Systematic Comparison of Beetle Luciferase-Luciferin Pairs as Sources of Near-Infrared Light for In Vitro and In Vivo Applications

Luciferases catalyze light-emitting reactions that produce a rainbow of colors from their substrates (luciferins), molecular oxygen, and often additional cofactors. These bioluminescence (BL) systems have afforded an incredible variety of basic research and medical applications. Driven by the importance of BL-based non-invasive animal imaging (BLI) applications, especially in support of cancer research, new BL systems have been developed by engineering beetle luciferase (Luc) variants and synthetic substrate combinations to produce red to near-infrared (nIR) light to improve imaging sensitivity and resolution. To stimulate the application of BLI research and advance the development of improved reagents for BLI, we undertook a systematic comparison of the spectroscopic and BL properties of seven beetle Lucs with LH2 and nine substrates, which included two new quinoline ring-containing analogs. The results of these experiments with purified Luc enzymes in vitro and in live HEK293T cells transfected with luc genes have enabled us to identify Luc/analog combinations with improved properties compared to those previously reported and to provide live cell BL data that may be relevant to in vivo imaging applications. Additionally, we found strong candidate enzyme/substrate pairs for in vitro biomarker applications requiring nIR sources with minimal visible light components. Notably, one of our new substrates paired with a previously developed Luc variant was demonstrated to be an excellent in vitro source of nIR and a potentially useful BL system for improved resolution in BLI.


Introduction
Bioluminescence (BL), the emission of visible light by living organisms, is widely distributed in nature and has piqued the curiosity of humans from ancient times to the present. One commonly observed example of BL is that of the beetles including the widely studied North American firefly Photinus pyralis. As is the case for all characterized bioluminescent beetles, P. pyralis produces light from an enzyme (luciferase, Luc)-catalyzed reaction of a substrate (luciferin, LH 2 ) requiring Mg-ATP and molecular oxygen. Luccatalyzed light production (Scheme 1), which very likely proceeds through a single-electron transfer (SET) process [1,2], is an interesting example of the "substrate-assisted oxygenases" concept [3], wherein an organic substrate drives the oxidation process in the absence of cofactors. While P. pyralis Luc (often referred to as PpyWT and Fluc) normally produces yellow-green light (λ max~5 60 nm) with LH 2 , Luc mutants and various wild-type enzymes can produce emissions with maxima ranging from~535 nm to~630 nm [4][5][6][7]. Longer Paired primarily with the wild-type Luc from P. pyralis and several variants, a wide variety of LH2 analogs have been reported that extend the range of BL emission maxima from ~520 nm to ~750 nm [9][10][11][12][13][14]. Longer wavelength photon production has been accomplished without the assistance of energy transfer processes by, in most instances, extending the π conjugation of the natural substrate in several distinct structural designs [10][11][12][13][14], including the incorporation of a naphthalene ring (NH2-NpLH2 and OH-NpLH2) [14] ( Figure 1). For substrate analogs AkaLumine-HCl (Aka) [15], infraluciferin (iLH2) [16,17], 4′-BrLuc [18,19], NH2-NpLH2 [14] and OH-NpLH2 [14], the initial BL properties determined with Fluc, a mammalian codon optimized version of P. pyralis Luc (Luc2), or click beetle red Luc (CBR) were optimized for noninvasive in vivo bioluminescence imaging (BLI) applications by mutagenesis strategies including directed evolution.
The advantages of expanding BLI applications with synthetic substrate analogs and optimized orthogonal Luc/luciferin analog pairs have been discussed in highly informative recent reviews [20,21]. BLI [22], especially with methods using beetle enzymes [15], is a particularly exciting and important application of BL offering extremely low background due to the absence of inherent light emission in mammals [23][24][25]. This advantage, however, is tempered somewhat by generally weak light emission, although highly sensitive cooled CCD-based detectors mitigate this shortcoming. While the Luc2/LH2 system with λmax ~605 nm at 37 °C has been widely used in BLI, it is limited due to absorption by hemoglobin, lipids, water and other cell components [26]. An active area of investigation to improve BLI resolution, especially for deep tissue imaging, is focused on the development of Luc/substrate pairs that emit in the near-infrared (nIR) "bio-optical" window between 650 nm and 900 nm [14][15][16][17]21,27].
A current major direction of our lab is to make highly efficient biological sources of nIR light with λmax values > 700 nm (and minimal emission in the visible) for in vitro use in biosensors and biomarkers detectable with night vision technology. Previously, we made good strides towards this goal by developing an intramolecular Bioluminescence Resonance Energy Transfer (BRET) system [28] that consisted of a highly engineered PpyWT variant covalently labeled with nIR fluorescent dyes. In one example, we constructed a system that emitted with λmax = 783 nm; however, ~20% of the total light emitted was at wavelengths below 700 nm due to incomplete BRET. To avoid the required, but undesirable and limiting, step of chemically modifying enzymes, we recently have focused on engineering Luc variant/luciferin analog pairs to accomplish our goal. Previously, we paired Luc2 variants with LH2 and a substrate analog to develop an economical and convenient dual color BL reporter gene assay to simultaneously monitor two gene expression events [29]. Paired primarily with the wild-type Luc from P. pyralis and several variants, a wide variety of LH 2 analogs have been reported that extend the range of BL emission maxima from 520 nm to~750 nm [9][10][11][12][13][14]. Longer wavelength photon production has been accomplished without the assistance of energy transfer processes by, in most instances, extending the π conjugation of the natural substrate in several distinct structural designs [10][11][12][13][14], including the incorporation of a naphthalene ring (NH 2 -NpLH2 and OH-NpLH2) [14] (Figure 1). For substrate analogs AkaLumine-HCl (Aka) [15], infraluciferin (iLH 2 ) [16,17], 4 -BrLuc [18,19], NH 2 -NpLH2 [14] and OH-NpLH2 [14], the initial BL properties determined with Fluc, a mammalian codon optimized version of P. pyralis Luc (Luc2), or click beetle red Luc (CBR) were optimized for noninvasive in vivo bioluminescence imaging (BLI) applications by mutagenesis strategies including directed evolution.
The advantages of expanding BLI applications with synthetic substrate analogs and optimized orthogonal Luc/luciferin analog pairs have been discussed in highly informative recent reviews [20,21]. BLI [22], especially with methods using beetle enzymes [15], is a particularly exciting and important application of BL offering extremely low background due to the absence of inherent light emission in mammals [23][24][25]. This advantage, however, is tempered somewhat by generally weak light emission, although highly sensitive cooled CCD-based detectors mitigate this shortcoming. While the Luc2/LH 2 system with λ max 605 nm at 37 • C has been widely used in BLI, it is limited due to absorption by hemoglobin, lipids, water and other cell components [26]. An active area of investigation to improve BLI resolution, especially for deep tissue imaging, is focused on the development of Luc/substrate pairs that emit in the near-infrared (nIR) "bio-optical" window between 650 nm and 900 nm [14][15][16][17]21,27].
A current major direction of our lab is to make highly efficient biological sources of nIR light with λ max values > 700 nm (and minimal emission in the visible) for in vitro use in biosensors and biomarkers detectable with night vision technology. Previously, we made good strides towards this goal by developing an intramolecular Bioluminescence Resonance Energy Transfer (BRET) system [28] that consisted of a highly engineered PpyWT variant covalently labeled with nIR fluorescent dyes. In one example, we constructed a system that emitted with λ max = 783 nm; however,~20% of the total light emitted was at wavelengths below 700 nm due to incomplete BRET. To avoid the required, but undesirable and limiting, step of chemically modifying enzymes, we recently have focused on engineering Luc variant/luciferin analog pairs to accomplish our goal. Previously, we paired Luc2 variants with LH 2 and a substrate analog to develop an economical and convenient dual color BL reporter gene assay to simultaneously monitor two gene expression events [29].  We report here the systematic evaluation of several purified luciferase proteins as potential in vitro sources of nIR light including: Luc2; RedFluc (Targeting Systems mammalian codon optimized red-shifted L. italica Luc); Akaluc (a Luc2 variant containing 28 mutations) [15]; Fluc_red (a mammalian codon optimized P. pyralis Luc variant containing 16 mutations) [17]; Mut51 (the Luc2 variant F243M/S347G) [30]; CBR2 (the mouse codon optimized CBR R334S/G351R variant) [14]; and PLR3 [31] (the G246A/F250H variant of the mammalian codon optimized PLR1) [29]. The seven enzymes were tested with LH2 and nine selected substrate analogs, including two novel quinoline ring-containing compounds that we are reporting here (Figure 1). With the exception of racemic iLH2, all the substrates in this study contain a 4-carboxylic acid substituted thiazoline ring in the Dconfiguration.
Recognizing that enzyme/substrate pairs that produce nIR light are of great importance to the continued development and improvement of in vivo BLI methods, we focused on combinations that, with the exception of the quinoline analogs NH2-QLH2 and OH-QLH2, had been successfully employed in BLI studies [14][15][16][17][18][19]30,[32][33][34][35][36][37][38][39]. PLR3 was matched with the quinoline-containing analogs based on the complete in vitro and live cell testing results with the seven Lucs (Table S1). While we intended to include in this study as many representative luciferase/substrate pairs as possible, a report by Viviani et We report here the systematic evaluation of several purified luciferase proteins as potential in vitro sources of nIR light including: Luc2; RedFluc (Targeting Systems mammalian codon optimized red-shifted L. italica Luc); Akaluc (a Luc2 variant containing 28 mutations) [15]; Fluc_red (a mammalian codon optimized P. pyralis Luc variant containing 16 mutations) [17]; Mut51 (the Luc2 variant F243M/S347G) [30]; CBR2 (the mouse codon optimized CBR R334S/G351R variant) [14]; and PLR3 [31] (the G246A/F250H variant of the mammalian codon optimized PLR1) [29]. The seven enzymes were tested with LH 2 and nine selected substrate analogs, including two novel quinoline ring-containing compounds that we are reporting here (Figure 1). With the exception of racemic iLH 2 , all the substrates in this study contain a 4-carboxylic acid substituted thiazoline ring in the D-configuration.
Recognizing that enzyme/substrate pairs that produce nIR light are of great importance to the continued development and improvement of in vivo BLI methods, we focused on combinations that, with the exception of the quinoline analogs NH 2 -QLH 2 and OH-QLH 2 , had been successfully employed in BLI studies [14][15][16][17][18][19]30,[32][33][34][35][36][37][38][39]. PLR3 was matched with the quinoline-containing analogs based on the complete in vitro and live cell testing results with the seven Lucs (Table S1). While we intended to include in this study as many representative luciferase/substrate pairs as possible, a report by Viviani et al. [40] that the purified P. hirtus railroad worm R215K variant paired with 6 -(1-pyrrolidinyl) luciferin [10] produced bright 650 nm BL appeared after this investigation was completed. Moreover, additional interesting LH 2 analogs have been synthesized [27,41,42] and several important additional comparative studies aimed at choosing optimal Luc/substrate pairs for in vivo BLI have appeared recently. Mezzanotte et al. compared [39] four Lucs and four substrates for this purpose, and a review by Saito-Moriya and coworkers emphasized [43] selected luciferin analogs, several of which are not included in our study. Our results have enabled us to identify strong candidate enzyme/substrate pairs for in vitro biomarker applications, to find Luc/analog combinations with improved properties compared to those previously reported, and to provide live cell BL data that may be relevant to in vivo imaging applications.

Luciferin Analog Synthesis
Based on prior results with simple quinoline-and naphthalene-substituted LH 2 analogs [44,45], we designed the amino-and hydroxy-substituted fused quinoline ring substrates NH 2 -QLH 2 and OH-QLH 2 , intending to similarly extend the long wavelength emission and improve the brightness of the isosteric fused naphthalene ring substrates NH 2 -NpLH2 and OH-NpLH2 [14] (Figure 1). The synthetic pathways (Scheme 2) that we developed for the preparation of the novel analogs were modeled after the approach to the corresponding naphthyl substrates [14]. The key Appel closures were accomplished using Prescher's methodology [46] that we adapted to microwave conditions, which for NH 2 -QLH 2 resulted in concomitant BOC group deprotection. The NH 2 -QLH 2 compound (K + salt) was prepared in six steps in 8% overall yield with 98.6% enantiomeric excess (ee) ( Figure S1), while the OH-QLH 2 (K + salt) was obtained in 13% overall yield in eight steps with 97.2 % ee ( Figure S2). After trials with several methods for the final condensation reaction with D-cysteine to produce the analogs, we found Miller's procedure [47] provided highly pure products with excellent % ee and a more readily scalable approach not requiring HPLC purification. Moreover, our yields of the final condensation steps and conversion of the NH 2 -NpLH2 and OH-NpLH2 products into the corresponding potassium salts were accomplished in~3.5-fold higher yields and with equal or greater % ee compared to the published values [14]. The syntheses of NH 2 -NpLH2, OH-NpLH2, NH 2 -QLH 2 , and OH-QLH 2 are described in the Supplementary Materials. al. [40] that the purified P. hirtus railroad worm R215K variant paired with 6′-(1-pyrrolidinyl) luciferin [10] produced bright 650 nm BL appeared after this investigation was completed. Moreover, additional interesting LH2 analogs have been synthesized [27,41,42] and several important additional comparative studies aimed at choosing optimal Luc/substrate pairs for in vivo BLI have appeared recently. Mezzanotte et al. compared [39] four Lucs and four substrates for this purpose, and a review by Saito-Moriya and coworkers emphasized [43] selected luciferin analogs, several of which are not included in our study.
Our results have enabled us to identify strong candidate enzyme/substrate pairs for in vitro biomarker applications, to find Luc/analog combinations with improved properties compared to those previously reported, and to provide live cell BL data that may be relevant to in vivo imaging applications.

Luciferin Analog Synthesis
Based on prior results with simple quinoline-and naphthalene-substituted LH2 analogs [44,45], we designed the amino-and hydroxy-substituted fused quinoline ring substrates NH2-QLH2 and OH-QLH2, intending to similarly extend the long wavelength emission and improve the brightness of the isosteric fused naphthalene ring substrates NH2-NpLH2 and OH-NpLH2 [14] ( Figure 1). The synthetic pathways (Scheme 2) that we developed for the preparation of the novel analogs were modeled after the approach to the corresponding naphthyl substrates [14]. The key Appel closures were accomplished using Prescher's methodology [46] that we adapted to microwave conditions, which for NH2-QLH2 resulted in concomitant BOC group deprotection. The NH2-QLH2 compound (K + salt) was prepared in six steps in 8% overall yield with 98.6% enantiomeric excess (ee) ( Figure S1), while the OH-QLH2 (K + salt) was obtained in 13% overall yield in eight steps with 97.2 % ee ( Figure S2). After trials with several methods for the final condensation reaction with D-cysteine to produce the analogs, we found Miller's procedure [47] provided highly pure products with excellent % ee and a more readily scalable approach not requiring HPLC purification. Moreover, our yields of the final condensation steps and conversion of the NH2-NpLH2 and OH-NpLH2 products into the corresponding potassium salts were accomplished in ~3.5-fold higher yields and with equal or greater % ee compared to the published values [14]. The syntheses of NH2-NpLH2, OH-NpLH2, NH2-QLH2, and OH-QLH2 are described in the Supplementary Materials.

Spectral Characterization of Substrates
The UV-visible ( Figure S7) and fluorescence (FL) ( Figure S8) spectra, and BL emission maxima (Table S1) of all Luc/substrates were measured. The BL emission spectra of representative Luc/substrate pairs that were obtained with purified Lucs at 37 • C ( Figure 2) illustrate the wide range of "red" BL emission that is achievable. All substrates contain long wavelength UV peaks between 330 nm and 386 nm that somewhat correlate with the corresponding BL emission maxima. In contrast, the relationship of FL maxima to the highest BL emission maxima achieved for each substrate ( Figure S9) produced an excellent linear fit (R 2 = 0.99). While the corresponding oxyluciferins (Scheme 1) are the actual BL light emitters, the FL emission maxima of the substrates seem to have reasonably good predictive value for achievable long wavelength BL. Given that all of the substrates share the carboxy-substituted thiazoline ring and are likely to proceed through the same chemistry to form the oxyluciferin emitters, the substrate FL properties are apparently relevant. The FL quantum yields of the oxyluciferins are a key factor in determining the maximum BL efficiency, i.e., the conversion of reacted molecules of substrate into photons. While we were pleased that the FL quantum yields of the quinoline-containing substrates were~2-fold greater than the corresponding naphthalene ones, the expected longer wavelength emission maxima were not achieved. Instead, both quinoline ring-containing substrates had~25 nm shorter FL wavelength maxima ( Figure S8). It is noteworthy that the six substrates with extended conjugation had very low FL quantum yields (0.02-0.16); whereas LH 2 and the three analogs with equivalent conjugation had much higher values, ranging from 0.67 to 0.83 ( Figure S8).

Spectral Characterization of Substrates
The UV-visible ( Figure S7) and fluorescence (FL) ( Figure S8) spectra, and BL emission maxima (Table S1) of all Luc/substrates were measured. The BL emission spectra of representative Luc/substrate pairs that were obtained with purified Lucs at 37 °C (Figure 2) illustrate the wide range of "red" BL emission that is achievable. All substrates contain long wavelength UV peaks between 330 nm and 386 nm that somewhat correlate with the corresponding BL emission maxima. In contrast, the relationship of FL maxima to the highest BL emission maxima achieved for each substrate ( Figure S9) produced an excellent linear fit (R 2 = 0.99). While the corresponding oxyluciferins (Scheme 1) are the actual BL light emitters, the FL emission maxima of the substrates seem to have reasonably good predictive value for achievable long wavelength BL. Given that all of the substrates share the carboxy-substituted thiazoline ring and are likely to proceed through the same chemistry to form the oxyluciferin emitters, the substrate FL properties are apparently relevant. The FL quantum yields of the oxyluciferins are a key factor in determining the maximum BL efficiency, i.e., the conversion of reacted molecules of substrate into photons. While we were pleased that the FL quantum yields of the quinoline-containing substrates were ~2fold greater than the corresponding naphthalene ones, the expected longer wavelength emission maxima were not achieved. Instead, both quinoline ring-containing substrates had ~25 nm shorter FL wavelength maxima ( Figure S8). It is noteworthy that the six substrates with extended conjugation had very low FL quantum yields (0.02-0.16); whereas LH2 and the three analogs with equivalent conjugation had much higher values, ranging from 0.67 to 0.83 ( Figure S8).

Figure 2.
Normalized bioluminescence (BL) emission spectra of purified Luc protein/substrate pairs. BL was initiated and emission spectra were recorded at 37 °C as described in Materials and Methods, with the exception that 25 μg of protein was used for the RedFluc/NH2-NpLH2 measurements to enhance signal intensity.

Systematic Study
A major objective of this study is to provide comparable data on the BL properties of our NH2-QLH2 and OH-QLH2 analogs with PLR3 and a series of Luc/substrate pairs that have been successfully applied in various in vivo BLI applications. While there is great value in applying standardized assay conditions in a systematic study, we recognize that even small changes in protocols could make it problematic to rigorously compare the results presented here to published data from other labs. Normalized bioluminescence (BL) emission spectra of purified Luc protein/substrate pairs. BL was initiated and emission spectra were recorded at 37 • C as described in Materials and Methods, with the exception that 25 µg of protein was used for the RedFluc/NH 2 -NpLH2 measurements to enhance signal intensity.

Systematic Study
A major objective of this study is to provide comparable data on the BL properties of our NH 2 -QLH 2 and OH-QLH 2 analogs with PLR3 and a series of Luc/substrate pairs that have been successfully applied in various in vivo BLI applications. While there is great value in applying standardized assay conditions in a systematic study, we recognize that even small changes in protocols could make it problematic to rigorously compare the results presented here to published data from other labs.

BL Properties of Luc2 with LH 2 and Substrate Analogs In Vitro
The results of BL testing of ten substrates with the widely used Luc2 enzyme are presented in Table 1. The in vitro measurements were made at 23 • C and BL emission spectra also were acquired at 37 • C to better estimate the emission profiles under live cell conditions. With the exception of LH 2 , whose emission shifted to~600 nm, BL peaks remained within ± 5 nm at the higher temperature.  (133) 0.03 ± 0.003 1.8 ± 0.6 1 Specific activities were obtained from assays at pH 7.4 and 23 • C by monitoring and integrating signal intensity for 2 min using purified enzyme (2.5 µg), 50 µM substrate, 1 mM ATP, and 3 mM MgSO 4 . Data are expressed as the mean ± standard deviation and are reported relative to the Luc2/LH 2 value, defined as 100 and were corrected for the spectral response of the CCD detector. 2 Data determined from BL emission spectra measured in in vitro assays were obtained as described in Materials and Methods. Bandwidths at full width at half-maximum values are given in parentheses. 3 Average radiance [p/s/cm 2 /sr] of BL measured from luc2 transfected HEK293T cells with the open filter setting of an IVIS Spectrum III instrument. BL is measured at the highest level attained within 5 min of 0.5 mM substrate addition. Data are expressed as the mean ± standard deviation and are reported relative to the Luc2/LH 2 value, as described in Materials and Methods. b.d.; below detection having a relative activity of <0.03. 4 V 0 values used to determine the constants were obtained from peak-height measurements as described in detail in the Supplementary Materials.
Not surprisingly, the natural substrate LH 2 is~10 to~3000 times brighter with Luc2 than all of the other substrates tested with this enzyme in vitro. The range of BL emission maxima for the analogs was an impressive 603 nm to 720 nm. For substrates with at least~2% of the intensity of the Luc2/LH 2 pair, only Aka produced light in the nIR optical window (λ max = 677 nm). With the exception of 4 -BrLuc, all substrates had lower K m values than LH 2 , and for the longest wavelength BL emitters they were 0.13 to 3.3 µM. While Luc2 can accommodate a wide array of substrate structures and can produce extraordinarily red-shifted emission (+158 nm compared to LH 2 ), a great deal of brightness is lost with the longest wavelength emitters.

BL Properties of LH 2 with Lucs In Vitro
The results of the evaluation of the BL properties of the seven Lucs with the natural LH 2 substrate are included in Table 2. The BL spectral data confirm that long wavelength emission beyond~625 nm is highly unlikely to be achievable with the natural substrate.

Enhanced In Vitro nIR Sources for Biosensor and Biomarker Applications
Focusing on our goal to make highly efficient biological sources of nIR light with λ max values > 690 nm and with minimal emission in the visible for use as biosensors and biomarkers, we measured the % BL emission > 690 nm of the Luc/substrates and adjusted the data for their relative in vitro specific activities. The top three enzyme/substrate pairs (Figure 3c) had BL λ max , % visible, and % > 690 nm emission values of: PLR3/OH-QLH 2 (718 nm, 2%, 76%); CBR2/NH 2 -NpLH2 (721 nm, 3%, 73%); and Red_Fluc/NH 2 -NpLH2 (694 nm, 8%, 62%). PLR3/OH-QLH 2 and CBR2/NH 2 -NpLH2 are excellent sources of nIR and their greater BL maxima are clearly the major contributing factor. While the intensities of these sources are low, for in vitro applications it should be possible to significantly improve the sensitivity mainly by increasing the concentrations of the enzymes and maintaining sufficient excess of the substrates.

Enhanced In Vitro nIR Sources for Biosensor and Biomarker Applications
Focusing on our goal to make highly efficient biological sources of nIR light with λmax values > 690 nm and with minimal emission in the visible for use as biosensors and biomarkers, we measured the % BL emission > 690 nm of the Luc/substrates and adjusted the data for their relative in vitro specific activities. The top three enzyme/substrate pairs (Figure 3c) had BL λmax, % visible, and % > 690 nm emission values of: PLR3/OH-QLH2 (718 nm, 2%, 76%); CBR2/NH2-NpLH2 (721 nm, 3%, 73%); and Red_Fluc/NH2-NpLH2 (694 nm, 8%, 62%). PLR3/OH-QLH2 and CBR2/NH2-NpLH2 are excellent sources of nIR and their greater BL maxima are clearly the major contributing factor. While the intensities of these sources are low, for in vitro applications it should be possible to significantly improve the sensitivity mainly by increasing the concentrations of the enzymes and maintaining sufficient excess of the substrates.

BL Properties of Luc2 with LH 2 and Substrate Analogs in Live Cells
We obtained live HEK293T cell BL activity data for all of the Lucs (Tables 1-3 and  Table S1), and normalized them to an internal transfection efficiency control (Nluc activity). The data are reported relative to the Luc2/LH 2 pair with an underlying assumption that the expression levels of the other Lucs are very similar to that of Luc2. In the live cell experiments, Luc2 was expressed and assayed at 37 • C. The live cell BL emission in all experiments was consistently greatest when measured through the bandpass filter that corresponded to the 37 • C BL emission maximum measured with purified proteins (Section 3 and Table S1). Compared to the in vitro Luc protein results (Table 1) in which saturating concentrations of substrates were used, there are additional determinants of the intensity of the BL signals in live cells including the permeability of substrates across the cell membrane and the apparent affinity of the Lucs for the substrates (K m in the cell environment). By normalizing the specific activity results to the purified Luc2 protein/LH 2 pair and obtaining the live cell activities of the substrates with Luc2 (essentially eliminating protein expression as a variable), we could reasonably compare the cellular properties of the substrates ( Table 1). The live cell specific activities followed the same general trend as the in vitro values; however, the~4-to~5-fold greater live cell values with CycLuc1, Aka, and NH 2 -LH 2 were impressive. Interestingly, the K m values of these substrates werẽ 3-25-fold lower than that of LH 2 . For these substrates, the K m values may be correlated to cell permeability and, with the exception of NH 2 -LH 2 , the substrates have higher LogP values (Table S2) that are consistent with cell membrane permeability being an important determinant of specific activity. Notably, analogs containing amino group substituents had relative activities 2.2-5.7-fold higher in live cells. In particular, the membrane permeability of NH 2 -NpLH2 and NH 2 -QLH 2 was greater than that of LH 2 and OH-QLH 2 . (Figure S10).

BL Properties of LH 2 with Lucs in Live Cells
A comparison of the relative in vitro and live cell specific activities revealed that greater (than Luc2) values were obtained in HEK293T cells for PLR3 (1.5-fold), CBR2 (1.6-fold) and Fluc_red (2.7-fold). Interestingly, these results suggest that enhanced sensitivity over the widely used Luc2/LH 2 pair may be achieved in BLI with PLR3, as they have with CBR2 [14,39] and Fluc_red [17]. It is likely that the favorable properties of these enzymes reflect enzyme stability at 37 • C, which is greater than that of Luc2.

BL Properties of Optimized and Novel Luc/Substrate Combinations in Live Cells
With the exception of our recently developed NH 2 -QLH 2 and OH-QLH 2 , the substrates in this study have been widely employed in BLI applications. We determined the BL properties of these optimized enzyme/substrate pairs along with unreported combinations that, in some cases, exceeded them (Table 3). In PC3M cells expressing P. pyralis Luc, a ratio of~5.5 was reported for LH 2 /NH 2 -LH 2 [32]. Our results with Luc2 in live cells were similar (LH 2 /NH 2 -LH 2 : ratio = 2.8), while CBR2 produced the greatest signal strength with NH 2 -LH 2 (Table 3); a result also observed with the F247L variant of Fluc [48]. The commercial availability of this substrate makes it an interesting possibility for BLI applications. Additionally, NH 2 -LH 2 has been generated intracellularly from stable precursors in a BLI study with transfected MDA-MB-231 tumor cells [35].
We benchmarked our evaluation of CycLuc1 paired with Luc2 (Table 1). While pairing this cyclic amino group-containing substrate with PLR3 produced a 1.6-fold increase in in vitro activity, a more significant 5.5-fold enhancement was realized in HEK293T cells (Table 3). In other mammalian cells, an 18-fold improvement in BL has been reported for a P. pyralis Luc S347A mutant [48]. Interestingly, CycLuc1 has excellent bioavailability in mice in vivo, based on its outstanding performance compared to LH 2 in BLI brain and deep tissue imaging studies [49]. The prospects of further improving BLI with this substrate by pairing it with PLR3 are exciting. Similarly, RedFluc displayed greater (1.8-fold) activity in live cells using Cycluc1 than with LH 2 , albeit with modestly blue-shifted emission (Table S1). Among the longest wavelength BL systems, the optimized Fluc_red/iLH 2 pair produced modest BL intensity at 707 nm (Table 3), yet has provided excellent sensitivity in a dual BLI format when paired with a green light emitting Fluc mutant and LH 2 [17]. Even greater sensitivity and longer wavelength nIR BLI may be possible with CBR2/iLH 2 as it increased live cell BL 2.3-fold, with maximum emission increased by 20 nm (Table 3) compared to Fluc_red/iLH2.
The Akaluc enzyme contains 28 amino acid changes introduced through directed evolution [15]. Paired with Aka in HEK293T cells, we found an Akaluc/Luc2 ratio of~15 in BL activity. While the activity of this optimized enzyme/substrate pair is poor in vitro, its live cell activity with 643 nm emission is the highest among all the long wavelength systems we measured ( Table 3). The remarkable BLI imaging of single cells that has been realized with the Akaluc/Aka pair is a reminder of the limitations of using only in vitro assays under ideal conditions to predict strong BLI candidates.
The nIR-emitting combination CBR2/NH 2 -NpLH2 has been successfully employed in BLI applications [14,39] despite its unimpressive activity using purified enzyme. Only PLR3 and Fluc_red produced good BL with this substrate in live cells, but at the expense of 30 nm blue shifts (Table S1). The activity of the OH-NpLH2 substrate paired with CBR2 was 1.4-fold lower than that of CBR2/NH 2 -NpLH2 in live cells (Table 3). However, the emission maximum obtained with OH-NpLH2 was an impressive 750 nm. NH 2 -QLH 2 and OH-QLH 2 paired optimally and very effectively with PLR3. Interestingly, as with the isosteric naphthyl-containing substrates, the hydroxyl substituent produced longer wavelength nIR (λ max = 716 nm) BLI; however, with OH-QLH 2 , the BL intensity was also greater. Moreover, the PLR3/OH-QLH 2 pair produced live cell BL~1.3-fold greater than the CBR2/NH 2 -NpLH2 pair with similar nIR emission (λ max = 718 nm) ( Table 3). These encouraging results ( Figure S11) will be pursued in future BLI studies. With CBR2, wild-type CBR was improved by the introduction of the amino acid changes R337S and G354R (Luc2 numbering) [14]. Introduction of these changes into PLR3 did not improve its BL properties. Based on sequence comparisons and the testing of OH-QLH 2 , with several previously published Lucs [5,31], it appears that the mutations G246A and F250H in PLR3 are major contributors to enabling the considerably larger quinoline ring-containing substrates to be accommodated productively at the enzyme's active site. We had shown previously that these residues which occupy the first turn of an active site defining helix [50], can form side chain to main chain H-bonds [50]. Perhaps this type of interaction allows local conformation changes that enables productive binding of the larger aromatic ring systems.
Mut51 was discovered in the Prescher lab using a novel parallel screening strategy [30]. In BLI of mice expressing the Mut51 variant in DB7 cells, 4 -BrLuc and LH 2 were shown to be an excellent orthogonal pair with 4 -BrLuc providing a sufficiently strong signal and LH 2 an essentially undetectable one [30]. We found the specific activity ratio (4 -BrLuc/LH 2 ) with purified Mut51 was~1 and in HEK293T cells the ratio was~2 (Tables 2 and 3). Unfortunately, our methods were unable to predict the selectivity of Mut51 observed in the BLI study. While differences in light measuring methodology and/or BL in the different cell lines is important, it is also possible that the bioavailability of 4 -BrLuc in mice is an important determinant. Notably, the LogP and RP-HPLC retention times (Table S2) strongly suggest that the halogen-containing substrate is much more hydrophobic than LH 2 . In addition, while pairing PLR3 with 4 -BrLuc may not produce a useful orthogonal pair with LH 2 , PLR3 produces 23-and 32-fold greater BL in vitro and in HEK293T cells, respectively, than does Mut51 (Table 3).

Enhanced nIR Sources for BLI Applications
We addressed the question of improving the resolution of the widely employed Luc2/LH 2 combination in BLI that would require increased photon output and/or greater emission in the nIR optical window. Using experiments in live HEK293T cells as a model, we measured the percentage of visible and nIR emission of the Luc/substrate pairs and adjusted the data for their relative specific activities (Figure 3a). Seven of the enzyme/substrate combinations had relative nIR window output that exceeded that of Luc2/LH 2 . Excellent BLI results have been reported for Akaluc/Aka [15], CBR2/LH 2 [14,39] and Fluc_red/LH 2 [17]. Our model data show that the 4.4-fold Fluc_red/LH 2 improvement is almost entirely due to its higher specific activity; whereas Akaluc/Aka (5.0-fold) and CBR2/LH 2 (2.9-fold) have greater specific activity and considerably longer wavelength emission. The live cell model data indicated that the previously unreported combinations Fluc_red/Aka (4.6-fold), CBR2/NH 2 -LH 2 (1.6-fold), PLR3/LH 2 (2.3-fold), and PLR3/CycLuc1 (4.5-fold) are potential candidates to achieve high resolution BLI. Extraordinarily high percentages of nIR optical window emission were observed for CBR2/NH 2 -NpLH2 (97%) and PLR3/OH-QLH 2 (98%) (Figure 3b). The former pair has successfully provided higher resolution in deeper tissue BLI images [14,39] that may also be realized with PLR3/OH-QLH 2 .

Materials
The following materials were obtained from the sources indicated: Mg-ATP (bacterial source) and ATP (disodium salt hydrate) from Sigma-Aldrich ( Akaluc in the pCDH-EF1-MCS-T2A-copGFP vector was obtained from Gene Dynamics, LLC (Portland, OR, USA), and PLR3 in the pF4Ag vector was described previously [31].
The following Luc proteins were expressed in Escherichia coli strain BL21(DE3) pLysS from the pGex-6P-2 vector: Luc2, Akaluc, CBR2, RedFluc, and PLR3. The Luc2, Akaluc and CBR2 genes were subcloned from their respective vectors. The RedFluc gene previously described in US Patent number 7,807,429 B2; LitS-S-11/F467R was used directly and the PLR3 gene was constructed by inserting the G246A and F250H mutations into the PLR1 gene [29]. The Fluc_red gene was constructed by introducing the S284T, H354R, and A357Y mutations into the x11a Fluc gene in the pET16b His-tag vector (Amit Jathoul, Cardiff University, Cardiff, Wales, UK), and the Mut51 gene in the pET28a His-tag vector was used as received.
The HEK293T cell studies were performed using mammalian codon optimized luc genes under the CMV promoter in the pNL3.1 [Nluc/minP] Vector. The genes encoding Luc2, CBR2, PLR3, and Mut51 were subcloned from the pF4Ag vector by ligation into the multiple cloning region of the pNL3.1 vector using the XhoI and EcoRV restriction sites. The Fluc_red, Akaluc and RedFluc genes were subcloned by replacing the Luc2 gene in the CMV-Luc2-pNL3.1 vector using the AsiSI and EcoRV restriction sites.
The structures of the substrates used in this study are shown in Figure 1.

General Methods
All luc gene sequences were verified by DNA sequencing at the W. M. Keck Biotechnology Laboratory (Yale University, New Haven, CT, USA). The Luc2, Akaluc, RedFluc, CBR2 and PLR3 genes were expressed in the pGEX-6P-2 expression vector and the purified enzymes contain the N-terminal peptide extension GlyProLeuGlySer-. The Mut51 and Fluc_red genes were expressed in His-tag expression vectors pET28a and pET16b, respectively. Concentrations of purified proteins were determined using a NanoDrop™ Lite Spectrophotometer. Detailed protocols for protein expression and purification and K m measurements are found in the Supplementary Methods.

Bioluminescence Emission Spectra
BL was initiated by mixing equal volumes (0.25 mL) of a solution of 50 mM Tricine pH 7.4 containing 2 mM ATP and 6 mM MgSO 4 with a solution of assay buffer containing 12.5 µg of enzyme and 0.1 mM of the indicated analog in a quartz cuvette. All solutions were pre-warmed to either 23 • C or 37 • C. The final concentrations of the mixture (0.5 mL) in 50 mM Tricine pH 7.4 were 0.4 µM enzyme, 50 µM of the indicated analog, 1 mM ATP, and 3 mM MgSO 4 . Emission spectra were acquired at 23 • C and 37 • C after a 1 min delay with a Horiba Jobin-Yvon iHR 320 imaging spectrometer equipped with a liquid N 2 cooled CCD detector. Data were collected over the wavelength range 450-925 nm, with the excitation source turned off and the emission slit width set to 10 nm, and were corrected for the spectral response of the detector using a correction curve provided by the manufacturer. The pH values were confirmed before and after spectra were obtained (Figure 2).

In Vitro Specific Activities
All assays were performed in triplicate in white 96-well microtiter plates containing 2.5 µg of purified enzyme and 50 µL of 0.1 mM analog in 50 mM Tricine pH 7.4. BL was initiated by the automated injection of 50 µL of 50 mM Tricine pH 7.4 containing 2 mM ATP, and 6 mM MgSO 4 . Signals were monitored over 2 min using a Synergy™ 2 microplate luminometer (BioTek, Winooski, VT, USA). Data were integrated and corrected for the spectral response of the Hamamatsu R928 PMT detector.

In Vitro Bioluminescence Wavelength Emission Distribution
Assays were performed in 96-well black clear bottom plates containing 0.5 µg purified Lucs (except that 0.1 µg of Luc2 was used with LH 2 ). Assay mix (0.1 mL) containing 0.1 mM analog, 0.1 mM Na-ATP, 1 mM MgSO 4 in 50 mM Tricine, pH 7.4 was then added to each well. After a 45 s incubation to ensure that emission decay was minimal, BL was measured with an IVIS Spectrum III (Perkin Elmer, Waltham, MA, USA) using the auto setting, FOV B, and filters set at OPEN, 520 ± 20 nm, 570 ± 20 nm, 620 ± 20 nm, 670 ± 20 nm, 710 ± 20 nm, 755 ± 15 nm and 790 ± 20 nm. Data were analyzed with the Living Image 4.7 Software (Perkin Elmer, Waltham, MA, USA) by selecting the appropriate region of interest (ROI) and were reported as radiance (p/s/cm 2 /sr). The radiance values were corrected for any emission decay that occurred as the filters were sequentially imaged. Experiments were performed in triplicate and individual trials were repeated at least three times. The intensities used to calculate the values presented in Figure 3 are based on the sum of the data through the following filters: visible (520 nm, 570 nm, 620 nm, and 670 nm); nIR optical window (670 nm, 710 nm, 755 nm, and 790 nm); and nIR (710nm, 755 nm, and 790 nm).

Cell Culture and Transfection
HEK293T cells were grown in Dulbecco's modified Eagles medium supplemented with 10% fetal bovine serum. Cells were counted using a TC10 automated cell counter (BioRad, Hercules, CA, USA) and plated at 1,250,000 cells per well in a 6-well plate and grown at 37 • C with 5% CO 2 for 4-6 h prior to transfection.

Live Cell Imaging in Transfected HEK293T Cells
Transfected cells were grown for 20 h at 37 • C with 5% CO 2 and were released from the plate with 0.05% trypsin-EDTA (Gibco, Waltham, MA, USA), resuspended in growth media, and counted. Each well of a black clear-bottom 96-well plate was seeded with 50,000 cells in 0.1 mL growth media and grown for an additional 24 h. For all series of substrates examined by each Luc, a separate set of quadruplicate wells was seeded for Nluc activity analysis using the Nano Glo Luciferase Assay Kit (Promega, Madison WI, USA). Growth media was removed and replaced with 0.05 mL DMEM without phenol red plus 10% FBS. Nluc assay buffer mix (0.05 mL) was added, mixed by pipetting and BL was measured after a 3 min incubation at room temp. BL was measured using an IVIS Spectrum III (Perkin Elmer) with the auto exposure, FOV B, and OPEN filter settings. For the Luc-substrate assays, growth media was removed and 0.1 mL of 0.5 mM solutions of LH 2 or analogs in assay buffer (50 mM Tricine (pH 7.4), growth media (1:1, v/v), and 5 µM ATP) were added to each well. After a 30 s incubation, BL was measure at 37 • C. Signals were monitored for 8 min with a measurement at 30 s intervals. Data were analyzed with the Living Image 4.7 Software (Perkin Elmer) after selecting the appropriate region of interest (ROI). The highest average Luc-substrate activity recorded within the first 5 min was used to calculate the final average radiance (p/s/cm 2 /sr). In order to account for daily variations in transfection efficiencies, the reported radiance values were calculated from the mean ± standard deviation of BL signals corrected by the respective Nluc activities. Each experiment was repeated at least 3 times.