Bioluminescence imaging in mice with synthetic luciferin analogues

Abstract

Luciferase enzymes from bioluminescent organisms can be expressed in mice, enabling these rodents to glow when treated with a corresponding luciferin substrate. Light emission occurs where the expression of the genetically-encoded luciferase overlaps with the biodistribution of the administered small molecule luciferin. Here we discuss differences between firefly luciferin analogues for bioluminescence imaging, focusing on transgenic and adeno-associated virus (AAV)-transduced mice.

Summary

Bioluminescence will only occur where the expression of the luciferase overlaps with the biodistribution of the luciferin. It is thus important to consider differences in substrate behavior, and how these differences might affect the parameters under study. Is the goal of the experiment to accurately report on where a luciferase is expressed, or is it to detect changes in bioluminescence only in a particular cell, tissue, or in response to a particular analyte or enzyme? The number of luciferin analogues is increasing rapidly, as is the number of luciferases and methods for their introduction. Together, these tools improve the ability to utilize bioluminescence imaging to report on specific molecular events in live animals, but also necessitate a more nuanced understanding of their interpretation.

1. Introduction

Bioluminescence occurs when a luciferase enzyme oxidizes its luciferin substrate to produce an excited-state oxyluciferin. When this molecule relaxes to the ground state, a photon is emitted. Unlike fluorescence, which requires an incident photon to access an excited-state molecule, bioluminescence occurs in the absence of light. The energy needed to access an excited state comes not from a photon, but from breaking an oxygen-oxygen bond.

There are multiple examples of luciferases in bacteria, insects, fungi, dinoflagellates, and a variety of marine organisms (Kaskova et al., 2016), but only a handful of luciferins. The majority of in vivo bioluminescence imaging studies to date have focused on just two molecules: d-luciferin, the natural substrate for beetle luciferases, and coelenterazine, the natural substrate for many marine luciferases (Figure 1). Here we describe how synthetic analogues of the firefly luciferase substrate d-luciferin differ in their in vivo behavior, and the general implications for other classes of luciferins.

Figure 1.

Luciferins. A) Firefly luciferase mechanism and examples of synthetic luciferin substrates (differences from d-luciferin denoted in blue); B) CycLuc1-amide and PCL-1 are examples of bioluminogenic reporters. CycLuc1-amide is hydrolyzed by the enzyme fatty acid amide hydrolase (FAAH) to unmask the luciferin substrate CycLuc1, whereas PCL-1 is oxidized by hydrogen peroxide to release d-luciferin; C) Examples of luciferins for marine luciferases from Renilla and the Oplophorus-derived mutant NanoLuc.

1.1. Distinctions between bioluminescence and fluorescence

Bioluminescence has several advantages over fluorescence for imaging in small animals such as mice. First, fluorescence requires excitation light, which is attenuated by scattering and depth-dependent tissue absorption by endogenous chromophores such as the hemoglobin in blood (Weissleder and Ntziachristos, 2003). Bioluminescence does not require excitation, so there is no attendant signal loss. Secondly, mice exhibit autofluorescence background upon photoexcitation, as endogenous chromophores can be excited and relax to the ground state by emitting light. Furthermore, the excited state of these endogenous chromophores, as well as the exogenous fluorescent probes, can lead to phototoxicity. Bioluminescence avoids these caveats.

Although there is no excitation light, the bioluminescence emission wavelength is still important, as the light exiting the animal will experience the same scattering and tissue attenuation mentioned above. Consequently, the blue bioluminescence of marine luciferases such as Renilla (Loening et al., 2010), Gaussia (Tannous et al., 2005), and the Oplophorus-derived NanoLuc (Hall et al., 2012), which all utilize coelenterazine and similar luciferin substrates (Figure 1), is more strongly attenuated than that of beetle luciferases with peak emission in the green and red. Maximal depth penetration will be achieved by near-infrared wavelengths, beyond the visible range, where few if any endogenous chromophores can absorb the emitted light. In general, the spectral emission of luciferases is broad, and luciferases that appear to be green in color will also emit photons that extend into the red and near-infrared wavelengths. Firefly luciferase itself has peak emission of ~560 nm at room temperature, but shifts to a peak of ~612 nm at 37 °C, and has substantial emission >650 nm (Zhao et al., 2005). Furthermore, the bioluminescence signal that exits the mouse is red-shifted in a depth-dependent manner, as the intervening tissue acts as a filter for the shorter wavelengths. This effect has some advantages, as it can be used to estimate the depth of the light source, but also makes it challenging to perform multi-color bioluminescence imaging at anything other than superficial depths. Efforts to red-shift the emission of firefly luciferase further into the red and near-infrared have largely focused on synthetic luciferins (Iwano et al., 2013; Jathoul et al., 2014; Mofford et al., 2014; Reddy et al., 2010), whereas efforts to red-shift the emission of NanoLuc have mostly focused on resonance energy transfer to yellow, orange, or red fluorescent proteins (Chu et al., 2016; Schaub et al., 2015; Yeh et al., 2019b).

1.2. The importance of luciferin biodistribution

On their own, luciferases are not useful for imaging. It is only when a luciferase contacts a suitable luciferin substrate that it is capable of generating an excited-state emitter. The need for this interaction means that the Km, biodistribution and pharmacokinetics of the small molecule luciferin substrate are crucial. In many ways, one can consider the luciferin to be a drug, and the luciferase the drug target (Mofford and Miller, 2015). The result of target engagement is light emission.

It is useful to consider the difference between common small-molecule and genetically-encoded fluorescent reporters. For a fluorescent protein like GFP, the fluorophore is self-contained, and the fluorescent signal is localized exclusively to the protein (Ormo et al., 1996). However, if a small molecule fluorescent probe is used in vivo, the fluorescent signal tracks with the small molecule, and essentially reports on its biodistribution. Often, dominant signals are observed in organs responsible for drug metabolism and excretion: the liver and bladder (Choi et al., 2011).

In the case of bioluminescence, there are both small-molecule (luciferin) and genetically-encoded (luciferase) components. Light emission only occurs where the luciferase expression overlaps with luciferin biodistribution – which is not necessarily where the luciferin or luciferase is primarily located. This distinction can be useful for the design of specific probes, but as discussed in more detail below, should also be kept in mind when interpreting bioluminescence imaging experiments.

1.3. Development of synthetic beetle luciferin substrates

d-luciferin is the natural substrate for firefly luciferase and other similar beetle luciferases. All beetle luciferases function by adenylating d-luciferin, then oxidizing the adenylated intermediate to access an excited-state oxyluciferin (Figure 1). This light-emitting chemistry has evolved several times in beetles, with apparent advantages to these insects (Adams and Miller, 2020; Fallon et al., 2018). Scientists have appropriated this bioluminescent reaction to image cell proliferation, gene expression, and other specific molecular events in a wide variety of contexts, including in live animals (Contag and Bachmann, 2002; Prescher and Contag, 2010). However, what is good for beetles (e.g., room temperature, no blood, specialized light organs, signals meant to be readily detectable by others) is not likely to be optimal for imaging in mice (37 °C, blood that strongly absorbs visible wavelengths, deep tissues with transport barriers, exogenous introduction of luciferin-luciferase systems).

In the past several years, numerous luciferin analogues have been reported that retain the ability to emit light, but allow the modulation of properties of interest for imaging, such as wavelength and access to tissues (Adams and Miller, 2014; Evans et al., 2014; Jathoul et al., 2014; Kuchimaru et al., 2016; Mofford et al., 2014). Corresponding mutations to luciferases have also been utilized to improve the emissive properties with specific luciferins (Adams et al., 2016; Harwood et al., 2011; Iwano et al., 2018; Jones et al., 2017). Furthermore, additional modifications to d-luciferin or its synthetic analogues have been used to create a wide variety of “caged” bioluminescent sensors (Figure 1). Such molecules can be used to detect specific analytes or enzymes by virtue of their release of a functional luciferin substrate from a non-luminogenic precursor (Su et al., 2019). As described above, the location of luciferin release and luciferase expression may not completely overlap. This distinction can be utilized for proximity detection (Sellmyer et al., 2013), but in many cases the biodistribution of luciferin analogues and caged luciferin reporters is not fully characterized.

New luciferin analogues continue to be developed at a rapid pace, many of which have yet to be fully characterized in vivo (Hall et al., 2018; Ikeda et al., 2020; Miller et al., 2018; Sharma et al., 2019, 2017; Zhang et al., 2018). Here we illustrate and discuss the effect of the luciferin on bioluminescence imaging using the natural firefly luciferase substrate d-luciferin, the synthetic luciferins CycLuc1 (Reddy et al., 2010) and CycLuc6 (Mofford et al., 2014), and the synthetic “caged” luciferin CycLuc1-amide (Mofford et al., 2015).

2. General considerations for in vivo imaging with synthetic luciferins

2.1. Substrate preparation and dosing

Historically, drug dosing in mice is described in “mg/kg”, or milligrams of drug per kilogram of mouse. However, for chemists and the more molecularly-inclined, drug dosing is typically expressed as a molar quantity. Hence, with the recent development of new luciferins of different molecular weights, dosing has been described in micromoles of luciferin per kilogram of mouse (μmol/kg), or simply in molar units, to facilitate molecule-to-molecule comparisons.

d-luciferin is very water soluble, and is typically supplied to mice via i.p. injection of a 100 mM aqueous solution into the lower right quadrant of the abdomen. CycLuc1 is more hydrophobic, but can be directly dissolved in PBS to 5 mM. Luciferin amides such as CycLuc1-amide lack a free carboxylate and are considerably less soluble in PBS. We typically dilute a 50 mM DMSO stock into PBS to give a final concentration of 0.1–0.25 mM luciferin amide, and a final DMSO concentration that is 0.5% or lower. Synthetic aminoluciferins are generally effective at much lower doses than d-luciferin, due in part to lower Km values and potentially greater ability to traverse biological barriers (Adams and Miller, 2014; Mofford et al., 2015). On the other hand, reducing the dose of d-luciferin substantially reduces the bioluminescent signal. We thus typically compare mice dosed with the standard 100 mM d-luciferin to PBS-soluble concentrations of aminoluciferins such as 5 mM CycLuc1 and 0.25 mM CycLuc1-amide.

Since not all mice are the same mass, we inject 4 μL of each PBS stock luciferin solution per gram of mouse, so that a “typical” 25 g mouse gets a 100 μL i.p. injection, 400 μmol/kg for d-luciferin. The corresponding doses using 5 mM CycLuc1 and 0.25 mM CycLuc1-amide are 20 μmol/kg and 1 μmol/kg. We make stock solutions in a slightly larger volume than needed, and just prior to injection we use a low-holdup 0.22 μm syringe filter (Millex-GV, SLGV004SL) to ensure removal of any particulates or microbes while minimizing sample loss.

Some luciferin analogues have been injected using high concentrations of DMSO (e.g., 50%) (Van de Bittner et al., 2010). Although this can improve the solubility of caged and synthetic luciferin analogues, such high concentrations of DMSO are not generally recommended in IACUC protocols. The synthetic luciferin Akalumine has been supplied as the hydrochloride salt, which greatly increases its aqueous solubility (Iwano et al., 2018). Potentially this can be more generally applied to increase the aqueous solubility of other aminoluciferin analogues, but there is some concern that the acidity of such preparations could cause irritation or toxicity (Iwano et al., 2018; Yeh et al., 2017), especially upon repeated injection.

2.2. Choice of animals

The most common animal used for bioluminescence imaging is the mouse, but many of the general principles and guidelines are also applicable to other small animals. For example, it is best to avoid using animals with black fur. The fur greatly attenuates the bioluminescent signal, and the need for shaving limits the signal location to the shaved area as the fur will mask signal elsewhere. A better practice is to use strains with white fur, or breed animals into a white/albino background. We typically use FVB/NJ, BALB/cJ, and white C57BL/6J mice. Sometimes black C57BL/6 mice must be used when that is the only model that is available (as it is the most common strain for transgenics), but the best practice for optical imaging is to breed these animals into a white background.

We acquire background images of each luciferase-expressing mouse before administering any luciferin. Mice are obviously not naturally bioluminescent, but contact with phosphorescent materials can give background signal. More importantly, synthetic luciferins can have a considerably longer half-life than d-luciferin. For example, after i.p. injection of 4 μmol/kg CycLuc6 (Mofford et al., 2014), a particularly lipophilic substrate (Figure 1), we found residual bioluminescent signal that persisted for up to three days (Figure 2). Furthermore, mice are coprophagic. Indeed, with some substrates, we have observed bioluminescence from mice that had not previously been injected with a luciferin, but had simply been co-housed with treated mice. It is therefore prudent to clean mouse cages frequently, particularly when using synthetic luciferin substrates that can be active at very low concentration.

Figure 2.

Bioluminescence imaging with the lipophilic substrate CycLuc6 (100 μL of 1 mM in PBS, i.p.) in the ubiquitously-expressing luciferase mouse strain FVB-Tg(CAG-luc,-GFP)L2G85Chco/J. Biodistribution is sluggish and emission persists for days.

The kinetics of bioluminescence emission can vary between luciferins and organs. After luciferin administration, we typically alternate between dorsal and ventral views to best capture signal over time from both dorsal and ventral organs. In some cases these signals can be viewed from both orientations, whereas in others, strong signal emanating from one side (e.g., the liver) can reflect so that it appears at a different location on the opposite side (i.e., near the kidneys). Generally, peak emission for D-luciferin occurs ~10 minutes after i.p. injection and is cleared in ~2 hours, whereas peak emission from luciferin amide analogues is achieved more rapidly, after ~5 min in the brain. However, more lipophilic substrates can deviate significantly from these norms (Figure 2).

3. Imaging of transgenic luciferase mice

3.1. Background

Bioluminescence imaging requires the expression of a luciferase. Perhaps the most common in vivo application is for monitoring luciferase-expressing tumor cells that have been implanted into a mouse that does not express luciferase. Here we focus instead on mice that express firefly luciferase within their own cells and tissues. A number of transgenic luciferase mice are commercially available from JAX or Taconic. For example, FVB-Tg(CAG-luc,-GFP)L2G85Chco/J (“CAG-luc”) mice, sometimes also called “FVB-luc+” mice, express luciferase in most cells and tissues behind a synthetic CAG promoter consisting of a human CMV early enhancer fused to a hybrid chicken beta-actin/rabbit beta-globin promoter (Cao et al., 2004). FVB/N-Tg(Gfap-luc)-Xen (“GFAP-luc”) mice express luciferase behind a glial fibrillary acidic promoter, which is selectively expressed in astrocytes (Zhu et al., 2004). B6.129S6-Per2^tm1Jt/J (“mPER2::LUC”) mice express a fusion protein between mouse period circadian clock 2 (mPER2) and luciferase, which reports on circadian rhythms (Yoo et al., 2004). In addition, Cre-Lox technology can be used to generate transgenic mice with restricted luciferase expression, for example by crossing a Cre-driver line with the commercially available Cre reporter luciferase mouse FVB.129S6(B6)-Gt(ROSA)26Sor^tm1(Luc)Kael/J (Safran et al., 2003). Examples include B6.SJL-Slc6a3^{tm1.1(cre)Bkmn}/J (DAT^IREScre; dopaminergic neurons) (Bäckman et al., 2006; Evans et al., 2014) and B6.Cg-Speer6-ps1^{Tg(Alb-Cre)21Mgn}/J (Albumin-Cre; liver) (Heffern et al., 2016; Safran et al., 2003).

Of these examples, the transgenic luciferase mouse strain that has been most commonly used with synthetic luciferins is the ubiquitously-expressing CAG-luc mouse. With appropriate controls, meaningful data can be obtained from the application of caged luciferin reporters in these mice (Heffern et al., 2016; Mofford et al., 2015; Van de Bittner et al., 2010). However, a challenge with interpreting in vivo data from these mice is that overlapping and superficial signals can swamp signals from deeper tissues. So, for example, imaging the “brain area” to report on the brain bioluminescence of these mice is not particularly useful for substrates that will also emit light from the skin above the skull. The use of transgenic mice with more restricted luciferase expression can be helpful in this regard, as described above (Evans et al., 2014; Heffern et al., 2016). Conclusions regarding luciferin access based on the exogenous introduction of luciferase-expressing tumor cells are also fraught with caveats. For example, injecting human cells into the brain and then imaging them does not provide meaningful information about the ability of luciferins to traverse the endogenous mouse blood-brain barrier (BBB) (Mofford and Miller, 2015). Similar concerns apply to the imaging of subcutaneous luciferase-expressing tumor cells by injection of a luciferin adjacent to, or even directly into, the tumor.

3.2. Case study: imaging of glial fibrillary acidic protein (GFAP)-luc mice

Glial fibrillary acidic protein (GFAP) is selectively expressed in astrocytes, and FVB/N-Tg(Gfap-luc)-Xen (“GFAP-luc”) mice have been used as a model to detect astrogliosis in the progression of neurodegenerative diseases such as ALS (Keller et al., 2009) and Alzheimer’s (Watts et al., 2011). However, the ability to detect GFAP expression in the brain using d-luciferin is complicated by high basal GFAP-luciferase expression in the ears, thought to be due to GFAP-expressing cells in the cochlear spiral ganglion (Kanzaki et al., 2012). Researchers often attempt to cover the ears in an effort to reduce this problematic background signal (Keller et al., 2009; Watts et al., 2011). When the ears are not covered, they often overwhelm any endogenous signal from the brain (Kong et al., 2012).

We compared the behavior of d-luciferin, CycLuc1, and CycLuc1-amide in the GFAP-luc mouse model (Figure 3). Mice were injected i.p. with each respective luciferin as described above, then anesthetized with 2% isofluorane and imaged with a sensitive CCD camera (IVIS-100, Perkin-Elmer). When d-luciferin is used, the brightest bioluminescent signal emanates from the ears rather than the astrocytes in the brain (Figure 3). This is consistent with previous reports, but is also a little confounding – why should the bioluminescent signal from the brain be weak when the astrocytes are expressing luciferase? Is GFAP expression really higher in the ears than in the brain, or is this an artifact caused by the inability of d-luciferin to accurately report on luciferase expression in different tissues? Notably, imaging the same GFAP-luc animals with CycLuc1 enables detection of bioluminescence from the brain (Figure 3), as one would expect if astrocytes are expressing luciferase. This is consistent with an enhanced ability to cross the BBB compared to d-luciferin (Evans et al., 2014), rather than an actual change in GFAP expression, and suggests that d-luciferin is underreporting GFAP expression in the brain.

Figure 3.

Representative GFAP-luc mouse (male) imaged 10 minutes after i.p. injection with d-luciferin (400 μmol/kg), CycLuc1 (20 μmol/kg), and CycLuc1-amide (1 μmol/kg).

CycLuc1-amide lacks a free carboxylate, and is not a direct substrate for firefly luciferase (Figure 1). Conversion of this luciferin amide to CycLuc1 requires fatty acid amide hydrolase (FAAH), restricting its activation to FAAH-expressing tissues such as the brain (Adams et al., 2016; Mofford et al., 2015). Consistent with this requirement, treatment of GFAP-luc mice with CycLuc1-amide results in bioluminescence exclusively from the brain and spinal cord, and not the ears (Figure 3). In addition to illustrating inherent differences in luciferin substrate access, these results suggest that the use of GFAP-luc mice to detect astrogliosis in neurodegenerative diseases or after lipopolysaccharide (LPS) treatment could be confounded by breakdown of the BBB and the concomitant greater access of d-luciferin to brain tissue (Ayzenberg et al., 2015; Mofford and Miller, 2015). Detection of BBB breakdown has potential utility, but must be clearly differentiated from changes in GFAP-positive cells and expression levels.

4. Imaging using adeno-associated viral (AAV) reporters

4.1. Background

AAV is widely used as a gene therapy vector due to its lack of pathogenicity and low immunogenicity, combined with its ability to transduce both diving and non-dividing cells. AAV viral vectors can allow tissue-selective delivery of genes into a wide variety of animal models (e.g., mouse, rat, rabbit, cat, dog, monkey) as well as humans, depending on the route of administration and the capsid serotype (Srivastava, 2016). Tail-vein injection of AAV9-luciferase reporters into mice primarily transduces leg muscles, the heart, and the liver (Mofford et al., 2015; Zincarelli et al., 2008). Stereotactic injection into the brain transduces cells near the site of injection, limiting expression to the brain (Evans et al., 2014). In these mice, d-luciferin is notably dimmer than CycLuc1 (Evans et al., 2014), which is consistent with its biodistribution (Berger et al., 2008).

Recent reports of modified AAV9 capsids with improved ability to cross the BBB have allowed more facile expression in the brain after systemic administration (e.g., via retro-orbital or tail-vein injection). AAV-PHP.eB is a modified AAV9 capsid that can readily cross the BBB in C57BL/6 mice (Chan et al., 2017; Deverman et al., 2016). Follow-up work has shown that its ability to traverse the BBB requires the presence of a specific variant of a GPI-anchored protein, lymphocyte antigen 6 complex, locus A (Ly6a) (Hordeaux et al., 2019; Huang et al., 2019). This variant is present in C57BL/6 mice, the mouse strain in which the initial screen for this modified capsid was conducted, but is notably absent in BALB/c mice. Although PHP.eB is unable to traverse the BBB in BALB/c mice, primates, and several other mouse strains, it can cross the BBB in FVB mice (Challis et al., 2019; Hordeaux et al., 2019). The full scope of its utility in other rodent models has not yet been firmly established.

4.2. Case study: imaging of a luciferase reporter delivered with AAV-PHP.eB

The pAAV-CMV-luc2 plasmid (Adams et al., 2016; Evans et al., 2014; Mofford et al., 2015) was packaged into AAV-PHP.eB by the UMass Viral Vector Core. Six-week old FVB/NJ mice (two female, one male) were injected in the lateral tail vein with 1×10¹¹ vector genomes of AAV-PHP.eB-CMV-luc2 in 200 μL of sterile-filtered PBS. The animals were imaged at least three weeks later and the data analyzed using Living Image software.

Unlike AAV9-CMV-luc2 (Mofford et al., 2015), tail-vein injection of AAV-PHP.eB-CMV-luc2 results in strong brain transduction (Figure 4), consistent with the literature (Challis et al., 2019; Chan et al., 2017). Although the PHP.eB vector effectively transduces the brain and has considerably lessened delivery to leg muscles compared to AAV9, we find that that it retains tropism for the heart. Imaging of the AAV-PHP.eB-transduced mice with different luciferins on separate days also revealed important differences. First, CycLuc1 gives much brighter signals from the brain and spinal cord than d-luciferin (Figure 4). On the other hand, the bioluminescent signal in the heart is similar for both substrates, and similar to what we have observed for tail-vein injection of AAV9-CMV-luc2 (Mofford et al., 2015). CycLuc1-amide also yields bright bioluminescence from the brain and spinal cord, but no bioluminescence from the heart, consistent with the lack of FAAH expression in this organ (Mofford et al., 2015).

Figure 4.

Bioluminescence imaging of three FVB/NJ mice transduced with AAV-PHP.eB-CMV-luc2 (from left to right: two female, one male). Dorsal (A–C) and ventral (D–F) views were taken at 10 and 7 min respectively after i.p. injection of the same set of mice with 400 μmol/kg d-luciferin (A and D), 20 μmol/kg CycLuc1 (B and E), and 1 μmol/kg CycLuc1-amide (C and F).

In experiments involving animals, it is important to consider sex as a biological variable. When imaging male and female mice, we have observed some differences that are glaring and obvious, such as the detection of FAAH activity in the testes with CycLuc1-amide (Miller et al., 2018). Other differences can be more subtle. Sex-dependent liver expression is evident with all of the luciferins (Figure 4). Indeed, with AAV-transduced animals, male mice are known to yield higher gene expression in the liver (Davidoff et al., 2003). Furthermore, FAAH is expressed in the liver, and thus bioluminescence from CycLuc1-amide in the liver is detected only in male mice (Figure 4).

Current transgenic luciferase mice are largely restricted to firefly luciferase. Construction of new AAVs is considerably easier than the generation of transgenic mice, and offers an efficient way to introduce new luciferase reporters. Indeed, several mutant luciferases have already been incorporated into AAVs and used for bioluminescence imaging (Adams et al., 2016; Iwano et al., 2018). Furthermore, AAVs can be readily introduced into different tissues in the mouse model of choice as well as a variety of other animals. AAVs are thus well suited for the evaluation of new luciferases in vivo as well as the ready introduction of AAV-based luciferase reporters into new mouse models. In addition to modification of the viral capsid and route of administration, the use of tissue-specific promoters (Domenger and Grimm, 2019) and microRNAs (Xie et al., 2011) could be used to further refine and restrict expression as needed.

5. Considerations for marine luciferases and luciferins

Imidazopyrazinone-based luciferins such as coelenterazine (Figure 1) do not require ATP for activation, and can be substrates for a variety of structurally-distinct marine luciferases that can yield brighter bioluminescence in vitro than beetle luciferases. However, these substrates have posed some challenges for in vivo use, as they are generally unstable and poorly soluble in aqueous solutions. A number of different solvents and additives have been included to help address this challenge, but there is limited information on which is optimal for maximizing luciferin delivery while minimizing toxicity. When injected i.p., there is also mixed data on whether furimazine and coelenterazine faithfully report on the location of cells that express NanoLuc or NanoLuc-FP fusions such as Antares2 (Iwano et al., 2018; Stacer et al., 2013; Yeh et al., 2019b). In a mouse model in which Antares2-expressing cells were delivered to the lungs, bioluminescence was not confined to the lungs (Iwano et al., 2018). This may reflect poor biodistribution of the substrate after i.p. injection, and/or an ability to detect NanoLuc that is secreted outside of live cells or released from lysed cells (Yeh et al., 2019b, 2019a). The ability to image marine luciferases outside of cells opens up possibilities that are not readily accessible for ATP-dependent beetle luciferases, but can also complicate the detection of populations of live cells. On the other hand, studies in which NanoLuc-expressing cells were detected in the lungs have been reported using i.v. injection of the substrate (Stacer et al., 2013). Pending further studies, i.v. injection is recommended for the most faithful reporting of bioluminescence signal within cells from NanoLuc and NanoLuc-based reporters. Although i.v. injection is operationally more difficult and results in rapid signal kinetics, this route has been successfully used in the past for imaging with Renilla luciferase and coelenterazine. Going forward, the construction of transgenic animals or AAV-based reporters with well-defined, restricted expression of NanoLuc and other marine luciferases could help establish the optimal parameters for using these enzymes. Potentially, new imidazopyrazinone luciferin analogues with improved solubility and biodistribution could further expand the flexibility of these luciferases for in vivo imaging.