing regimen. The typical behavior of the noise _p, the standard deviation of the protein number, is shown for the three regimens.

Full figure and legend (172K)Figures & Tables index

The analogy between biochemical circuits and signal processing systems in engineering utilized in this study provides a general framework for characterizing the signal-transduction pathways found in biology. Different biological tasks place different requirements on signal-transduction circuits. For example, in chemotaxis, bacteria must respond quickly to changing input signals, whereas in quorum sensing or stress response, reliability may be more crucial than speed. One suspects that biological networks exhibit a harmony between network architecture and network function. For this reason, understanding the comparative advantages and disadvantages of different network architectures and the components that make up these networks is likely to yield new insights into biological function, as well as new schemes for synthetic circuits.

MORE ARTICLES LIKE THIS

These links to content published by NPG are automatically generated.