Single-fluorophore membrane transport activity sensors with dual-emission read-out

We recently described a series of genetically encoded, single-fluorophore-based sensors, termed AmTrac and MepTrac, which monitor membrane transporter activity in vivo (De Michele et al., 2013). However, being intensiometric, AmTrac and Meptrac are limited in their use for quantitative studies. Here, we characterized the photophysical properties (steady-state and time-resolved fluorescence spectroscopy as well as anisotropy decay analysis) of different AmTrac sensors with diverging fluorescence properties in order to generate improved, ratiometric sensors. By replacing key amino acid residues in AmTrac we constructed a set of dual-emission AmTrac sensors named deAmTracs. deAmTracs show opposing changes of blue and green emission with almost doubled emission ratio upon ammonium addition. The response ratio of the deAmTracs correlated with transport activity in mutants with altered capacity. Our results suggest that partial disruption of distance-dependent excited-state proton transfer is important for the successful generation of single-fluorophore-based dual-emission sensors. DOI: http://dx.doi.org/10.7554/eLife.07113.001

In AmTrac, the cpEGFP was inserted between residue 233 and 234 of the Arabidopsis thaliana ammonium transporter AtAMT1;3 ( Figure 1A, top). The AmTrac sensors differ only by two residues at the N-terminus of the cpEGFP, also referred to as left linker, the right linker being composed of phenylalanine (F) and asparagine (N) in all sensors. The first AmTrac engineered, also named AmTrac-LE, carries leucine (L, Leu 234 ) and glutamate (E,Glu 235 ) in the left linker. Brighter versions carry a serine (S,Ser 235 ) in the second position of the left linker, and variable amino acids in the first position. AmTrac-GS carries a glycine (G, Gly 234 ), AmTrac-LS a leucine (L, Leu 234 ) and AmTrac-IS an isoleucine (I, Ile 234 ) ( Figure 1A).
Despite the minimal differences in their amino acid sequence, AmTrac-GS, -LS, -IS and LE differ greatly in the shape and intensities of their fluorescence spectra, when expressed in yeast ( Figure 1B). The fluorescence spectra of all AmTracs show two excitation maxima, at λ exc ∼390 nm and λ exc ∼500 nm,  also referred to as A-and B-band hereafter. In AmTrac-LE, the A-state dominates the fluorescence spectrum, whereas all other AmTracs (-GS, -LS and -IS) show an equilibrium shift toward the B-state. In wild type GFP (wtGFP) (44), the A-band derives from the A-state, the protonated chromophore, while the B-state indicates the deprotonated chromophore (43). Both states are coupled by a protonation equilibrium in both wtGFP and the AmTrac sensors.
The generally accepted mechanism of single-FP sensors assumes changes in the solvent accessibility and consequent quenching of the fluorescence (Akerboom et al., 2009;Chen et al., 2013). However, AmTracs are substantially different from the soluble sensors characterized so far. Membrane localization of the sensors reduces motion and mobility. Therefore, we performed timeresolved fluorescence measurements to detect if non-radiative processes such as solvent quenching or homo-FRET are involved in the sensor response. This analysis was performed for AmTrac-GS and -LE ( Figure 1C,D). AmTrac-GS showed single exponential decay kinetics with lifetime values of τ = 2.55 ± 0.04 ns and τ = 2.48 ± 0.1 ns when treated with water and ammonium, respectively ( Figure 1C top and Figure 1-figure supplement 1). These values are in agreement with most reports on a single lifetime of about τ = 2.5 to 2.8 ns for EGFP in solution (Stepanenko et al., 2004;Arpino et al., 2012). AmTrac-LE showed bi-exponential decay kinetics with fluorescence decay times of τ 1 = 1.96 ± 0.4 ns and τ 2 = 3.98 ± 0.32 ns and τ 1 = 1.69 ± 0.07 ns and τ 2 = 3.96 ± 0.17 ns when treated with water or ammonium, respectively ( Figure 1C bottom and Figure 1-figure supplements 1, 2). The two lifetime components found for AmTrac-LE are indicative of the presence of two species in the excited state. We thus conclude that the response mechanism of AmTrac-LE vs -GS differs with regard to which state (A-or B-state) is excited, and postulate that either different excited state reactions and/or different excited state species are involved. Since the lifetime values in absence and presence of ammonium differed, at most, by less than 1 ns for AmTrac-LE and -GS (Figure 1-figure supplement 1) we exclude external quenching effects due to altered solvent access in AmTracs, which is the widely accepted mechanism of single-FP sensors.
The distance of the cpEGFP moieties in the sensors allows for homo-FRET, but anisotropy decay kinetics changed only minimally in the presence of ammonium for AmTrac-GS and -LE ( Figure 1D). The large initial r-values found are most likely artificial and probably due to weak signal of the green fluorescence and strong light scattering of the yeast cells. We thus conclude that the FI decrease is not the result of a change in the efficiency of (a possible) homo-FRET and postulate that structural rearrangements affect the protonation states and thus the FI.
Excitation of the A-state of wtGFP leads to green emission due to excited-state proton transfer (ESPT) (Chattoraj et al., 1996). ESPT in wtGFP is a photophysical phenomenon that describes the rapid deprotonation of the neutral chromophore in the excited state leading to loss of blue fluorescence and formation of a green-emitting intermediate state (I-state). The proton travels from the chromophore hydroxyl group, via a buried water molecule and neighboring side chain serine 205 (Ser 205 ), to the final acceptor Glu 222 after excitation of the A-form. ESPT is usually prevented in EGFP due to the S65T mutation (Brejc et al., 1997).
To test for ESPT in AmTracs, we recorded fluorescence emission spectra of yeast transformed with AmTrac-LE and -GS after excitation of the A-state (λ exc 395 nm), in presence of varying ammonium concentrations. Green fluorescence was detected, further supporting the hypothesis of ESPT occurrence ( Figure 2B and Figure 2-figure supplement 1A). However, AmTrac-LE failed to fully respond to ammonium treatment when the A-state was excited, since the FI decreased by 30% only ( Figure 2B). Conversely, when the B-state was excited, the FI decreased up to ∼50% (Figure 2A). The emission spectrum of AmTrac-LE was characterized by another unique feature: the excitation of the A-state showed two emission maxima, one higher energy shoulder-like maximum at λ em 490 nm and a distinct peak at λ em 515 nm. Upon ammonium addition, the relative intensity of the peaks was altered and an iso-emissive point was observed ( Figure 2B). Interestingly, this iso-emissive behavior was detected for AmTrac-LE but not for AmTrac-GS ( Figure 2-figure supplement 1A).
Crystal structures of AmTrac sensors are not available yet. However, crystal structures have been solved for cpEGFP, as free protein or within sensors (Wang et al., 2008;Akerboom et al., 2009;Leder et al., 2010) and can provide clues on how the linker residues are involved in the structural rearrangements during the sensing mechanism. The crystal structure of cpEGFP (PDB 3EVP) shows two rotamers for Glu 148 in close proximity to the chromophore Thr65-Tyr66-Gly67, which indicates a high degree of flexibility at position 148. We assume that due to re-arrangements in the cpEGFP barrel, Glu 148 may exist in different orientations and distances to the phenyl-group of the chromophore ( Figure 2C). Since Glu 148 corresponds to Glu 235 in AmTrac-LE, we consider the solved structure of cpEGFP as a proxy for AmTrac-LE. Theoretical modifications of the linker sequence were performed using Pymol by replacing the linker residues LE of cpEGFP (PDB 3EVP) with GS. The result shows only one conformational state for serine of the GS linker ( Figure 2-figure supplement 1B) indicating more stability during conformation changes. This also indicates the impact of different amino acid linkers on the structural environment of the chromophore. We postulate that Glu 235 of AmTrac-LE relocates during the sensing process, most likely altering the efficiency of ESPT and in turn leading to blue and green fluorescence emission.
The spectral behavior of AmTrac-LE shows similarities with the dual-emission pH sensor deGFP1 (Hanson et al., 2002) as well as the GFP mutant S65T/H148E (Shu et al., 2007). Therefore, we reasoned that it should be possible to generate improved dual-emission sensors with a larger ratio change compared to AmTrac-LE. We modified the left linker amino acids only, since they appeared to be involved in the dual-emission behavior. We replaced LE with two random amino acids, to check for all possible combinations, and analyzed these constructs in yeast. Out of more than 500 yeast colonies screened, we found four candidates with the new linker composition CP, FP, RP, YP that displayed distinct dual-emission patterns and two of these candidates showed improved ratios upon ammonium treatment. Interestingly, all four constructs, named deAmTracs, contained a proline residue at the linker position of the former Glu 235 , while the first amino acid of the linker varied (Figure 3 and Figure 3-figure supplement 1). The two candidates with the highest ratio changes, deAmTrac-CP and -FP, are described in more detail (Figure 3). deAmTrac-CP carried a cysteine and proline, while deAmTrac-FP contained phenylalanine and proline in the left linker. Upon excitation with λ exc 395 nm, blue and green emission maxima at 490 nm and 515 nm, respectively, were detected, which showed opposing changes in FI upon ammonium treatment. A clear iso-emissive point in the dual-emission spectra was observed ( Figure 3A). Plotting the response ratio of the intensity of the A-state /I-state vs the external ammonium concentration showed that the ratio values almost doubled for both deAmTrac-CP and -FP ( Figure 3B). The affinity constants from the response ratio towards ammonium for deAmTrac-CP (EC 50 54 ± 9 μM) and deAmTrac-FP (EC 50 36 ± 6 μM) were similar to AmTrac (EC 50 55 ± 7 μM) (De Michele et al., 2013). Due to the appearance of blue emission we speculate that proline or glutamic acid at position 235 are involved in hindering deprotonation or stabilizing the protonated state of the chromophore. Theoretical modification of the LE linker in the crystal structure of cpEGFP (PDB 3EVP) for either CP or FP using Pymol yields different rotamers (Figure 3-figure supplement 2). We postulate that the residues CP and FP allow for more flexibility due to the possibility of different conformational states. Differently from GS in AmTrac-GS, these linker residues may partially disrupt the ESPT pathway, preventing rapid proton transfer and trapping the excited A-state which consequently emits blue light. The blue emission is rather small, since the green emission from successful ESPT dominates the spectrum.
Several lines of evidence demonstrated that AmTrac's fluorescence response is strictly correlated to the ammonium transporter activity (De Michele et al., 2013). To verify that such correlation was maintained in deAmTracs, we generated a series of mutants with altered transporter activity, based on the suppressor screen for the inactivating mutation T464D in AmTrac-LE (De Michele et al., 2013). In deAmTrac-CP and -FP, the T464D mutation was still sufficient to block growth and response. In the T464D mutant background the presence of the suppressing mutation A141E increased the ratiometric response and transport capacity, shown as reduced growth of the mutants on high ammonium concentrations. Conversely, mutation Q61E restored growth and response of deAmTrac-T464D, although not to the full extent (Figure 4-figure supplement 1). Overall, the mutations had an effect on transport activity and response similar to AmTrac-LE (De Michele et al., 2013) ( Figure 4A and Figure 4-figure supplement 1). Spectral analysis of the mutant deAmTracs showed the impact of the introduced mutations on the dual-emission pattern. In the T464D mutant the blue emission disappeared, and only green emission was detected. In presence of A141E, blue emission increased and green emission decreased. Q61E shows the opposite effect with decreased blue and increased green emission at high ammonium concentrations ( Figure 4B). These results support the utility of deAmTracs in assessing ammonium transporter activity. However, our findings also point out that such sensors are complex systems and that a single mutation affecting the structure and function of the transporter may also affect the direct environment of the chromophore.
Based on spectral analyzes and the ESPT model reported by Chattoraj et al. (1996), we postulate a mechanistic model with three co-existing processes: (i) structural rearrangements in the sensor protein upon ammonium perception, (ii) a predominantly protonated state of the cpEGFP chromophore and (iii) partial disruption of the ESPT pathway to prevent efficient proton transfer. In the absence of ammonium transport activity, an efficient hydrogen bond network is established in close proximity to the chromophore, allowing for fast proton transfer from the excited state of the neutral chromophore to the final acceptor. Under these conditions, green emission from the deprotonated chromophore is observed. With increasing ammonium transport activity, the wellestablished ESPT pathway is partially inhibited, so that the proton cannot be efficiently transferred. Emission from the protonated intermediate is observed as blue light ( Figure 5).
The backbone rearrangements must be extremely fast to disrupt and re-establish the ESPT network. Still, the motion is not dramatic enough to completely prevent ESPT, explaining why green emission is still observed as the dominant peak during conditions of high ammonium transport. Therefore, deAmTracs only show modest ratio changes compared to other published dual-emission sensors, possibly due to only small spatial rearrangements or averaging of the ensembles during the ammonium binding/transport process. Future efforts will address and explore alternative approaches for the design of improved ratiometric sensors for ammonium transporters.

DNA constructs and mutagenesis
All cpEGFP-based sensor plasmid constructs described are based on pDRF'-AmTrac-LE and -GS; described previously (De Michele et al., 2013).

Random mutagenesis
The deAmTrac sensors were generated by substituting the LE-linker of AmTrac-LE with random amino acids to test all the possible linker sequences and effects on dual-emission patterns. The cpEGFP fragment was amplified with forward primer AmXX FW containing the variable sequence encoding the linker, and the reverse primer AmFN RV, encoding the C-terminal FN-linker (Table 1). DNA-fragments were gel-purified (Machery-Nagel) and yeast was co-transformed with linearized pDRF'-AtAMT1;3 (cleaved after amino acid 233). The cpEGFP-fragment with varying linkers was inserted after amino acid 233 of AtAMT1;3 by homologous recombination (De Michele et al., 2013).
For the complementation assay, liquid cultures were diluted 10 −1 , 10 −2 , 10 −3 , 10 −4 and 10 −5 in water and 5 μl of each dilution was spotted on solid YNB medium buffered with 50 mM MES/Tris, pH 5.2, supplemented with 3% glucose and either NH 4 Cl, or 1 mM arginine as the sole nitrogen source. After 3 days of incubation at 30˚C, cell growth was documented by flatbed scanning the plate at 300 dpi in grayscale mode.
For fluorescence measurements, liquid yeast cultures were washed twice in 50 mM MES pH 6.0, and resuspended to OD 600nm ∼0.5 in 50 mM MES pH 6.0, supplemented with 5% glycerol to delay cell sedimentation (Ast et al., 2015).