Quantifying heterologous gene expression during ectopic MazF production in Escherichia coli

MazF is a sequence-specific endoribonuclease-toxin of the MazEF toxin–antitoxin system. MazF cleaves single-stranded ribonucleic acid (RNA) regions at adenine–cytosine–adenine (ACA) sequences in the bacterium Escherichia coli. The MazEF system has been used in various biotechnology and synthetic biology applications. In this study, we infer how ectopic mazF overexpression affects production of heterologous proteins. To this end, we quantified the levels of fluorescent proteins expressed in E. coli from reporters translated from the ACA-containing or ACA-less messenger RNAs (mRNAs). Additionally, we addressed the impact of the 5′-untranslated region of these reporter mRNAs under the same conditions by comparing expression from mRNAs that comprise (canonical mRNA) or lack this region (leaderless mRNA). Flow cytometry analysis indicates that during mazF overexpression, fluorescent proteins are translated from the canonical as well as leaderless mRNAs. Our analysis further indicates that longer mazF overexpression generally increases the concentration of fluorescent proteins translated from ACA-less mRNAs, however it also substantially increases bacterial population heterogeneity. Finally, our results suggest that the strength and duration of mazF overexpression should be optimized for each experimental setup, to maximize the heterologous protein production and minimize the amount of phenotypic heterogeneity in bacterial populations, which is unfavorable in biotechnological processes.


Introduction
MazF is the toxin part of the bacterial toxin-antitoxin MazEF module, neutralized by the MazE antitoxin and thus inactive in non-stressful conditions [1,2]. MazF acts as an endoribonuclease that cleaves single-stranded RNA regions at ACA trinucleotide sites in Escherichia coli [3,4]. As artificially produced MazF degrades the rRNA precursors, as well as mRNAs [4][5][6][7][8], overall translation is reduced, which leads to decline in bacterial growth [9]. Even though cells grow slowly during mazF overexpression, cellular processes such as transcription and translation are not halted during this growth reduction [10,11]. A previous study has shown that bacterial populations maintain transcription during MazF production, possibly to ensure synthesis of important components of translational machinery and the antitoxin MazE, so the cells could recover from the stress rapidly [11]. Moreover, during mazF overexpression protein synthesis is possible from mRNAs that do not contain ACA sequences, and thus are not targeted by MazF [10]. Ectopic MazF production has been included in various experimental frameworks in biotechnology and synthetic biology. For instance, mazF overexpression has been employed to improve high-yield production of the protein of interest [12,13], for manipulation of cellular resources [14], and in examining interaction networks within synthetic microbial communities [15]. In this study, we investigate how basic mRNA properties, such as the mRNA sequence and length of the untranslated region, affect synthesis of the heterologous protein during mazF overexpression. Additionally, this study underlines the importance of addressing the connection between expression of the gene of interest and phenotypic heterogeneity in bacterial populations during mazF overexpression.

Bacterial strains and reporter systems
We employed three constitutively expressed reporter systems: a plasmid-based gfp reporter gene devoid of ACA sites (gfp ΔACA ), transcribed into (1) a canonical or (2) a leaderless mRNA [16], and (3) a chromosomally encoded mCherry reporter gene with its native ACA sites [9]. The coding gfp ΔACA mRNA region is thus not targeted by MazF [16], while the mCherry mRNA is prone to the MazF-mediated cleavage [9]. Low-or high-copy plasmids harboring gfp ΔACA reporter systems were transformed into strain TB212, which is a derivative of E. coli BW27784 that constitutively transports l-arabinose (Ara) without metabolizing it [17]. Strain TB212 carries a chromosomally integrated mCherry reporter gene placed under the phage λ promoter [18], and it is additionally transformed with plasmid pBAD-mazF [19]. All gfp ΔACA reporter systems [16,20], strains [17,21] and plasmids are listed in Additional file 1: Table S1. Bacterial growth was monitored by measuring optical density at 600 nm (OD 600 ), and flow cytometry analysis was performed with LSR Fortessa (BD, USA). Detailed experimental protocols, flow cytometry setups and analyses are described in [9].

Fluorescence analysis
As a negative control for GFP and mCherry fluorescence, we measured autofluorescence of strain BW27784 pBAD-mazF. As an additional negative GFP fluorescence control, we employed strain TB212 pBAD-mazF with a gfp ΔACA reporter gene system in which a stem loop structure was placed closely upstream of the start codon to prevent ribosome binding and consequently translation, located on a high-copy or a low-copy plasmid (Additional file 2). Normalized GFP and mCherry fluorescence was calculated as the mean fluorescence level of a TB212 pBAD-mazF gfp ΔACA reporter strain divided by the mean fluorescence level of the reporterless strain BW27784 pBAD-mazF, measured at the same time point. The percentage increase in fluorescence between two measurements, e.g. between two time points t, was calculated on normalized fluorescence values as increase = [(fluoresc ence(t 2 ) − fluorescence(t 1 ))/fluorescence(t 1 )] * 100. Error bars in all graphs present standard deviation. To evaluate differences in fluorescence datasets we used two-tailed, paired Student's t-test (induced vs. uninduced cultures, or 6 h-induction vs. 2 h-induction).

Sequence analysis
A 910 nucleotide-region comprising the mazEF locus was analyzed in the strains K-12 MG1655 (NCBI ID:

Low-level translation of the leaderless gfp mRNA throughout bacterial growth phases
In this study, we sought to determine how the length of the untranslated region (UTR) of an mRNA affects the synthesis of the corresponding protein throughout bacterial growth phases, and specifically during mazF overexpression. Bacterial canonical mRNAs harbor a 5′-UTR comprising ribosome recognition regions and other translational signals, as reviewed in [22]. Leaderless mRNAs lack 5′-UTRs or possess very short 5′-UTRs, and are, in general, translated less efficiently than canonical mRNAs. Nonetheless, previous in vitro and in vivo studies have shown that E. coli leaderless mRNAs can be translated by different ribosome variants [23][24][25]. In our experiments, we analyzed the GFP fluorescence as a proxy for translation of the leaderless gfp ΔACA reporter, referred to as ll-gfp ΔACA reporter, and the canonical mRNA gfp ΔACA reporter, referred to as can-gfp ΔACA reporter. Analysis of the ll-gfp ΔACA reporters in the early exponential phase showed that GFP fluorescence was not significantly higher in populations harboring the fluorescent reporter compared to control populations that did not carry the fluorescent reporter (Fig. 1A, B). This indicates very low ll-gfp ΔACA expression in the early exponential phase, in contrast to the higher levels of expression of the can-gfp ΔACA reporters (Fig. 1C). However, we detected an increased fluorescent signal as a result of translation of the ll-gfp ΔACA mRNA in the later phases of bacterial growth, as well as 2 and 6 h after inducing mazF expression (Table 1, Part A). After 2 h, GFP fluorescence of mazF-induced cultures increased on average by 34% compared to the respective uninduced cultures, when the ll-gfp ΔACA reporter was encoded on a high-copy plasmid (see "Methods" for the calculation). Six hours after mazF overexpression, GFP fluorescence of mazFinduced cultures increased by 133%. These results were further corroborated with the biochemical analysis (Additional file 3: Fig. S1). When the ll-gfp ΔACA reporter was encoded on a low-copy plasmid, GFP fluorescence of mazF-induced cultures did not significantly change 2 h after mazF overexpression, however after 6 h GFP fluorescence of mazF-induced cultures increased by 17%. In parallel, we analyzed GFP fluorescence encoded by the can-gfp ΔACA reporters (Fig. 1C, Table 1, Part A). Six hours after mazF overexpression, GFP fluorescence of mazFinduced cultures increased on average by 79% when the can-gfp ΔACA reporter was encoded on a high-copy plasmid (one replicate culture did not yield a significant GFP fluorescence increase), and by 134% when the cangfp ΔACA reporter was encoded on a low-copy plasmid. In all cases, the increase in GFP fluorescence indicates the fluorescent protein synthesis and its accumulation inside bacterial cells during mazF overexpression. Together, this analysis suggests that translation of leaderless mRNAs occurs throughout the bacterial growth phases as well as during mazF overexpression, albeit at low levels.

Increased cellular concentration of GFP proteins translated from ACA-less mRNAs during mazF overexpression
We analyzed differences in fluorescence intensity of bacterial cells between two measurements during mazF overexpression, i.e. 2 and 6 h after inducing mazF expression ( Fig. 2A, Table 1, Part B). We measured the fluorescence from reporter proteins encoded by the ACA-containing mCherry, the can-gfp ΔACA reporter genes transcribed into canonical mRNAs, as well as the ll-gfp ΔACA reporter gene transcribed into a leaderless mRNA. Our analysis shows that mCherry fluorescence increased only slightly during 4 h of mazF overexpression, by 35% on average (red column in Fig. 2A). In the same experimental setup, GFP fluorescence from the Fig. 1 Flow cytometry analysis of GFP fluorescence encoded by gfp ΔACA reporters. The leaderless mRNA of the ll-gfp ΔACA reporter entirely lacks a 5′-UTR, and this reporter construct has the start sequence ATG of the gfp ΔACA gene following directly after the promoter region [16,20]. The canonical mRNA of the can-gfp ΔACA reporter comprises a 5′-UTR, which includes a strong ribosome binding site. A Green distributions depict measurements of the E. coli strain TB212 harboring the plasmid pBAD-mazF and the ll-gfp ΔACA reporter encoded on a high-copy plasmid. Light grey distributions depict measurements of the strain harboring only the plasmid pBAD-mazF. Here, one replicate is presented, for further results see Additional file 2. Ectopic mazF overexpression from plasmid pBAD-mazF [19] was induced by adding 0.1% Ara in the early exponential phase, at OD 600 = 0.18-0.22. Flow cytometry analysis was performed in the early exponential phase, and 2 h [average OD 600 (uninduced) = 2.45, OD 600 (mazF-induced) = 0.45] and 6 h after mazF overexpression [average OD 600 (uninduced) = 4.42, OD 600 (mazF-induced) = 0.80]. B Normalized GFP fluorescence from the ll-gfp ΔACA reporters or C can-gfp ΔACA reporters encoded on a high-copy (HC, dark green) or a low-copy (LC, light green) plasmid, measured in different phases of bacterial growth and after adding arabinose (Ara) to induce mazF expression (N = 3 independent replicate cultures). Altogether, the growth of mazF-induced cultures was reduced by 77-86% after 2 h, and by 71-90% after 6 h, compared to the respective uninduced controls, see Additional file 2 ll-gfp ΔACA reporter increased by 78% when the reporter was encoded on a high-copy plasmid, and by 12% when encoded on a low-copy plasmid. The highest fluorescence increase was measured for cells carrying the cangfp ΔACA reporter: GFP fluorescence increased by 124% when the reporter was encoded on a high-copy plasmid, and by 113% when encoded on a low-copy plasmid. Firstly, the overall fluorescence increase corroborates previous findings that transcription and translation carry on during mazF overexpression [10,11]. The growth rate reduction during mazF overexpression also indirectly contributes to the increased level of fluorescence because highly stable reporter proteins, such as GFP and mCherry, are less diluted through slower cell division [9]. Secondly, a higher fluorescence increase for cells harboring gfp ΔACA reporters indicates a higher rate of protein synthesis from mRNAs devoid of ACA sites than from ACA-containing mRNAs that can be cleaved by MazF, such as the mCherry mRNA, which was already implied in [10]. Finally, these results suggest considerably higher expression of proteins translated from the canonical mRNA than the leaderless mRNA form during ectopic mazF expression.

Interplay between gene expression and population heterogeneity during mazF overexpression
A previous study has established that fluorescence encoded by the reporter gene placed under the phage λ promoter P R , can be employed to quantify population heterogeneity during mazF overexpression [9]. Specifically, differences in the reporter protein fluorescence reflect changes in the single-cell growth rate, indicating that increased variation in the growth rates of single cells within the population underlies increased population heterogeneity. In this study, population heterogeneity measured as the variation in mCherry fluorescence increased by 32% during 4 h of mazF overexpression (Fig. 2B). In addition, it has been previously shown that the formation of bacterial subpopulations of different fluorescent intensities occurs 5.5-6.5 h after inducing mazF overexpression, regardless of whether the constitutively expressed fluorescent gene reporters are encoded in the chromosome or on a plasmid [9]. Our analysis indicates that bacterial subpopulations exhibited distinct levels of can-gfp ΔACA expression 22 h after inducing mazF overexpression (Fig. 2C). Longer mazF overexpression likewise promoted larger differences in the fluorescence measurements between the replicate cultures (see error bars in Figs. 1C and 2A), and in one case even resulted in an insignificant fluorescence increase in can-gfp ΔACA expression (Table 1), possibly due to pleiotropic effects of mazF overexpression. To conclude, even though prolonged mazF overexpression commonly increases GFP fluorescence encoded by both ll-gfp ΔACA and can-gfp ΔACA reporters measured at the population level, it also promotes bacterial population heterogeneity. In general, E. coli strains overexpressing an ACA-less gene have been used for commercial production of the respective protein under optimized fermentation conditions and mazF overexpression [12,13,26]. Recombinant protein production during mazF overexpression can be maintained for 3 [10], 4 [13] or even 7 days [12]. However, ectopic MazF production non-uniformly alters growth rates of bacterial cells [9]. Bacterial population heterogeneity is typically unfavorable in biotechnological processes as it reduces the yield of recombinant protein production, and causes process instability especially during large-scale cultivation [27]. Inducing mazF expression at lower levels or shorter periods of time has a lesser impact on the population heterogeneity [9,28,29]. Furthermore, one of the most widely used hosts were encoded on a high-copy (HC) or a low-copy plasmid (LC), transcribed to a leaderless (without 5′-UTR) or a canonical (containing 5′-UTR) mRNA that contained ACA sites (mCherry) or was devoid of ACA sites (gfp ΔACA ). The highest increase in fluorescence was detected from the can-gfp ΔACA reporter encoded on a high-copy plasmid, which was almost twofold higher than the fluorescence increase measured from the ll-gfp ΔACA reporter encoded on a high-copy plasmid (N = 3 independent replicate cultures for GFP fluorescence analysis, N = 12 for mCherry fluorescence analysis). B Coefficient of variation (CV) in mCherry fluorescence was calculated as standard deviation divided by the mean of the log 10 -transformed fluorescence data, for different phases of bacterial growth, and it is a proxy for population heterogeneity [9]. CV in mCherry fluorescence increased by 32.4 ± 19.4% in mazF-induced cultures, during 4 h of mazF overexpression (N = 12 independent replicate cultures, p-value = 0.0001). C Green distributions depict measurements of the E. coli strain BW27784 harboring the plasmid pBAD-mazF and the can-gfp ΔACA reporter encoded on a low-copy plasmid. Light grey distributions depict measurements of the strain harboring only the plasmid pBAD-mazF. 0.02% Ara was added to exponentially growing cultures to induce mazF overexpression, and flow cytometry analysis was performed in the early exponential phase, and for recombinant protein production, E. coli BL21(DE3) [10,12,13], contains mazEF locus 100% identical to the mazEF locus of the here studied K-12 strain (see "Methods"). Therefore, a genetically engineered host E. coli strain with deleted mazEF locus could be employed in biotechnological setups, as the presence of the native mazEF locus has been shown to be the main source of population heterogeneity during ectopic mazF overexpression [9].

Conclusions
Our results show that mazF overexpression considerably increases cellular concentration of fluorescent proteins translated from mRNAs devoid of ACA sites. The higher reporter protein fluorescence is observed when the reporter gene is expressed at higher levels, which can be achieved by (1) employing a reporter system with a strong promoter and a strong ribosome binding site, and (2) inserting the reporter system on a high-copy plasmid. This suggests that genetic systems with different transcriptional and translational properties can be used to study cellular resource allocation during mazF overexpression [14]. Moreover, current efforts in synthetic biology and biotechnology focus on engineering bacterial systems with reduced phenotypic population heterogeneity [30,31]. In order to avoid increased population heterogeneity emerging during ectopic MazF production, and to maintain the stability of recombinant protein synthesis, it is necessary to optimize experimental setups that employ MazF by adjusting the strength and duration of mazF overexpression.

Limitations
This study would benefit from further analysis of different types of fluorescent reporter systems in different E. coli strains, to provide a better understanding of the limits of experimental frameworks when employing mazF overexpression for the production of the specific protein and manipulation of synthetic circuits.