We built and tested two different bifunctional reporter designs (Figure S1). In the first design, the N-terminus of the β-gal α-fragment (BBa_E0038) is fused directly to the C-terminus of full-length GFP (BBa_E0043). In the second design, the N-terminus of full-length GFP is fused to the C-terminus of the β-gal α-fragment via a linker (Figure 1b). Initial experiments using the first design yielded functional enzymatic activity, but poor fluorescence activity (not shown). Initial experiments with the second design yielded marked fluorescence and enzymatic activity. We named the second design Gemini and studied it further.

We first confirmed that Gemini encodes functional fluorescence and enzymatic activities using microscopy and X-gal plates, respectively. We wanted to fairly compare the fluorescence and enzymatic activities of Gemini to the individual activities of GFP and the β-gal α-fragment alone across a range of commonly used expression levels. To do this, we placed the coding sequences for each of the three reporters under the control of the same expression cassettes and vector backbones. Specifically, we placed all three reporters under the control of a medium strength RBS (BBa_B0032) and four different promoters of variable strength (in order of decreasing strength BBa_J23119, 101, 106, 115) on a low-copy vector (BBa_PSB4A5).

Thus, we built four promoter variants for each reporter: Gemini (Figure 1C), the β-gal α-fragment alone (Figure S2a), and GFP alone (Figure S2b). The four promoters were chosen for two reasons. First, the four promoters are commonly used standard BioBrick parts. Second, the four promoters span a wide range of transcriptional activity (Methods).

To test enzymatic activity we grew cells containing promoter variants for Gemini, GFP, and the β–gal α-fragment on a plate containing X-gal and IPTG. The plate-based experiments indicated that the enzymatic activity of Gemini is stronger than that of β-gal α-fragment alone (Figure 2a). To test fluorescence activity, we grew cells containing Gemini, GFP, and α-fragment expression constructs with promoter J23101 in supplemented M9 media at 37°C. We sampled an aliquot in mid-exponential growth phase for microscopy measurements, which indicated that the fluorescence activity of Gemini is weaker than that of GFP alone (Figure 2b).

Next, we more carefully measured the fluorescence activity of Gemini relative to GFP alone. We used a Wallac Victor 3 fluorimeter to measure absorbance (Figure S3) and fluorescence (Figure S4) during a period of log phase growth (Methods). For GFP promoter variants, we measured fluorescence activity as the change in GFP molecules per cell per unit time (Figure S5a). For Gemini promoter variants, we report the fluorescence activity as the change in molecules of GFP equivalent per cell per unit time (Figure S5b). We use molecules of GFP equivalent for two reasons. First, it allows direct comparison with the activity measured for GFP alone. Second, we do not have a calibration curve that allows us to calculate molecules of Gemini from arbitrary fluorescence data. We averaged fluorescence activity for each promoter variant across thirty minutes of mid-log phase growth for both GFP and Gemini (highlighted regions in Figure S5). We report fluorescence activities with error bars that represent the standard deviation of the averaged activity measurements (Figure 3a).

We measured the enzymatic activity of Gemini relative to β-gal α-fragment alone (Figure S6). We used a fluorogenic substrate for the β-galactosidase enzyme to quantify enzymatic activity in a Wallac Victor 3 fluorimeter (Methods). For the β-gal α-fragment and Gemini variants, we fit the fluorescence data with a linear regression. From this regression, we determined the enzymatic activity as the change in arbitrary fluorescence with respect to time (Figure S7 and Figure S8). We calculated the average enzymatic activity for β-gal and Gemini across the three replicates for each promoter variant (Figure 3b). We report enzymatic activities with error bars that represent the standard deviation of the averaged activity measurements.

We measured the relative activity of each promoter via the fluorescence activity of GFP alone and Gemini, and via the enzymatic activity of the β-gal α-fragment alone and Gemini. We used promoter J23101 as a reference standard for all relative promoter measurements by dividing each measured activity (Figure 3) by the activity of promoter J23101 for the same reporter (Methods). We report relative promoter units (RPUs) with error bars that represent the standard deviation of the relative activity measurements (Figure 4).

We describe the construction and quantitative characterization of Gemini, a fusion of the β-galactosidase (β-gal) α-fragment to the N-terminus of full-length green fluorescent protein (GFP) that is designed to support biological part characterization in microbes. We demonstrate that Gemini exhibits functional β-gal activity, which we assay with plates or fluorimetry, and GFP activity, which we assay with fluorimetry or microscopy. We show that complementation with the α-fragment fused to the N-terminus of full-length GFP is successful, and that Gemini exhibits more sensitive enzymatic activity than the β-gal α-fragment alone. We show that RPU measurements obtained via the fluorescence activity of Gemini are reasonably consistent with RPU measurements obtained via the fluorescence activity of GFP. We provide a calibration curve that allows engineers to relate RPU measurements obtained via the fluorescence activity of Gemini with RPU measurements obtained via enzymatic activity of Gemini.

The enzymatic activity of Gemini is more sensitive than the activity of the β-gal α-fragment alone and the fluorescence activity of Gemini is less sensitive than the activity of GFP alone. This observation may be explained by the relative stabilities of GFP and the β-gal α-fragment alone. GFP is a relatively stable reporter; for example, the reported in vivo half-life of GFP is ∼26 hours in cultured mouse cells [19]. The β-gal α-fragment is known to be a relatively unstable reporter; for example, the reported enzymatic activity of the α-fragment is ∼25% that of full-length β-galactosidase [20]. With these points in mind, the α-fragment may be stabilized by its fusion to GFP, explaining the greater sensitivity of the β-gal α-fragment in the context of Gemini relative to the β-gal α-fragment alone. Conversely, GFP may be relatively destabilized by fusion to the α-fragment. Regardless of the mechanism, Gemini exhibits a wider dynamic range than either GFP or the α-fragment alone. Because it is more sensitive, the enzymatic activity of Gemini can be used to measure the activity of promoters that cannot be easily detected with the β-gal α-fragment alone (for example, Figure 2a, promoter J23115). Because the fluorescence activity of Gemini is weaker than that of GFP alone, Gemini may be used to measure stronger promoter activity than GFP alone; under the control of a strong promoter, GFP alone may saturate whereas the weaker fluorescence activity of Gemini may still be quantifiable (not shown).

Calculating the activity of a promoter that gives rise to a measured reporter protein synthesis rate requires a mathematical model. The parameter values within the model are sensitive to the experimental conditions, which often change from lab to lab and are generally difficult to measure. Thus, the transcriptional activity calculated for the same promoter often differs across laboratories [2]. Yet even if the parameter values governing reporter synthesis rate under a specific set of experimental conditions are unknown, the values may not change significantly between different cultures measured simultaneously within the same laboratory. The reporter synthesis rate calculated for a culture containing an uncharacterized promoter can be compared to the synthesis rate calculated for a culture containing a standard reference promoter grown in parallel. Ideally, the resulting ratio captures only the relative transcriptional activity between the uncharacterized promoter and the reference promoter, since all relevant parameters, except the promoter itself, are consistent between the two experiments. With this in mind, relative promoter units (RPUs) were developed to improve consistency in promoter activity measurements across laboratories and conditions. Using such as approach, a common reference promoter (J23101), and GFP, one study reduced the reported variability in measurements of promoter activity across laboratories by 50% [2].

Ideally, the RPU framework could be extended to reporter proteins besides GFP, so that different laboratories could make consistent RPU measurements with arbitrary reporters, and so that the synthetic biology community would only have to agree upon a common reference promoter. Clearly, parameter values such as maturation and degradation rates differ between reporter proteins. For example, there is a marked difference between the absolute fluorescence activity of GFP alone and Gemini, when both are expressed under the control of the same strong promoters J23119 and J23101, ceteris paribus (Figure 3a). Even if the parameter values that quantify the expression, maturation, and degradation rates of a reporter protein, such as Gemini, are unknown, the values may not change significantly when the reporter is placed under the control of different promoters. If these parameters do not change with respect to promoter activity, then RPU measurements obtained via Gemini should capture the relative transcriptional activity between the test and reference promoter, and should be consistent with measurements obtained via GFP alone.

RPU measurements obtained via the fluorescence activity of GFP are consistent within the bounds of measurement error with RPU measurements obtained via the fluorescence activity of Gemini (Figure 4a). This suggests that parameter values such as maturation and degradation rates for Gemini are reasonably insensitive to changes in promoter activity. If these parameters were sensitive to changes in promoter activity, a more significant divergence between RPU measurements obtained via GFP and Gemini would be expected. Instead, the results suggest that RPU measurements obtained via reporters other than GFP can be consistent with GFP, therefore extending the RPU framework to fluorescent reporters besides GFP.

However, we found an inconsistency between RPU measurements obtained via the fluorescence and enzymatic activities of Gemini. Specifically, we observed that at high expression levels the relative promoter measurements obtained via the fluorescence activity of Gemini (Figure 4a) diverge from the relative promoter measurements obtained via the enzymatic activity of Gemini (Figure 4b). One explanation for this divergence is that the enzymatic activity of Gemini may saturate if the quantity of expressed Gemini-encoded α-fragment exceeds the amount of complementing omega-fragment present within cells. Careful characterization of a functional full-length β-gal GFP fusion (unpublished results, Kortemme Lab, UCSF) would help to consider this model.

Finally, any future efforts to improve Gemini should note that Western blot experiments on cell extracts revealed two bands for Gemini (Figure S9). One band was near the size expected for Gemini (∼35.7 kDa) and the other appeared to be a cleavage product (∼30.7 kDa). Further Western blot experiments using a protocol optimized to improve size discrimination across this mass range yielded a band for the cleavage product at ∼30.1 kDa (Figure S9b). Given that the predicted mass for GFP plus the linker is ∼28.7 kDa, we expect that Gemini is being cut within the β-gal α-fragment (that is, ∼1.4–2.0 kDa beyond the linker). Thus, for Gemini and perhaps for any full-length β-gal GFP fusion, proteolytic cleavage may decouple the 1∶1 stoichiometric coupling of the two reporter epitopes. For now, and until perfect 1∶1 stoichiometric coupling can be demonstrated, RPU measurements obtained via the enzymatic activity of Gemini can be related to RPU measurements obtained via the fluorescence activity of Gemini using calibration curves (such as Figure S10).

We assembled Gemini via overlap PCR. We made constructs containing GFP (BBa_E0043) and LacZ-α (BBa_E0038) via separate PCR reactions. We performed one PCR reaction using LacZ-α as template with the primers one through three (below). We performed a second PCR using GFP as template with primers four through six (below). We gel purified each PCR product. We used a third PCR reaction to stitch the two products together using primers one and six to form the bifunctional reporter, Gemini (BBa_E0051). We Qiaquick gel purified the PCR product, digested it EcoRI and PstI restriction endonucleases, and cloned the digested PCR product into BioBrick vector, pSB1A3.

(1) tgccacctgacgtctaagaa

(2) gatgttcgctgccgccgccttcgctgccgcccgcgcgccattcgccattc

(3) cgctgccatgatgatgatgatgatgttcgctgccgccgccttc

(4) gaacatcatcatcatcatcatggcagcgaacgtaaaggagaagaacttttcac,

(5) gaaggcggcggcagcgaacatcatcatcatcatcatggcag

(6) attaccgcctttgagtgagc

We added RBS B0032 to Gemini via PCR. We performed PCR of the bifunctional reporter plasmid template using primers 6 and 7. We Qiaquick gel purified this PCR product, digested it with XbaI and PstI, and then performed a Qiagen ERC cleanup on the cut PCR product. We cut the destination vector pSB4A5 with EcoRI/PstI and then performed gel purification on the cut vector.

(7) gaattctctagagtcacacaggaaagtactagatgaccatgattacggattcac

We ligated different promoters to the digested PCR products containing the B0032 RBS and Gemini. We chose standard BioBrick promoters J23119, J23101, J23106, and J23115, because these promoters span a wide range of transcriptional activity and are commonly used by synthetic biologists. We synthesized both strands of each promoter variant as oligos containing an EcoRI overhang in front and SpeI overhang in the back. We normalized these oligos to 50 µM. For each oligo, we added 1 µl of the oligo to a 10 µl reaction with 1 µl 10x T4 DNA ligase buffer, 8 µl water, and 0.5 µl T4 PNK to phosphorylate the 5′ ends. After PNK treatment, we mixed the two complementary oligos together, heated them to 80°C, and allowed them to cool slowly. We ligated these promoters with the cut pSB4A5 vector and PCR products to form each of the promoter variants.

We wanted to fairly compare the fluorescence and enzymatic activities of Gemini to the individual activities of GFP and the β-gal α-fragment alone. To do this, we placed the coding sequences for GFP (BBa_E0043) and LacZ-α (BBa_E0038) under the control of the same expression cassettes, and on the same vector backbone, as Gemini. We used the Bricklayer assembly service provided by Ginkgo BioWorks to build the four GFP and four LacZ-α promoter variants, resulting in three sets of expression constructs: Gemini (Figure 1C), the β-gal α-fragment alone (Figure S2a), and GFP alone (Figure S2b).

We transformed the twelve expression constructs into E. coli TOP10 cells (Invitrogen). We grew the transformed E. coli on LB agar plates containing ampicilin (1∶1000 dilution) antibiotic. After overnight incubation, we inoculated 5 ml test tubes containing LB medium and ampicilin (1∶1000 dilution) with a single colony for each expression construct. After overnight incubation, we purified plasmid DNA for each expression construct using a mini-prep kit (Qiagen) and sequenced all expression constructs.

We used supplemented M9 medium (M9 salts, 1 mM thiamine hydrochloride, 0.2% casamino acids, 0.1 M MgSO4, 0.5 M CaCl2) with glycerol (0.4%) added as a carbon source and ampicilin (1∶1000 dilution) antibiotic for measurements of fluorescence activity for all expression constructs. We inoculated three test tubes containing 5 ml of sterile supplemented M9 medium with a single colony for each of the twelve reporter constructs. As a result, we grew three replicates for each of the twelve expression constructs. We grew the 36 cultures for approximately 18 hrs at 37°C. We then diluted each culture by a factor of 1∶100 into 5 ml of fresh media. We grew the diluted cultures for approximately five hours under the previous conditions.

After five hours, we measured the optical density of a 200 µl aliquot from each culture on a Wallac Victor3 multi-well fluorimeter (Perkin Elmer). We converted the optical density measurements from the fluorimeter to OD600 measurements using a standard calibration curve (not shown). We diluted the 36 cultures to the same optical density in 5 ml of fresh media and grew them for one hour at 37°C. We transferred a 200 µl aliquot from each of the 36 cultures into a well on a flat-bottomed 96 well plate (Nunc). We incubated the plate in a Wallac Victor3 multi-well fluorimeter at 37°C and assayed the samples with an automatically repeating protocol of absorbance measurements (600 nm absorbance filter, 0.1 second counting time), fluorescence measurements (485 nm excitation filter, 525 nm emission filter, 0.1 second measurement time), and shaking (20 second interval between measurements with 2.0 mm shaking diameter and double orbital shaking type).

We determined background absorbance by measuring wells containing only media, and background fluorescence by measuring the fluorescence of TOP10 cells without a GFP expressing vector. After background subtraction, we converted time-series fluorescence (Figure S4) and absorbance (Figure S3) measurements to GFP molecules per well and CFUs per well respectively, using standard calibration curves previously reported for the fluorimeter used in these experiments [1]. We calculated GFP synthesis rates (GFP molecules per cell per second) for all expression constructs containing GFP (Figure S5a). We also calculated GFP synthesis rates (molecules of GFP equivalent per cell per second) for all expression constructs containing Gemini (Figure S5b). We used molecules of GFP equivalent were used for two reasons. First, it allows direct comparison with the activity measured for GFP alone. Second, we do not have a standard calibration curve that allows us to calculate molecules of Gemini from arbitrary fluorescence data. We averaged the synthesis rates across thirty minutes of mid-log phase growth for both GFP and Gemini promoter variants (highlighted regions in Figure S5). We report the resulting fluorescence activities for GFP and Gemini with error bars that represent the standard deviation of the averaged activity measurements (Figure 3a).

After measurement of fluorescence activities, we measured enzymatic activity for each reporter in the three sets of expression constructs. We used E. coli TOP10 cells that contain the LacZ omega fragment necessary for LacZ-α complementation. We induced the LacZ omega fragment by growing cells in supplemented M9 medium with IPTG inducer (1∶1000 dilution). We used the flourogenic substrate 4-methylumbelliferyl-D-galactopyranoside (Sigma Aldrich #M1633), otherwise known as MUG, to measure β-gal activity. We dissolved the MUG in DMSO at a concentration of 2 mg/ml and used the resulting solution as a 10x stock. After measurement of fluorescence activities, we transferred a sample from each culture (10 µl) to a new well containing 20 µl of the MUG stock solution and PBS 170 µl for a total final volume of 200 µl. We were careful to ensure that cell density was consistent between the cultures prior to measuring β-gal activity (Figure S6). We measured fluorescence produced by reaction between β-gal and the MUG substrate using an automatically repeating protocol of measurements (355 nm excitation filter, 460 nm emission filter) on the Wallac Victor3 multi-well fluorimeter (Figure S7). We fit a linear regression to the time-series fluorescence data to measure the enzymatic activity for the β-gal α-fragment and Gemini expression constructs (Figure S8).

We measured the relative activity of each promoter via the fluorescence activity of GFP alone and Gemini, and via the enzymatic activity of the β-gal α-fragment alone and Gemini. We performed relative measurements using the method reported by Kelly et al [1]: we divided the measured activity (Figure 3) for each promoter by the activity of promoter J23101. We did this for the fluorescence activity of GFP alone, the fluorescence activity of Gemini, the enzymatic activity of the β-gal α-fragment alone, and the enzymatic activity of activity of Gemini. We report promoter activities in relative promoter units (RPU) with error bars that represent the standard deviation of the relative activity measurements (Figure 4).