In vivo fluorescence lifetime imaging of macrophage intracellular metabolism during wound responses in zebrafish

  1. Veronika Miskolci
  2. Kelsey E Tweed
  3. Michael R Lasarev
  4. Emily C Britt
  5. Alex J Walsh
  6. Landon J Zimmerman
  7. Courtney E McDougal
  8. Mark R Cronan
  9. Jing Fan
  10. John-Demian Sauer
  11. Melissa C Skala  Is a corresponding author
  12. Anna Huttenlocher  Is a corresponding author
  1. Department of Medical Microbiology and Immunology, University of Wisconsin-Madison, United States
  2. Morgridge Institute for Research, United States
  3. Department of Biomedical Engineering, University of Wisconsin-Madison, United States
  4. Department of Biostatistics & Medical Informatics, University of Wisconsin-Madison, United States
  5. Department of Nutritional Sciences, University of Wisconsin-Madison, United States
  6. Department of Molecular Genetics and Microbiology, Duke University School of Medicine, United States
  7. Department of Pediatrics, University of Wisconsin-Madison, United States

Abstract

The function of macrophages in vitro is linked to their metabolic rewiring. However, macrophage metabolism remains poorly characterized in situ. Here, we used two-photon intensity and lifetime imaging of autofluorescent metabolic coenzymes, nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and flavin adenine dinucleotide (FAD), to assess the metabolism of macrophages in the wound microenvironment. Inhibiting glycolysis reduced NAD(P)H mean lifetime and made the intracellular redox state of macrophages more oxidized, as indicated by reduced optical redox ratio. We found that TNFα+ macrophages had lower NAD(P)H mean lifetime and were more oxidized compared to TNFα− macrophages. Both infection and thermal injury induced a macrophage population with a more oxidized redox state in wounded tissues. Kinetic analysis detected temporal changes in the optical redox ratio during tissue repair, revealing a shift toward a more reduced redox state over time. Metformin reduced TNFα+ wound macrophages, made intracellular redox state more reduced and improved tissue repair. By contrast, depletion of STAT6 increased TNFα+ wound macrophages, made redox state more oxidized and impaired regeneration. Our findings suggest that autofluorescence of NAD(P)H and FAD is sensitive to dynamic changes in intracellular metabolism in tissues and can be used to probe the temporal and spatial regulation of macrophage metabolism during tissue damage and repair.

Editor's evaluation

Immunometabolism is an emerging field, and to understand immune cell metabolism during inflammation and infection is of great interest. In this report, cutting edge microscopy techniques and innovative zebrafish models are used to characterize the metabolism of macrophages in situ. In the future, fluorescence microscopy approaches pioneered using zebrafish may illuminate strategies to therapeutically manipulate metabolism in human immune cells.

https://doi.org/10.7554/eLife.66080.sa0

Introduction

Macrophages are innate immune cells that play key functions in tissue repair (Krzyszczyk et al., 2018; Wynn and Vannella, 2016). The heterogeneity and diversity of macrophage phenotypes and functions are well documented in vitro (Martinez and Gordon, 2014; Mills et al., 2014; Murray, 2017). However, there is a gap in understanding macrophage phenotypes in interstitial tissues during tissue damage and repair. This is particularly important because distinct macrophage populations play important roles in wound healing and tissue regeneration.

Macrophages are commonly described as classically (M1) or alternatively (M2) activated, with both subsets playing critical roles in wound healing (Krzyszczyk et al., 2018; Wynn and Vannella, 2016). The M1/M2 classification, especially in the context of in vivo biology, is controversial (Murray et al., 2014; Orecchioni et al., 2019), and more likely represents a continuum of activation states. The importance of metabolic regulation of macrophage function was not appreciated until more recently, when it was recognized that some metabolic pathways are profoundly altered in classically activated macrophages (Jha et al., 2015; Tannahill et al., 2013). For example, classically activated macrophages are glycolytic, while oxidative phosphorylation is the main fuel source during alternative activation in vitro (O’Neill et al., 2016; Van den Bossche et al., 2017). This recent progress has led to the emergence of metabolic reprogramming as a hallmark of immune cell activation, and supports the premise that the metabolic state is not an outcome but rather a determinant of immune cell activation and function (O’Neill and Pearce, 2016; Ryan and O’Neill, 2020).

While it is well documented that macrophages exhibit plasticity as wounds repair and convert from M1 to M2 over the course of wound healing (Krzyszczyk et al., 2018), the metabolic regulation of macrophage function within interstitial tissue during wound repair remains unclear (Caputa et al., 2019). We need additional tools to detect the polarization and metabolic phenotypes of macrophages within interstitial tissues in live animals.

Autofluorescence imaging of the intensities and lifetimes of metabolic coenzymes is an attractive method to monitor macrophage metabolism and function in vivo. The reduced forms of nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and oxidized flavin adenine dinucleotide (FAD) are endogenous metabolic coenzymes that are autofluorescent. The fluorescence intensities of NAD(P)H and FAD can be used to determine the optical redox ratio (Table 1), which provides a label-free method to monitor the oxidation-reduction state of the cell (Chance et al., 1979). Multiple definitions of the optical redox ratio exist, but here we use NAD(P)H/(NAD(P)H + FAD), since an increase in the optical redox ratio intuitively corresponds with a more reduced intracellular environment, suggestive of an increase in glycolysis, and it normalizes the values to be between 0 and 1 (Walsh et al., 2021). The fluorescence lifetime measures the time a molecule spends in the excited state before decaying back to the ground state. The fluorescence lifetimes of NAD(P)H and FAD are distinct in their free and protein-bound states, which provide a label-free measurement of their enzyme-binding activities (Table 1; Georgakoudi and Quinn, 2012; Kolenc and Quinn, 2019). Fluorescence lifetime imaging microscopy (FLIM) has several advantages over intensity measurements, because FLIM provides additional biological information by distinguishing the protein-bound and free states, and is not dependent on the cellular concentrations of the coenzymes (Datta et al., 2020; Walsh and Skala, 2015). Importantly, FLIM is a label-free and noninvasive method to detect metabolic changes in situ and can also resolve metabolic heterogeneity within a cell population (Heaster et al., 2019; Sharick et al., 2019; Walsh et al., 2021; Walsh and Skala, 2015).

Table 1
Definition of autofluorescence imaging endpoints.
EndpointsDefinitionInterpretation
Optical redox ratio
Chance et al., 1979
INAD(P)HINAD(P)H+IFAD
Increase in optical redox ratio (ORR) = more reduced intracellular environment; likely increase in glycolysis; decrease in ORR = more oxidized intracellular environment; likely decrease in glycolysis. (I, intensity)
Nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) Short lifetime of NAD(P)HFree/unbound NAD(P)H
NAD(P)H τ2Long lifetime of NAD(P)HNAD(P)H bound to a protein
NAD(P)H α1Fractional component of free NAD(P)Hα2 is fractional component of bound NAD(P)H, α1+α2=1; quantifies the pools of NAD(P)H in free and bound states
NAD(P)H mean lifetime ()τm=τ1α1+τ2α2Weighted average of individual lifetime endpoints (); one can look at changes in individual endpoints to see what drives changes in τm; for instance, a decrease in τm can be due to increase in α1, decrease in τ1, and/or decrease in τ2
Flavin adenine dinucleotide (FAD) τ1Short lifetime of FADFAD bound to a protein
FAD τ2Long lifetime of FADFree/unbound FAD
FAD α1Fractional component of bound FADα2 is fractional component of free FAD, α1+α2=1; quantifies the pools of FAD in free and bound states
FAD τmτm=τ1α1+τ2α2
Weighted average of individual lifetime endpoints (τ1,τ2,α1,α2); one can look at changes in individual endpoints to see what drives changes in τm; for instance, a decrease in τm can be due to increase in α1, decrease in τ1, and/or decrease in τ2
Optical Metabolic Imaging (OMI) index
Walsh and Skala, 2015
ORRi<ORR>+NAD(P)Hτmi<NAD(P)Hτm>FADτmi<FADτm>Composite measure of mean-centered optical redox ratio and mean lifetimes of NAD(P)H and FAD; increase in the OMI index corresponds to increased redox ratio, and increased NAD(P)H and FAD protein-binding activities

Zebrafish represents a powerful system to study macrophage polarization and tissue repair. Live imaging has revealed the presence of both M1 (TNFα+) and M2 (TNFα−) macrophages in wounded tissues (Miskolci et al., 2019; Nguyen-Chi et al., 2017; Nguyen-Chi et al., 2015). Here, we performed autofluorescence imaging of NAD(P)H and FAD to assess changes in the metabolic activity of macrophages in response to tissue damage in live zebrafish. We show that these measurements detect metabolic changes in macrophages within interstitial tissue in response to sterile damage and microbial cues with temporal and spatial resolution. We also show that perturbations that modulate macrophage polarization and metabolism affect tissue repair.

Results

Autofluorescence imaging detects oxidized intracellular redox state in macrophages in vivo upon 2-deoxy-d-glucose treatment

To determine if a known glycolysis inhibitor alters macrophage metabolism in vivo, we imaged NAD(P)H and FAD in wounded zebrafish larvae in the absence and presence of 2-deoxy-d-glucose (2-DG). 2-DG is a glucose analog and acts as a competitive inhibitor of glycolysis at the step of phosphorylation of glucose by hexokinase (Pelicano et al., 2006). To isolate autofluorescence signals associated with macrophages from the whole tissue, we used mCherry and green fluorescent protein (GFP) transgenic reporter lines. GFP is suitable to image in conjunction with NAD(P)H, but it excludes the acquisition of FAD because they have overlapping spectra (Datta et al., 2020; Qian et al., 2021), while mCherry is compatible for simultaneous imaging with NAD(P)H and FAD (Heaster et al., 2021; Hoffmann and Ponik, 2020). The traditional serial acquisition of NAD(P)H and FAD was not suitable for imaging motile cells, such as macrophages, in live larvae. To accommodate cell movement during image acquisition in live larvae, we employed wavelength mixing that allows for simultaneous acquisition in three different channels (Stringari et al., 2017). We performed simple tail fin transection on transgenic larvae (Tg(mpeg1:mCherry-CAAX) that labels the plasma membrane of macrophages with mCherry), and performed autofluorescence imaging of NAD(P)H and FAD at the wound region (Figure 1A) at 3–6 hr post tail transection (hptt) in the absence or presence of 2-DG. As inhibiting glycolysis reduces NADH levels (Georgakoudi and Quinn, 2012; Kolenc and Quinn, 2019), we expected the optical redox ratio to decrease in macrophages of 2-DG-treated larvae compared to untreated control. Indeed, the optical redox ratio was significantly lower in macrophages in the 2-DG-treated larvae (Figure 1B and C). This change was driven by a decrease in NAD(P)H intensity in treated larvae, while FAD intensity remained similar to control levels (data not shown). Inhibition of glycolysis was associated with a significant reduction of the mean lifetime (τm) of NAD(P)H in macrophages, with only a marginal reduction in FAD τm (Figure 1D and E). We also observed significant reduction for NAD(P)H and FAD τ2, and increase for NAD(P)H α1 (Figure 1—figure supplement 1A-F). These effects on NAD(P)H and FAD lifetime endpoints were similar to the effects observed with 2-DG treatment of activated T cells (Walsh et al., 2021). In sum, macrophages were more oxidized following treatment with a glycolysis inhibitor, and these findings support the utility of using autofluorescence imaging of metabolic coenzymes to detect changes in metabolic activity of macrophages in situ.

Figure 1 with 1 supplement see all
Inhibition of glycolysis reduces the optical redox ratio of macrophages following simple transection.

Tail fin transection distal to the notochord was performed using transgenic zebrafish larvae (Tg(mpeg1:mCherry-CAAX) that labels macrophages in the plasma membrane with mCherry) at 3 days post fertilization, and autofluorescence imaging of nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and flavin adenine dinucleotide (FAD) was performed on live larvae at 3–6 hr post tail transection (hptt) that were either untreated (control) or treated with 5 mM 2-deoxy-d-glucose (2-DG) (glycolysis inhibitor) for 1 hr prior to imaging. (A) Schematic showing the area where wounding (black line) and imaging (blue box) were performed. (B) Representative images of mCherry (to show macrophages), optical redox ratio, and NAD(P)H and FAD mean lifetimes (τm) are shown; macrophages in mCherry channel were outlined with dashed lines and the area was overlaid in the optical redox ratio and lifetime images to show corresponding location; scale bar = 50 µm. Quantitative analysis of (C) optical redox ratio, (D) NAD(P)H and (E) FAD mean lifetimes (τm) from two biological repeats (control = 90 cells/9 larvae, 2-DG = 123 cells/9 larvae) is shown; quantitative analysis of associated individual lifetime endpoints (τ1,τ2,α1) and sample size for each repeat are included in Figure 1—figure supplement 1. The optical redox ratio and τm were log transformed prior to analysis. p values represent statistical analysis of the overall effects. Estimated means with 95% CI and overall effects with p values are included in Figure 1—source data 1.

Autofluorescence imaging detects metabolic changes in macrophages at the infected tail wound

To determine if autofluorescence imaging could distinguish different macrophage populations in a whole organism, we used a zebrafish Listeria monocytogenes (Lm)-infected tail wound model (Miskolci et al., 2019). This infection induces the recruitment of M1 macrophages as detected by a high level of TNFα expression (Miskolci et al., 2019). In contrast, most macrophages are devoid of TNFα expression following a simple transection (Miskolci et al., 2019), and likely represent a differentially activated M2-like population (Nguyen-Chi et al., 2015).

Based on the differential activation profiles, we hypothesized that we would detect differences in the metabolic activity of macrophages at the simple and Lm-infected transection wounds. We performed tail fin transection in the absence or presence of Lm on double transgenic (Tg(tnf:GFP) × Tg(mpeg1:mCherry-CAAX)) larvae and performed autofluorescence imaging of NAD(P)H at the wound region on live larvae at 48 hr post wound (hpw). 48 hpw was chosen as a representative timepoint when the proportion of TNFα+ cells at these wounds was sufficiently different between the two wound models; most macrophages (~70%) are TNFα− in response to simple transection, whereas most macrophages (~80%) are TNFα+ at the infected wound, and these proportions do not significantly change at later timepoints (Miskolci et al., 2019). We performed autofluorescence imaging in conjunction with the TNFα reporter line (tnf:GFP) in order to monitor and group macrophages by TNFα expression during image analysis. The TNFα reporter line relies on GFP expression to report transcriptional activity of tnfα (Marjoram et al., 2015), which precludes acquisition of FAD measurements. As a result, in this experiment we were not able to monitor changes in the intracellular optical redox ratio. Macrophages at the wound region were identified based on plasma membrane-localized mCherry expression as above. The infected tail wound recruited significantly more macrophages compared to the uninfected control (simple transection), and most macrophages at the infected tail wound expressed high levels of TNFα, while the majority lacked TNFα expression at the uninfected control wounds (Figure 2A, Figure 2—figure supplement 1D), consistent with our previous report (Miskolci et al., 2019). We detected a significant reduction in the mean lifetime (τm) of NAD(P)H for TNFα+ macrophages relative to TNFα− macrophages in both the uninfected control and Lm-infected tail wounds (Figure 2B). While both types of wounds showed similar trends for TNFα+ and TNFα− macrophages, NAD(P)H τm was further reduced in macrophages from Lm-infected tail wounds relative to uninfected control when comparing either the TNFα− or TNFα+ groups (Figure 2B). Similar trends were observed for the individual lifetime components (τ1,τ2) of NAD(P)H as those observed for τm (Figure 2—figure supplement 1A,B), while we did not detect any significant changes in the fractional component of free NAD(P)H (α1) in any of the comparisons (Figure 1—figure supplement 1C).

Figure 2 with 1 supplement see all
Mean lifetime of nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) is reduced in TNFα+ macrophages at the infected tail wound.

Tail fin transection distal to the notochord was performed using N-phenylthiourea-treated double transgenic zebrafish larvae (Tg(tnf:GFP) x Tg(mpeg1:mCherry-CAAX), a TNFα reporter line in combination with a line that labels macrophages in the plasma membrane with mCherry) at 3 days post fertilization in the absence or presence of Listeria monocytogenes (Lm). Autofluorescence imaging of NAD(P)H was performed on live larvae at 48 hr post wound (Figure 1A). (A) Representative images of mCherry expression to show macrophages, GFP to show TNFα expression, and NAD(P)H mean lifetime (τm) are shown for control or Lm-infected tail wounds. Macrophages in the mCherry channel were outlined with dashed lines and the area was overlaid in GFP and lifetime images to show corresponding location; in the infected condition only a few macrophages are outlined as examples; scale bar = 50 µm. (B) Quantitative analysis of NAD(P)H mean lifetime (τm) from three biological repeats (control TNFα− = 184 cells/16 larvae, control TNFα+ = 75 cells/16 larvae, infected TNFα− = 258 cells/16 larvae, infected TNFα+ = 789 cells/16 larvae) is shown; quantitative analysis of associated individual lifetime endpoints (τ1,τ2,α1) and sample size for each repeat are included in Figure 2—figure supplement 1. The τm was log transformed prior to analysis. Interaction between treatment and GFP expression was included to analyze whether either factor modified the effect of the other; no interaction was found. p values represent statistical analysis of the overall effects. Estimated means with 95% CI and overall effects with p values are included in Figure 2—source data 1.

Next, we repeated the same set of experiments but without the TNFα reporter, to acquire FAD measurements in order to monitor changes in the optical redox ratio. NAD(P)H τm, τ1 and τ2 were significantly reduced in macrophages at the Lm-infected wound (Figure 3D, Figure 3—figure supplement 1A,B), consistent with the measurements above (Figure 2B, Figure 2—figure supplement 1A,B). We found that NAD(P)H α1 significantly increased in macrophages at the Lm-infected wound (Figure 3—figure supplement 1C). The presence of infection at the tail wound did not induce any significant changes in FAD lifetime endpoints (Figure 3—figure supplement 1D-G). Interestingly, the optical redox ratio was significantly reduced in macrophages at the highly inflammatory Lm-infected wound compared to the uninfected control, indicating that TNFα+ (M1-like) macrophage population is more oxidized compared to TNFα− (M2-like) in vivo (Figure 3A and B). The Optical Metabolic Imaging (OMI) index, a composite measure of the optical redox ratio, NAD(P)H τm and FAD τm (Table 1), was also lower in macrophages at the Lm-infected wound (Figure 3C). These findings were unexpected considering the observed increase of the optical redox ratio in the context of in vitro infection of bone marrow-derived macrophages (BMDM) with Lm (Figure 3—figure supplement 1I,J), indicating a more reduced intracellular redox state, consistent with previous publications of Listeria infection of macrophages in vitro (Gillmaier et al., 2012). We reasoned this result may be influenced by the presence of an intracellular pathogen in macrophages, and not solely due to a more proinflammatory macrophage phenotype. To test this, we next measured changes in the intracellular metabolism of macrophages in the context of thermal injury that also induces a TNFα+ macrophage population, but in the absence of infection.

Figure 3 with 1 supplement see all
Optical redox ratio and mean lifetime of nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) are reduced in macrophages at the infected tail wound.

Tail fin transection distal to the notochord was performed using transgenic zebrafish larvae (Tg(mpeg1:mCherry-CAAX) that labels macrophages in the plasma membrane with mCherry) at 3 days post fertilization in the absence or presence of Listeria monocytogenes (Lm). Autofluorescence imaging of NAD(P)H and flavin adenine dinucleotide (FAD) was performed on live larvae at 48 hr post wound (Figure 1A). (A) Representative images of mCherry expression to show macrophages, optical redox ratio, and NAD(P)H and FAD mean lifetimes (τm) are shown for control or infected tail wounds; macrophages were outlined with dashed lines and the area was overlaid in the optical redox ratio and lifetime images to show corresponding area; in the infected condition only a few macrophages are outlined as examples; scale bar = 50 µm. Quantitative analysis of (B) optical redox ratio, (C) Optical Metabolic Imaging index, and (D) NAD(P)H mean lifetime (τm) from three biological repeats (control = 105 cells/16 larvae, infected = 761 cells/14 larvae) is shown; quantitative analysis of associated NAD(P)H and FAD mean (τm) and individual lifetime endpoints (τ1,τ2,α1), and sample size for each repeat are included in Figure 3—figure supplement 1. p values represent statistical analysis of the overall effects. Estimated means with 95% CI and overall effects with p values are included in Figure 3—source data 1.

Autofluorescence imaging resolves changes in the metabolic activity of macrophages over the course of thermal tissue damage

To measure metabolic activity of macrophages during robust tissue damage, we used our zebrafish thermal injury tail wound model (LeBert et al., 2018). The burn wound elicits the recruitment of an M1-like macrophage population, as detected by TNFα expression (Miskolci et al., 2019). Unlike at the infected wound where TNFα expression in macrophages persists, TNFα+ macrophages peak at 24 hpw and resolve thereafter, as most macrophages at the wound are TNFα− by 72 hpw following thermal injury (Miskolci et al., 2019). We chose these two timepoints to compare the metabolic activity of macrophages in response to simple transection and thermal injury. Since macrophages are mostly TNFα− throughout the course of the wound response following a simple transection, we hypothesized that the metabolic activity of macrophages would be different at 24 hpw, but similar at 72 hpw, when comparing the two wound models.

We performed tail transection or generated a burn wound distal to the notochord on transgenic (Tg(mpeg1:mCherry-CAAX)) larvae, and performed autofluorescence imaging at the wound region on live larvae at 24 and 72 hpw. As expected, we observed significant differences in the metabolic activity of macrophages between the wounds at 24 hpw, but the cellular metabolism was similar at 72 hpw. Importantly, macrophages at the burn wound had a more oxidized redox state relative to macrophages at the simple transection at 24 hpw, indicated by the lower optical redox ratio and OMI index (Figure 4A–C). In addition, NAD(P)H τm and τ1 were lower, while α1 was higher in macrophages at the burn compared to simple transection at 24 hpw (Figure 4D, Figure 4—figure supplement 1A,C); we did not detect any significant changes in FAD lifetime endpoints at 24 hpw (Figure 4—figure supplement 1D-G). Macrophages are mostly TNFα− at both wound types by 72 hpw (Miskolci et al., 2019), suggesting that the macrophage populations present at these wounds have similar activation states and are thereby likely to have similar metabolic activity. Accordingly, the optical redox ratio and OMI index were not different between macrophages of the simple transection and burn wound at 72 hpw (Figure 4B and C). The mean lifetime of NAD(P)H was significantly lower in macrophages at the burn wound relative to the simple transection at 72 hpw (Figure 4D), while it was similar for FAD (Figure 4—figure supplement 1D). Most of the individual lifetime endpoints (τ1,τ2 and α1) of NAD(P)H and FAD were also similar between the two wounds at 72 hpw (Figure 4—figure supplement 1A-G). Furthermore, we also detected temporal changes in the metabolic activity of macrophages during wound responses. The optical redox ratio and OMI index of macrophages increased over time at both wound types (Figure 4B and C), indicating a more reduced redox state. This would be expected as TNFα+ macrophages resolve at both wound types over time (Miskolci et al., 2019). In line with this, NAD(P)H τm,τ1 and τ2 increased, while α1 decreased at both wound types over time (Figure 4D, Figure 4—figure supplement 1A-C). We also detected time-related changes in FAD endpoints; FAD τm, and τ1 decreased over time at both wounds (Figure 4—figure supplement 1D,E). Collectively, we found that TNFα+ macrophage population was more oxidized, as indicated by a decrease in the redox ratio, and was associated with a decrease in NAD(P)H mean lifetime relative to a TNFα− macrophage population in context of both infected and sterile injury (Table 2).

Table 2
Summary of changes in optical redox ratio, Optical Metabolic Imaging (OMI) index, and nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) lifetime endpoints.

Changes in treated samples are shown relative to the control. Control, simple tail transection (uninfected); nd, not different.

Control versus2-Deoxy-d-glucoseLm-infected wound TNFα−Lm-infected wound TNFα+Lm-infected wound
Burn wound24 hrBurn wound72 hr
Optical redox ratio--nd
OMI index---nd
NAD(P)H τm
NAD(P)H τ1ndnd
NAD(P)H τ2ndnd
NAD(P)H α1ndnd
Figure 4 with 1 supplement see all
Autofluorescence imaging resolves temporal changes in the metabolic activity of macrophages at sterile tail wounds.

Tail fin transection (Tt) or thermal injury (burn) distal to the notochord was performed using transgenic zebrafish larvae (Tg(mpeg1:mCherry-CAAX) that labels macrophages in the plasma membrane with mCherry) at 3 days post fertilization. Autofluorescence imaging of nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and flavin adenine dinucleotide (FAD) was performed on live larvae at 24 and 72 hr post wound (Figure 1A). (A) Representative images of mCherry expression to show macrophages, optical redox ratio, and NAD(P)H and FAD mean lifetimes (τm) are shown for Tt or burn wounds; macrophages were outlined with dashed lines and the area was overlaid in the optical redox ratio and lifetime images to show corresponding location of macrophages; scale bar = 50 µm. Quantitative analysis of (B) optical redox ratio, (C) Optical Metabolic Imaging index, and (D) NAD(P)H mean lifetime (τm) from three biological repeats (Tt-24 hr = 322 cells/16 larvae, burn-24 hr = 850 cells/14 larvae, Tt-72 hr = 213 cells/12 larvae, burn-72 hr = 578 cells/11 larvae) is shown; quantitative analysis of associated NAD(P)H and FAD mean (τm) and individual lifetime endpoints (τ1,τ2,α1), and sample size for each repeat are included in Figure 4—figure supplement 1. Interaction between treatment and time was included to analyze whether either factor modified the effect of the other; strong interaction was detected for the optical redox ratio. p values represent statistical analysis of the overall effects. Estimated means with 95% CI and overall effects with p values are included in Figure 4—source data 1. (E) Tail fin tissue was collected distal to the caudal vein/artery loop (blue box) 24 hr following either tail transection or thermal injury distal to the notochord (black line) for mass spec analysis of small metabolites to compare the global trend of changes in redox metabolites with that measured by autofluorescence imaging; metabolomics data shown in (E) and (F) are from four biological repeats. (F) Metabolite abundance measured by either autofluorescence imaging or mass spec in transection sample was normalized by that in burn or (G) was used to calculate the redox ratio in transection (Tt) or burn samples. We included NADPH abundance in the redox ratio calculated using mass spec measurements. *NADPH and NADH intensities could not be collected separately by autofluorescence imaging as their fluorescence spectra overlap, thereby were measured collectively.

To substantiate the metabolic changes observed by autofluorescence imaging in macrophages at the sterile tail wounds, we tested if we would see similar changes in tail fin tissue using targeted liquid chromatography-mass spectrometry (LS-MS)-based method to analyze the abundance of NADH, NAD(P)H, and FAD. We performed simple transection or burn wound distal to the notochord on unlabeled wild-type zebrafish larvae and collected the tail fin tissue distal to the caudal vein/artery loop (to remain close to the wound microenvironment) 24 hr following injury for targeted LC–MS metabolite analysis (Figure 4E). The technical limitation here is that we analyzed the abundance of these small metabolites in the whole tail fin tissue, not macrophages alone, because it is difficult to collect enough macrophages from such a small region to reach the detection limit of the mass spectrometer. We calculated the relative abundances of NAD(P)H and FAD in burn wound compared to transection, and found the trends measured by autofluorescence imaging and mass spectrometry to be consistent, in that the NAD(P)H levels are similar in both wound models, while FAD level is lower in transection (Figure 4F). Additionally, we compared the redox ratio measured in each wound model by both methods (Figure 4G). The two methods gave similar results, with both showing a trend toward a higher redox ratio in the transection model. These findings suggest that the redox state of macrophages and the whole tail fin tissue are similar after tissue damage.

Metformin increases the optical redox ratio of wound macrophages

To further validate the use of autofluorescence imaging to characterize macrophage metabolism in vivo, we used metformin to modulate macrophage polarization. Metformin attenuates proinflammatory activation of macrophages in vitro and promotes an M2-like phenotype (Schuiveling et al., 2018). We previously reported that metformin also reduces the number of TNFα+ cells in zebrafish liver cancer models (de Oliveira et al., 2019). We found that metformin treatment reduced the proportion of TNFα+ macrophages at the burn wound, without decreasing the number of macrophages (Figure 5A and B, Figure 5—figure supplement 1A). We performed thermal injury of the tail fin on control or metformin-treated larvae and performed autofluorescence imaging on live larvae at 24 hr following injury. By decreasing TNFα+ cells, we expected to observe an increase in the optical redox ratio and NAD(P)H τm. Indeed, we measured a significant increase in the optical redox ratio and OMI index of macrophages in metformin-treated larvae (Figure 5C–E). We also detected an increase in the mean lifetime of NAD(P)H, albeit not statistically significant (Figure 5F); NAD(P)H τ1 and τ2 also increased as expected, however, α1 did not change (Figure 5—figure supplement 1B-D). The modest changes in NAD(P)H lifetime endpoints are likely attributed to the small shift in macrophage polarization upon metformin treatment (Figure 5B).

Figure 5 with 1 supplement see all
Metformin treatment increases optical redox ratio of macrophages.

(A, B) Thermal injury (burn) distal to the notochord was performed using double transgenic zebrafish larvae (Tg(tnf:GFP x mpeg1:mCherry-CAAX) that labels macrophages in the plasma membrane with mCherry, and monitors TNFα expression using GFP-based TNFα reporter) at 3 days post fertilization (dpf), that were either untreated (control) or treated with 1 mM metformin starting at 1 dpf; larvae were fixed at 24 hr post burn (hpb). (A) Representative images of sum projections of z-stacks acquired by spinning disk confocal microscopy are shown; scale bar = 100 μm. (B) TNFα expression in macrophages were quantified by area thresholding of GFP and mCherry intensities, and is displayed as proportional area per larva with bars showing arithmetic mean and 95% CI; results are from three biological repeats (control = 42, metformin = 46 larvae). (C–F) Thermal injury distal to the notochord was performed using control or metformin-treated transgenic zebrafish larvae (Tg(mpeg1:mCherry-CAAX)) at 3 dpf. Autofluorescence imaging of nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and flavin adenine dinucleotide (FAD) was performed on live larvae at 24 hpb. (C) Representative images of mCherry expression to show macrophages, optical redox ratio, and NAD(P)H and FAD mean lifetimes (τm) are shown; scale bar = 50 µm. Quantitative analysis of (D) optical redox ratio, (E) Optical Metabolic Imaging index, and (F) NAD(P)H mean lifetime (τm) from three biological repeats (control = 632 cells/13 larvae, metformin = 670 cells/14 larvae) is shown; quantitative analysis of associated NAD(P)H and FAD mean (τm) and individual lifetime endpoints (τ1,τ2,α1), and sample size for each repeat are included in Figure 5—figure supplement 1. Log transformation was applied to optical redox ratio prior to analysis. p values represent statistical analysis of the overall effects. Estimated means with 95% CI and overall effects with p values are included in Figure 5—source data 1.

Macrophage metabolic switches are associated with changes in tissue repair

To determine if altering macrophage polarization and metabolic phenotype is associated with a change in wound healing, we treated thermal wounds with metformin or used STAT6-deficient zebrafish and characterized the effect on tissue regrowth. We first characterized the metabolic phenotype of macrophages in STAT6-depleted larvae, using a recently characterized stat6 mutant (Cronan et al., 2021). STAT6 is a known regulator of M2-like macrophage polarization (Murray, 2017). In accordance with in vitro studies, we found that depletion of STAT6 led to an increase in the proportion of TNFα+ macrophages at the wound at 72 hr post burn (hpb) (Figure 6A and B), although there were fewer total macrophages at the wound (Figure 6—figure supplement 1B). Importantly, STAT6 depletion did not affect the total number of macrophages in whole larvae (Figure 6—figure supplement 1A). In accordance with the increase in TNFα+ macrophages at the wound, we detected a significant reduction in the optical redox ratio and NAD(P)H τm of macrophages in the stat6 mutant larvae at 72 hpb (Figure 6D and F); the OMI index was also lower, albeit not statistically significant (Figure 6E). Although the proportion of TNFα+ macrophages was higher in the mutant larvae at 72 hpb, it was decreasing by 96 hpb (Figure 6B), suggesting that STAT6 depletion merely delays the switch from a TNFα+ macrophage population to TNFα−at the wound, but does not prevent it. To determine if this switch in the polarization of macrophages at the wound induces changes in wound healing, we quantified the impact of metformin and STAT6 depletion on regeneration after thermal injury. We found significantly improved wound healing in the presence of metformin, as indicated by the larger area of tissue regrowth (Figure 6G and H). By contrast, depletion of STAT6 resulted in more TNFα+ macrophages, reduced optical redox ratio, and impaired wound healing (Figure 6I and J). This is consistent with prior work suggesting that the presence of TNFα+ macrophages at the wound is associated with impaired wound healing (Krzyszczyk et al., 2018; Miskolci et al., 2019). Taken together, these findings suggest that the metabolic activity of macrophages in the wound microenvironment correlates with the ability of damaged tissues to heal.

Figure 6 with 1 supplement see all
Macrophage metabolic switches are associated with changes in tissue repair.

(A–B) Thermal injury distal to the notochord was performed using N-phenylthiourea-treated double transgenic wild-type (wt) or stat6-deficient zebrafish larvae (Tg(tnf:GFP x mpeg1:mCherry-CAAX)) at 3 days post fertilization (dpf); larvae were fixed at 72 and 96 hr post burn (hpb). (A) Representative images of sum projections of z-stacks acquired by spinning disk confocal microscopy are shown; scale bar = 100 μm. (B) TNFα expression in macrophages was quantified by scoring cells for GFP signal and is displayed as proportion of TNFα+ cells per larva with bars showing arithmetic mean and 95% CI; results are from three biological repeats (wt = 33, stat6−/− = 46 at 72 hpb, wt = 33, stat6−/− = 20 larvae at 96 hpb). Larvae were generated by a het incross and genotyped post imaging; only wt and stat6−/− larvae were analyzed. (C–F) Thermal injury distal to the notochord was performed using transgenic zebrafish larvae (Tg(mpeg1:mCherry-CAAX)) in wt- or stat6-deficient background at 3 dpf. Autofluorescence imaging of nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and flavin adenine dinucleotide (FAD) was performed on live larvae at 72 hpb. (C) Representative images of mCherry expression to show macrophages, optical redox ratio, and NAD(P)H and FAD mean lifetimes (τm) are shown; macrophages were outlined with dashed lines and the area was overlaid in the optical redox ratio and lifetime images to show corresponding location of macrophages; scale bar = 50 µm. Quantitative analysis of (D) optical redox ratio, (E) Optical Metabolic Imaging index, and (F) NAD(P)H mean lifetime (τm) from three biological repeats (wt = 217 cells/12 larvae, stat6−/− = 272 cells/12 larvae) is shown; quantitative analysis of associated NAD(P)H and FAD mean (τm) and individual lifetime endpoints (τ1,τ2,α1), and sample size for each repeat are included in Figure 6—figure supplement 1. p values represent statistical analysis of the overall effects. Estimated means with 95% CI and overall effects with p values are included in Figure 6—source data 1. (G) Control or metformin-treated larvae were fixed at 72 hpb following thermal injury of the tail fin. Representative single-plane brightfield images are shown; scale bar = 100 μm. (H) Quantitative analysis of tail fin tissue regrowth area per larva, displayed with bars showing arithmetic mean and 95% CI. Results are from three biological repeats (control = 59, metformin = 60 larvae). (I) Wt- or stat6-deficient larvae were fixed at 96 hpb following thermal injury of the tail fin. Representative single-plane brightfield images are shown; scale bar = 100 μm. (J) Quantitative analysis of tail fin tissue regrowth area per larva, displayed with bars showing arithmetic mean and 95% CI. Results are from three biological repeats (wt = 36, stat6−/− = 24 larvae). Larvae were generated by a het incross and genotyped post imaging; statistical analysis was performed on wt, stat6± and stat6−/−; only wt and stat6−/− larvae are shown.

Discussion

Understanding in vivo behavior has been limited by the lack of tools for the assessment of functional metabolic changes in live organisms. As a result, immunometabolism in vivo remains poorly characterized. Autofluorescense imaging of the endogenous fluorescence of metabolic coenzymes is an attractive approach because it allows for the quantitative analysis of metabolic changes on a single-cell level, while maintaining cells in their native microenvironment. Studies on the metabolic profiles of macrophages in vivo using autofluorescence imaging have been limited, with one study demonstrating that macrophages have distinguishable lifetime signatures from tumor cells in the tumor microenvironment (Szulczewski et al., 2016), and one study demonstrating changes in optical redox ratio, and mean lifetimes (τm) of NAD(P)H and FAD between dermal and tumor macrophages in vivo in mice (Heaster et al., 2021).

Here, we took advantage of transparent zebrafish larvae to image macrophage metabolism in live animals with temporal and spatial resolution. We found that a proinflammatory (TNFα+) macrophage population was more oxidized relative to a TNFα− population, as indicated by a lower optical redox ratio and OMI index. The TNFα+ population was also associated with lower NAD(P)H τm,τ1 and τ2, and these measurements were consistent across the proinflammatory wound models (Table 2). These results may reflect a reduction in glycolytic activity, as these changes are similar to what we found with 2-DG inhibition of glycolysis (Table 2).

In vitro infection of murine macrophages by Lm exhibited increased optical redox ratio relative to uninfected macrophages (Figure 3—figure supplement 1I,J), however, we found that infection of zebrafish larvae reduced the optical redox ratio in macrophages (Figure 3B). Surprisingly, we found that proinflammatory macrophages at sterile inflammatory sites were also associated with reduced optical redox ratios (Figure 4B), suggesting that in general, a proinflammatory macrophage population in vivo has a more oxidized redox state (Table 2). These findings suggest that macrophage metabolism in vitro and in vivo may differ and raises interesting questions for future investigation. Physiological functions in vitro do not always translate to in vivo settings; one example is the requirement of integrins for leukocyte migration on two-dimensional substrates in vitro, but it is dispensable for migration within three-dimensional interstitial tissues (Lämmermann et al., 2008). This metabolic variation most likely reflects the differences in the inherent nature of in vitro and in vivo microenvironments, such as interactions with other cells (Van den Bossche and Saraber, 2018). These findings underscore the importance of understanding macrophage metabolism directly in their native microenvironment and the development of new tools to probe these activities in situ.

We characterized macrophage populations in live zebrafish larvae using the TNFα reporter (Marjoram et al., 2015), a well-established marker of classically activated M1-like macrophages (Murray, 2017; Nguyen-Chi et al., 2015). The TNFα reporter has been used in several studies to identify M1-like macrophages in zebrafish (de Oliveira et al., 2019; Miskolci et al., 2019; Nguyen-Chi et al., 2017; Nguyen-Chi et al., 2015; Roh-Johnson et al., 2017). In light of the known in vitro metabolic profiles of macrophages, we expected a macrophage population with large numbers of TNFα+ cells to exhibit a more glycolytic state and thereby have higher redox ratio relative to a mostly TNFα− macrophage population. One caveat of intensity-based measurements is that other fluorophores, such as elastin and lipofuscin, could contribute to the intensity signals for the redox ratio and be a source of error (Datta et al., 2020). Another caveat is that NADH also exists in a phosphorylated form (NADPH) that is autofluorescent and has overlapping spectral properties with NADH Blacker and Duchen, 2016; hence for the sake of accuracy we use NAD(P)H to reflect their combined signal. The source and role of the observed oxidative metabolism in proinflammatory macrophages in the context of infection and sterile inflammation in vivo requires further analysis. Our results with metformin treatment, a drug known to inhibit the production of mitochondrial reactive oxygen species (mROS) at complex I of the electron transport chain in the mitochondria (Schuiveling et al., 2018), suggest that mROS contributes to the observed trends in the optical redox ratio and NAD(P)H lifetime profiles during sterile inflammation. It will be interesting to further explore the role of mROS in macrophage activation and function in vivo.

We found that a proinflammatory macrophage population is also associated with a decrease in mean lifetime of NAD(P)H (Table 2). Similarly, inhibition of metabolic reactions in the cell (e.g. rotenone for oxidative phosphorylation, 2-DG for glycolysis) decreases NAD(P)H τm in vitro and in vivo (Walsh et al., 2021; Yaseen et al., 2017). Weights are given to the free (α1) and bound (α2) components, so that pools of NAD(P)H in the free or bound state can be quantified. The lifetimes themselves (τ1,τ2) are affected by preferred protein-binding activities in the cell ( >300 proteins bind to NAD(P)H in the cell Berman et al., 2000), and microenvironmental factors (e.g. pH, viscosity, and temperature). Therefore, the NAD(P)H lifetime is a unique biophysical measurement that is influenced by the amount of NAD(P)H in the free and bound pools, preferred protein-binding activities in the cell, and microenvironmental factors. In the current study, we report the mean lifetime of NAD(P)H (τm) as the weighted average of the short and long lifetime components (τm=α1τ1+α2τ2). Therefore, a decrease in τm is due to an increase in the pool of free NAD(P)H, a decrease in τ1, and/or a decrease in τ2 (Table 1).

Interestingly, we found that TNFα expression in macrophages at the control and infected tail wounds was associated with a graded effect on NAD(P)H lifetime endpoints that was on continuum (Figure 2B, Figure 2—figure supplement 1A-C). TNFα− macrophages from uninfected wound (control) are on one end of the spectrum, while TNFα+ cells from the infected wound are on the opposite end. As we move on this spectrum, we see a graded change in the mean and individual lifetime endpoints in the same direction from one end to the other end, reminiscent of the concept that macrophage activation in vivo occurs in a continuum of activation states as opposed to a more strict M1 or M2 classification (Murray, 2017). These results suggest that autofluorescence imaging is sensitive to detect variations in macrophage populations across different levels of activation. The single cell-based approach is one key advantage of autofluorescence imaging as it allows for analyzing metabolic heterogeneity in a cell population. In future studies we will apply distribution density models as in Heaster et al., 2019; Walsh et al., 2021 to monitor heterogeneity in macrophages during wound responses.

Our findings show that autofluorescence imaging is also capable of resolving time-related changes in macrophage metabolism. We observed that the optical redox ratio and OMI index increased in macrophages over time, indicating that the intracellular redox state becomes more reduced over time, both at the simple transection and the burn wound (Figure 4B and C). This was expected based on our previous report that the early macrophage population at the burn wound is mostly TNFα+, however, over time the macrophage population becomes mostly TNFα−, similar to the simple transection, and this switch in activation phenotype coincided with a recovery in wound healing (Miskolci et al., 2019). The observed increase in the optical redox ratio over time is interesting. Macrophages polarize toward a prohealing M2-like state during wound healing, thus based on existing literature macrophages would be expected to rely on oxidative metabolism (Caputa et al., 2019; O’Neill et al., 2016) and thereby display a reduction in the optical redox ratio. However, it has been demonstrated that M2-like macrophages are more motile compared to M1-like cells (Hind et al., 2016) and glycolytic reprogramming has been shown to be important for macrophage migration (Semba et al., 2016). These reports suggest that our observed increase in the optical redox ratio at the wound, reflecting an increase in glycolytic activity, may be supporting the more motile nature of prohealing M2-like cells, however, this requires further investigation. Nevertheless, our data suggests that a reduced intracellular redox state in macrophages supports wound healing; this is supported by the results of the metformin treatment showing that an increase in the optical redox ratio (Figure 5D) is associated with improved wound healing (Figure 6H) and the converse by Stat6 depletion (Figure 6D and J). Our results are in line with recent zebrafish studies that found that tail transection leads to a shift to glucose metabolism and wound healing is blocked upon 2-DG treatment (Sinclair et al., 2021), and progenerative macrophages display a glycolytic phenotype in context of muscle injury as indicated by increased levels of intracellular NADH measured by two-photon autofluorescence imaging and bioluminescence-based assay (Ratnayake et al., 2021).

Finally, our findings demonstrate a correlation between macrophage M1 phenotype and impaired wound healing. Treatment with metformin reduced the TNFα expression in macrophages and improved wound healing. By contrast, depletion of STAT6 increased TNFα expression and impaired wound healing (Figure 6). It will be interesting to determine if macrophage intrinsic STAT6 is sufficient to regulate its metabolic activity and the fate of tissue repair.

With the emergence of immunometabolism it is now recognized that metabolic reprogramming underlies macrophage activation and function. Differential activation of macrophages plays a central role in host health and disease progression, underscoring the importance of studying macrophage metabolism in vivo. We have shown that fluorescence intensity and lifetime imaging of NAD(P)H and FAD can resolve metabolic changes in macrophages with distinct activation states in situ in a live organism, suggesting that this approach can be a valuable label-free imaging-based tool to study the metabolic regulation of immune cell function in vivo.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Strain, strain background (10,403 S)Listeria monocytogenes (strain 10,403 S)PMID:26468080Was be obtained from JD Sauer Lab, University of Wisconsin - Madison
Strain, strain background (Danio rerio)WT (AB)ZIRCZL1https://zebrafish.org/home/guide.php
Strain, strain background (D. rerio)Tg(tnf:GFP) (AB)PMID:25730872Was obtained from Michel Bagnat Lab, Duke University
Strain, strain background (D. rerio)Tg(mpeg1:histone2b-GFP) (AB)PMID:31259685Can be obtained from Anna Huttenlocher Lab, University of Wisconsin - Madison
Strain, strain background (D. rerio)Tg(mpeg1:mCherry-CAAX) (albino)PMID:26887656Was obtained from Leonard Zon Lab, Boston Children’s Hospital, Dana Farber Cancer Institute
Strain, strain background (D. rerio)stat6 mutant (AB)PMID:33761328Was obtained from David Tobin Lab, Duke University
Chemical compound, drugMetforminEnzo Life Sciences, cat no ALX-270–432 G005https://www.enzolifesciences.com/ALX-270-432/metformin/1 mM in E3, bathing, start treatment at 1 day post fertilization, refresh daily
Chemical compound, drug2-Deoxy-d-glucoseSigma, cat no D8375https://www.sigmaaldrich.com/US/en/product/sigma/d83755 mM in E3, bathing, 1 hr pretreatment right before imaging
Software, algorithmGraphPad PrismRRID:SCR_002798https://www.graphpad.com/scientific-software/prism/
Software, algorithmSASRRID:SCR_008567https://www.sas.com/en_us/home.html
Software, algorithmFiji, ImageJSchindelin et al., 2012RRID:SCR_002285https://fiji.sc/
Software, algorithmCell profilerRRID:SCR_007358https://cellprofiler.org/
Software, algorithmMatplotlibRRID:SCR_008624https://matplotlib.org/
Software, algorithmR project for statistical computingR Core Team, 2019RRID:SCR_001905https://www.r-project.org/
Software, algorithmMATLAB R2019bhttps://www.mathworks.com/
Software, algorithmSPCImage 7.4https://www.becker-hickl.com/products/spcimage/
OtherLine-powered thermal cautery instrumentStoelting59,005https://stoeltingco.com/Neuroscience/Thermal-Cautery-Instruments~9879
OtherType E tip for cautery instrumentStoelting59,010https://stoeltingco.com/Neuroscience/Thermal-Cautery-Instruments~9879

Zebrafish husbandry

Request a detailed protocol

All protocols using zebrafish and mice in this study have been approved by the University of Wisconsin-Madison Research Animals Resource Center (protocols M005405-A02/zebrafish, M005916/mouse). Adult zebrafish were maintained on a 14  hr:10  hr light/dark schedule. Upon fertilization, embryos were transferred into E3 medium and maintained at 28.5°C. To prevent pigment formation, larvae were maintained in E3 medium containing 0.2 mM N-phenylthiourea (PTU) (Sigma-Aldrich, St. Louis, MO) starting at 1 day post fertilization (dpf). Adult wild-type zebrafish, transgenic lines Tg(tnf:GFP) (Marjoram et al., 2015), and Tg(mpeg1:histone2b-GFP) (Miskolci et al., 2019) in wild-type or stat6-deficient zebrafish (Cronan et al., 2021) in AB background, and Tg(mpeg1:mCherry-CAAX) (Bojarczuk et al., 2016) in wild-type or stat6-deficient zebrafish in albino background were utilized in this study. To genotype stat6-deficient zebrafish, genomic DNA was used in a GoTaq Green PCR reaction (cat no. M7123; Promega, Madison, WI) with forward primer 5’-TATGCAGTTCCCTCCCTTCG-3’ and reverse primer 5’-AGCTGATGAAGTGTTTGGCG-3’ (Cronan et al., 2021); PCR products were resolved by a 3% MetaPhor agarose gel (Lonza Rockland, Rockland, ME).

Bacterial culture and preparation

Request a detailed protocol

Unlabeled or mCherry-expressing Lm strain 10,403 was used in this study (Vincent et al., 2016). Lm were grown in brain–heart infusion (BHI) medium (Becton, Dickinson and Company, Sparks, MD). A streak plate from frozen stock was prepared and grown overnight at 37°C; the plate was stored at 4°C. The day before infection, a fresh colony was picked from the streak plate and grown statically in 1 mL BHI overnight at 30°C to reach stationary phase and to flagellate bacteria. The next day bacteria were prepared to infect either primary macrophages or zebrafish larvae. To prepare for infection of primary cells, the 1 mL suspension was diluted with 1 mL sterile PBS, OD was determined to calculate the number of bacteria to infect cells at multiplicity of infection (MOI) of 2 (1 cell:2 bacteria) (OD 1 = 7.5 × 108 bacteria). To prepare for zebrafish tail wound infection, bacteria were subcultured for ~1.5–2.5  hr in fresh BHI (1:4 culture:BHI; 5 mL total) to achieve growth to midlogarithmic phase (OD 600 ≈ 0.6–0.8). From this subcultured bacterial suspension, 1 mL aliquot was collected, spun down at high speed for 30 s at room temperature, washed three times in sterile PBS, and resuspended in 100 µL of sterile PBS.

Lm infection of mouse bone marrow-derived macrophages

Request a detailed protocol

About 6- to 8-week-old C57BL/six female mice were obtained from NCI/Charles River NCI facility and BMDM were made as previously described (Sauer et al., 2011). Briefly, macrophages were cultured from bone marrow in the presence of M-CSF derived from transfected 3T3 cell supernatant for 6 days, with an additional supplement of M-CSF medium 3 days postharvest. Cells were frozen down for storage. The day before infection, frozen cells were thawed and plated in 35 mm glass bottom dishes (MatTek, Ashland, MA) at 1.6 × 106 in 2.4 mL BMDM medium (Roswell Park Memorial Institute (RPMI) media containing 10% fetal bovine serum, 10% CSF, 1% sodium pyruvate, 1% glutamate, and 0.1% β-mercaptoethanol) and allowed to recover overnight at 37°C, 5% CO2. The following day 1.6 mL BMDM medium with or without Lm at MOI 2 was added to cells and incubated at 37°C and 5% CO2. After 30 min, cells were rinsed once with BMDM medium and replaced with 2.4 mL medium containing 0.25 mg/mL gentamicin (Lonza, Walkersville, MD). Cells were maintained at 37°C and 5% CO2 and imaged live at 5–6 hr post infection.

Fixation

Request a detailed protocol

Larvae were fixed in 1.5% formaldehyde (Polysciences, Warrington, PA) in 0.1 M pipes (Sigma-Aldrich), 1.0 mM MgSO4 (Sigma-Aldrich), and 2 mM ethylene glycol tetraacetic acid (EGTA) (Sigma-Aldrich) overnight at 4°C, with gentle rocking. Next day samples were rinsed with PBS at least once and stored in PBS at 4°C until imaging.

Counting total number of macrophages in whole larvae

Request a detailed protocol

To count macrophages, the nucleus was labeled using transgenic line Tg(mpeg1:h2b-GFP). Larvae at 3 dpf (not treated with PTU) were fixed overnight at 4°C using 1.5% fixative. The z-stack images of whole larvae were acquired at room temperature using Zeiss Zoomscope (EMS3/SyCoP3; Zeiss, Oberkochen, Germany; Plan-NeoFluar Z objective; 25× magnification, 1.7 μm resolution, 9.2 mm field of view, 60 μm depth of field; 10 μm steps in z direction) and Zen software (Zeiss). Sum projections were generated and macrophages were counted manually using open source image processing package, Fiji (Schindelin et al., 2012).

Zebrafish tail fin wounding

Request a detailed protocol

Simple tail fin transection, infected tail fin transection, and thermal injury of the tail fin were performed on 3 dpf larvae as described previously (Miskolci et al., 2019). In preparation for wounding, larvae were anesthetized in E3 medium containing 0.16 mg/mL tricaine (ethyl 3-aminobenzoate; Sigma-Aldrich). Wounding was performed in plastic tissue culture dishes that were coated with milk to prevent larvae sticking to the plastic. Simple tail transection of the caudal fin was performed using surgical blade (Feather, no. 10) at the boundary of and without injuring the notochord; following transection, larvae were rinsed with E3 medium to wash away tricaine, placed in fresh milk-coated dishes with fresh E3 medium, and maintained at 28.5°C until live imaging. For infected tail transection, larvae were placed in 5 mL E3 medium containing tricaine in 60 mm milk-treated dish; 100 µL unlabeled bacterial suspension in PBS, or 100 µL PBS for control uninfected wounding, was added to the E3 medium and swirled gently to achieve even distribution of bacteria; tail fin transection of larvae in control or infected E3 medium was performed as described above; larvae were immediately transferred to a horizontal orbital shaker and shaken for 30 min at 70–80 rpm; control and infected larvae were then rinsed five times with 5 mL E3 medium without tricaine to wash away bacteria and maintained at 28.5°C until live imaging; larvae were not treated with antibiotics at any point during the experiment. To perform thermal injury, fine tip (type E) of a line-powered thermal cautery instrument (Stoelting, Wood Dale, IL) was placed into the E3 medium, held to the posterior tip of the caudal fin, and turned on for 1–2 s until tail fin tissue curled up without injuring the notochord; following injury, larvae were rinsed with E3 medium to wash away tricaine, placed in fresh milk-coated dishes with fresh E3 medium, and maintained at 28.5°C until live imaging or fixation.

Drug treatments

Request a detailed protocol

2-DG (Sigma-Aldrich, St. Louis, MO) treatment was empirically optimized by testing different doses and length of pretreatment. 2-DG was freshly prepared for each experiment, dissolved at 100 mM in E3 medium without methylene blue (E3−); treatment was performed by bathing zebrafish larvae in E3− containing 5 mM 2-DG for 1 hr before imaging; larvae were kept in the presence of the inhibitor during live imaging. Metformin (Enzo, Farmingdale, NY) was freshly dissolved at 50 mM in E3−; treatment was performed by bathing dechorionated larvae in E3− containing 1 mM metformin starting at 1 dpf; metformin was refreshed daily (prepared fresh stock and dilution daily) until completion of live imaging or fixation.

Embedding zebrafish larvae for live imaging

Request a detailed protocol

Larvae were embedded in 1 mL 1% low gelling agarose (Sigma-Aldrich) prepared in E3− in Ibidi µ-slide 2-well glass bottom chamber (Ibidi, Fitchburg, WI) and topped off with 1 mL E3−. Agarose and top-off solution were supplemented with 0.16 mg/mL tricaine to keep larvae anesthetized during live imaging. In drug treatment experiments, agarose and top-off solution were supplement with drugs at working concentrations.

Wound healing assay

Request a detailed protocol

Wound healing assay was performed as previously (Miskolci et al., 2019). Larvae were wounded at 3 dpf as above and maintained at 28.5°C until fixation at indicated times. Fixed samples were placed in plastic milk-coated tissue culture dish in 0.1% Tween-20-PBS solution and tails were cut past the tip of the yolk sac so that tail fin tissue would lay flat. Single-plane brightfield images were acquired using Zeiss Zoomscope (EMS3/SyCoP3; Zeiss, Oberkochen, Germany; Plan-NeoFluar Z objective; 112× magnification, 0.7 μm resolution, 2.1 mm field of view, 9 μm depth of field) and Zen software (Zeiss). Tissue regrowth area as a measure of wound healing was quantified using Fiji by outlining the tail fin tissue area distal to the notochord using the polygon tool.

TNFα expression in macrophages

Request a detailed protocol

Analysis of TNFα expression in macrophages at zebrafish tail wound was performed as previously (Miskolci et al., 2019) using TNFα reporter line (Marjoram et al., 2015). Double transgenic line Tg(tnf:GFP x mpeg1:mCherry-CAAX) was wounded at 3 dpf as above and maintained at 28.5°C until fixation at indicated times. Fixed samples were placed in Ibidi µ-slide 2-well glass bottom chamber in 0.1% Tween-20-PBS solution and z-stacks at 3 μm steps and 512 × 512 resolution were acquired at room temperature using a spinning disk confocal microscope (CSU-X, Yokogawa, Sugar Land, TX) with a confocal scanhead on a Zeiss Observer Z.1 inverted microscope, EC Plan-Neofluar NA 0.3/10× objective, a Photometrics Evolve EMCCD camera and Zen software (Zeiss). TNFα expression was quantified by scoring macrophages at the wound for GFP signal (TNFα− in the absence of GFP signal or TNFα+ when any GFP signal was detected within a cell), or by area thresholding in Fiji as previously described (Miskolci et al., 2019), and expressed as proportion per larva.

Fluorescence lifetime imaging of NAD(P)H and FAD

Request a detailed protocol

All samples were imaged live using a two-photon fluorescence microscope (Ultima, Bruker) coupled to an inverted microscope body (TiE, Nikon), adapted for fluorescence lifetime acquisition with time correlated single photon counting electronics (SPC-150, Becker & Hickl, Berlin, Germany). A 40× (NA = 1.15) water immersion objective was used. An Insight DS+ (Spectra Physics) femtosecond source with dual emission provided light at 750 nm (average power: 1.4 mW) for NAD(P)H excitation and 1040 nm (average power: 2.1 mW; for stat6 mutant experiment the average power was 5 mW to compensate for an excitation intensity drop caused by an underfilled objective resulting from adjustments to achieve a more uniform mixed wavelength excitation over a larger field of view) for mCherry excitation. FAD excitation at 895 nm was achieved through wavelength mixing. Wavelength mixing was achieved by spatially and temporally overlapping two synchronized pulse trains at 750 nm and 1040 nm (Stringari et al., 2017). Bandpass filters were used to isolate light, with 466/40 nm used for NAD(P)H and 540/24 nm for FAD, and 650/45 for mCherry which were then detected by GaAsP photomultiplier tubes (H7422, Hamamatsu). Fluorescence lifetime decays of NAD(P)H, FAD, and mCherry were acquired simultaneously with 256 time bins across 256 × 256 pixel images within Prairie View (Bruker Fluorescence Microscopy) with a pixel dwell time of 4.6 µs and an integration time of 60 s (image acquisition time) at an optical zoom of 2.00. No change in the photon count rate was observed, ensuring that photobleaching did not occur. Images were acquired in a single plane. Following each acquisition, we moved to a different plane to avoid measuring the same cell twice; each data point in the graphs represents a different macrophage. The second harmonic generation obtained from urea crystals excited at 890 nm was used as the instrument response function and the full width at half maximum was measured to be 260 ps. BMDM were imaged live in MatTek dishes while maintained at 37°C and 5% CO2 using a stage top incubator system (Tokai Hit, Bala Cynwyd, PA). Zebrafish larvae were embedded in 1% low-gelling agarose and imaged live at room temperature; double transgenic larvae Tg(tnf:GFP × mpeg1:mCherry-CAAX) were PTU treated, all other imaging was done in albino background.

Fluorescence lifetime data analysis

Request a detailed protocol

Fluorescence lifetime components were computed in SPCImage v7.4 (Becker and Hickl). For each image, a threshold was selected to exclude background. The fluorescence lifetime components were then computed for each pixel by deconvolving the measured instrument response function and fitting the resulting exponential decay to a two-component model, I(t)=α1et/τ1+α2et/τ2+C, where I(t) is the fluorescence intensity at time t after the laser excitation pulse, α1 and α2 are the fractional contributions of the short and long lifetime components, respectively (i.e. α1+α2=1), τ1 and τ2 are the fluorescence lifetimes of the short and long lifetime components, respectively, and C accounts for background light. A two-component decay was used to represent the lifetimes of the free and bound configurations of NAD(P)H and FAD (Lakowicz et al., 1992; Nakashima et al., 1980). Images were analyzed at the single cell level. For the in vitro macrophages, cell cytoplasm masks were obtained using a custom CellProfiler pipeline (v.3.1.8) (McQuin et al., 2018). Briefly, the user manually outlined the nucleus of the cells and those masks were then propagated outward to find cell areas. Cytoplasm masks were then determined by subtracting the nucleus masks from the total cell area masks. Bacteria masks were created in Fiji by thresholding the mCherry intensity images into bacteria and background. The resulting bacteria masks were then subtracted from the corresponding field of view’s masks to exclude bacterial metabolic data. The diffuse cytoplasmic fluorescence in the mCherry images is likely due to FAD autofluorescence (Szulczewski et al., 2016). Images of the optical redox ratio (intensity of NAD(P)H divided by the sum of the intensity of NAD(P)H and the intensity of FAD) and the mean fluorescence lifetime (τm=α1τ1+α2τ2, where τ1 is the short lifetime for free NAD(P)H and bound FAD, τ2 is the long lifetime of bound NAD(P)H and free FAD, and α1 and α2 represent relative contributions from free and protein-bound NAD(P)H, respectively, and the converse for FAD) of NAD(P)H and FAD were calculated and autofluorescence imaging endpoints were averaged for all pixels within a cell cytoplasm using RStudio v. 1.2.1335 (Team R, 2015). For the in vivo macrophages, a custom CellProfiler pipeline segmented the macrophage cell area. Briefly, the pipeline rescaled the mCherry intensity images to be between 0 and 1 by dividing by the brightest pixel value in the image. Background was excluded by manually setting a threshold (0.15). Cells were identified using CellProfiler’s default object identification. Then, each cell was manually checked and edited as necessary to exclude background fluorescence and to include all pixels of each macrophage. Images of the optical redox ratio (intensity of NAD(P)H divided by the sum of the intensity of NAD(P)H and the intensity of FAD) and the mean fluorescence lifetime (τm=α1τ1+α2τ2; defined above) of NAD(P)H and FAD were calculated and autofluorescence imaging endpoints were averaged for all pixels within a cell using MATLAB v.9.7.01296695 (R2019b; Mathworks, Natick, MA). OMI index (Walsh et al., 2014; Walsh and Skala, 2015), the linear combination of mean-centered optical redox ratio, NAD(P)H τm, and FAD τm (coefficients of 1, 1, and –1) was calculated for each cell.

LC–MS based metabolite analysis

Request a detailed protocol

To analyze intracellular metabolites, metabolites were extracted from 80 to 100 cut tails with cold (on dry ice) LC–MS grade 80/20 methanol/H2O (v/v). Samples were dried under nitrogen flow and subsequently dissolved in LC–MS grade water for LC–MS analysis methods. Protein pellets were removed by centrifugation. Samples were analyzed using a Thermo Q-Exactive mass spectrometer coupled to a Vanquish Horizon ultra-high performance liquid chromatograph. Metabolites were separated on a C18 (details below) at a 0.2 ml per min flow rate and 30°C column temperature. Data was collected on full scan mode at a resolution of 70 K. Samples were loaded in water and separated on a 2.1 × 100 mm, 1.7 μM Acquity UPLC BEH C18 Column (Waters) with a gradient of solvent A (97/3 H2O/methanol, 10 mM TBA, 9 mM acetate, pH 8.2) and solvent B (100% methanol). The gradient was: 0 min, 5% B; 2.5 min, 5% B; 17 min, 95% B; 21 min, 95% B; 21.5 min, 5% B. Data were collected on a full scan negative mode. The identification of metabolites reported was based on exact m/z and retention times, which were determined with chemical standards. Data were analyzed with Maven. Relative metabolite levels were normalized to protein content.

Statistical analyses

Request a detailed protocol

Biological repeats are defined as separate clutches of embryos collected on separate days. Statistical significance was set to 0.05. Statistical analyses of autofluorescence imaging data were performed using R v.3.6.2 (https://www.R-project.org) (R Core Team, 2019). General linear models with gaussian errors were fit to data, where every data point represented a macrophage. For zebrafish experiments, models included day (biological repeat) as a blocking factor and indicators for experimental treatments or conditions. An interaction was included in models where more than one experimental factor was present (e.g. time and treatment), to determine whether effects associated with one experimental factor modified the other. All models utilized cluster-robust standard errors to account for multiple macrophages being measured within the same larvae. Model assumptions (normality of errors and constant error variance) were assessed by checking residuals with normal quantile/percentile plots and inspecting residuals versus fitted values for constant variance. Log transformation was applied to certain lifetime endpoints when these model checks revealed an issue. It is indicated in the figure legend if log transformation was applied. No adjustment for multiplicity was done. Graphical displays were generated in Python using the open-source graphing package Matplotlib (https://matplotlib.org/). Each data point in the graphical display represents a macrophage and the data for each condition is presented as a composite dotplot and boxplot; each data point is a different macrophage; each biological repeat is displayed by a different color in the dotplot; boxplots show median (central line), first, and third quartiles (lower and upper lines), and the Tukey method was employed to create the whiskers (the farthest data points that are no further than 1.5 times the interquartile range); data points beyond whiskers are considered outliers. Statistical conclusions (p values) apply to the overall effects when comparing the groups and also account for the cluster-correlated structure of data as described above; this is denoted by italicizing the p values. Estimated means with 95% CI and comparison between groups showing the overall effect computed as fold change (ratio) or simple difference with 95% CI and p values are provided separately in supplemental source data files. Proportion of TNFα+ macrophages and total macrophages at the wound, total number of macrophages in whole larvae, and tissue regrowth area to measure wound healing in zebrafish larvae were analyzed in SAS/STAT 9.4 (SAS Institute Inc, Cary, NC) using a linear mixed model with treatment or genotype as the experimental effect, and replicate day as the random effect, with an interaction included when time was a second experimental factor. Normality and residuals were assessed and no transformations were required. Tukey adjustment for pairwise comparisons was utilized when more than two treatments were compared (i.e. wild-type, heterozygote, and homozyote mutant genotypes). Data were graphed using Prism (GraphPad Software, Inc, San Diego, CA); each data point represents a larva and arithmetic mean with 95% CI is shown in the graphical displays.

Code availability

Request a detailed protocol

All codes used for image and statistical analyses are deposited at https://github.com/skalalab/zebrafish_flim (copy archived at swh:1:rev:808637b36531d833e61c9495cb3aac4ae3723df0; Miskolci, 2022).

Data availability

Source data files containing numerical data used to generate the graphical displays are provided for all figures.

References

  1. Software
    1. R Core Team
    (2019)
    R: A langauge and environment for statistical computing
    R Foundation for Statistical Computing, Vienna, Austria.
  2. Software
    1. Team R
    (2015)
    RStudio: Integrated Development for R
    RStudio, Inc.

Decision letter

  1. Serge Mostowy
    Reviewing Editor; London School of Hygiene & Tropical Medicine, United Kingdom
  2. Didier YR Stainier
    Senior Editor; Max Planck Institute for Heart and Lung Research, Germany
  3. Robert Knight
    Reviewer; Kings College London, United Kingdom

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "In situ fluorescence lifetime imaging of macrophage intracellular metabolism during wound repair in zebrafish" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Didier Stainier as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Robert Knight (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) The authors use label free imaging methods to measure key components of immune cell metabolism. FLIM data presented appears promising, but alternative methods to support results / interpretations is lacking throughout this study.

2) Various key results appear unexpected / contradictory (for example macrophage results obtained in vitro versus in vivo (Figure 1 vs Figure 4); metabolomic monitoring of macrophages alone versus entire tissue (Figure 5))? To promote confidence in your new approach, support contradictory results with new data, alternative approaches, and/or more elaborate justification.

3) Where highlighted by the reviewers, please explain / justify technical and experimental details in depth.

Reviewer #1 (Recommendations for the authors):

I am enthusiastic about the imaging approach and its potential, but not yet convinced fluorescence lifetime imaging of macrophages in situ is reliable.

1. It is crucial to better understand why results in vitro do not mimic results in vivo. In the case of Listeria infection, controls can be performed. For example:

– Confirm that Listeria is intracellular during fluorescence lifetime imaging of macrophages (in vitro and in vivo).

Listeria is well known to invade epithelial cells. Could this influence results described in vivo?

– If to perform fluorescence lifetime imaging of Listeria infected HeLa cells, would these results mimic BMDM infection in vitro or infection in vivo?

2. Metabolomics of whole tail fin tissue (not macrophages alone)

It is surprising that mass spec of whole tail fin tissue gives similar result as for imaging of single macrophages. Considering these results suggest a whole tissue response (dominated by non-macrophages), what is the significance of monitoring macrophages only (if they behave the same as the rest of the tissue)?

3. Measure / manipulate classic markers of inflammation, eg mROS and cytokines (TNF), in parallel to fluorescence lifetime measurements to help build confidence in these new results.

4. Are fluorescence lifetime measurements for non-motile cells directly comparable to those from motile cells?

5. Line 162. 'We found that GFP is suitable to image in conjunction with NAD(P)H, but it excludes the acquisition of FAD as they have overlapping spectra. However, we found mCherry to be compatible for simultaneous imaging with NAD(P)H and FAD. We also optimized imaging on live larvae.'

-Please show these data for readers to follow how conclusions are made

6. Lines 227-229. In the previous work Miskolci et al. 2019, macrophage recruitment, TNFa+ cells and Listeria-infected cells peak between 72 and 96hpw. However in the present work, authors only analysed 48hpw. Why?

7. Figures

– For all images: Additionally provide inset image to zoom in on a single cell to better illustrate changes with this new technique.

– In the case of Figure 1, metabolomic changes of bacteria were excluded – why is this not done for all figures involving infection?

– Statistical information can be consistently presented across all figures (eg Figure 1 directly on plots; Figure 2 presented in Table).

8. Figures 3,4,5: Understandably the authors have counted cells in the same number of embryos. However, would it be more informative to count similar number of cells for the different conditions?

Reviewer #3 (Recommendations for the authors):

The approach used is novel and could potentially be helpful for researchers using zebrafish to measure cell metabolism. My major criticism is that the interpretation of the intensity and fluorescence lifetime data is simplistic and does not account for the complexity of NADH/NADPH bioavailability in cells. It is also very difficult to relate such measures to cellular metabolism without some additional method for confirming this.

The paper was somewhat hard to read, with technical details insufficiently explained and results only cursorily described or described using indirect language without specifically saying what was observed (e.g. a similar trend was observed as seen previously…). The manuscript would have benefited from much clearer descriptions of the actual data obtained and brief but to the point interpretations of these results.

On a technical note, I was not quite clear about how long it took to acquire sufficient photon counts from the zebrafish larvae – I note in the methods (pg. 17, line 562) that images acquired were 256 x 256 pixels and acquisition speed was 4.6 µs per pixel with 256 time bins. As macrophages are relatively motile I wondered how precise this measure of fluorescence lifetime was for each macrophage in these experiments and would like to have a clearer description of the acquisition process in the text to clarify this point.

I was also quite surprised that there was no mention of other methods used to measure NADH/ NADPH activity in zebrafish such as the iNap biosensor described by Tao et al. (doi: 10.1038/nmeth.4306). How do the measures of NADH/NADPH by intensity measures compare to such biosensors? It would be important to know how comparable FLIM measures of NAD(P)H are to such a biosensor to enable other researchers to make an informed choice as to whether FLIM is sensitive enough to resolve biologically meaningful differences in NAD(P)H bioavailability.

Specific comments are as follows:

pg. 2: the abstract is rather opaque and would benefit from some rewriting to clarify certain statements e.g. line 46 'FLIM also resolved temporal changes in the optical redox ratio and lifetime variables of NAD(P)H..'. There is no explanation of what these lifetime variables are making this sentence hard to understand.

pg. 5 (line 143): it would be helpful to clarify that initially the authors have obtained an intensity based measure of oxidation/ reduction to measure the optical redox ratio and they also obtain lifetime measures for the different molecules. The introduction implies they are only aiming to measure lifetime so it is not obvious this is not lifetime that is initially described.

pg. 5 (line 148) and pg. 26 (line 853): When describing calculations for the optical redox ratio in mouse macrophage infected by Listeria the authors appear to have adjusted this relative to expression of mCherry in bacteria: 'mCherry expression in bacteria was used to subtract from lifetime images in order to exclude metabolic changes in the pathogen from that of macrophages.' I assume this means they used the mCherry as a mask to remove the lifetime signals from bacteria expressing mCherry but it hard to understand and needs to be clarified and explained.

pg. 5 (line 151): I could not find a description for FLIM measures for the in vitro macrophage culture experiments and noticed the authors only briefly commented on the changes to fluorescence lifetime that were observed. The emphasis of the paper is that FLIM is a powerful method to identify changes to the redox state of the cell so I would consider it important to address this in the text and describe what the changes were.

pg. 5 (line 163): the authors cite unpublished data that argues mCherry emissions are compatible with FLIM imaging of NAD(P)H and FAD+ but do not show any evidence for this. This statement should be justified.

pg. 6 (line 175): Calculations to measure changes to NAD(P)H fluorescence lifetime are performed by using an average of the short and long fluorescence lifetime signals NAD(P)H. This does not account for the known differences of the long lifetime (tau2) component of NADH when it is bound to different proteins. It would be helpful to know how the authors have accounted for this potential source of error when calculating the proposed FLIM redox ratio using a mean fluorescence lifetime value of NAD(P)H.

pg. 6 (line 177): how many macrophages were measured in total to obtain the relative alpha and tau values for fluorescence lifetime shown in figure 2? And were the same macrophages measured at multiple times or are the datapoint shown in Figure 2 representative of different macrophages? These results should be more fully described to allow the reader to understand how much sampling was performed when measuring FLIM.

pg. 7 (line 223): when measuring FAD+ and NAD(P)H fluorescence lifetime the authors describe a difference in the relative proportion of signals in the alpha1 fractional component of free NAD(P)H between infected and uninfected animals (shown in Figure 4) compared to the alpha1 fractional component measured when measuring NAD(P)H alone in animals expressing GFP (shown in Figure 3). It is difficult to discriminate this difference when comparing Figure 3 and 4 supplements as the significance of the difference between control and infected animals for the NAD(P)H alpha1 fraction is only shown in Figure 4 supplement. Comparable data should be shown for both sets of data to allow a comparison of the alpha1 component between them.

pg. 8 (line 250): it is difficult to discern from the text which measures show significant differences in NAD(P)H and FAD+ between burn compared to transection injury models. It would be helpful in this instance to state whether the lifetime measures are increased or decreased and how these differences reflect the redox state. As an example, the authors state that 'The trends for the differences in the mean lifetime and individual lifetime components of NAD(P)H and FAD between the burn wound and simple transection at 24 hpw were comparable to the observed differences between the infected and simple transection (Figure 5C, D, Figure S5A-F).' This does not provide the reader with detail as to what is actually different and should clearly state what such differences are.

pg. 8 (line 254): The authors state that there is a significant difference for the NAD(P)H lifetime in macrophages in the tail of burn compared to transection injury models. This is calculated from the relative contributions of tau1 and tau2 to the mean fluorescence lifetime of NAD(P)H (Figure 5 Supplement). What was not discussed is that tau1 for NAD(P)H clearly shows a difference between sterile injuries (306, 95% CI: 289-325) and burn injuries (274, 95% CI: 264-284) at 24 hpi but tau2 is almost identical between these two models. Is this change to the tau1 component important for understanding differences in the mean NAD(P)H lifetime between the two models at 24 hpi and not the tau2 component?

pg. 9 (line 303): the authors describe changes to NAD(P)H lifetimes in animals treated with Metformin and state 'the changes in NAD(P)H tau1 and tau2 also trended as expected'. It is not clear what should be expected and how these trended. The authors should clarify this statement to state what the results actually showed.

pg. 19 (line 624): the authors state that a general linear model was used to test for significant changes. How was the data fitted to the models, what were the models used, with which assumptions were the models generated and where are the results from the models? I would expect this to be made more explicit so this work could be examined critically, but the level of detail of the statistical approaches used is not sufficient to enable this and should be addressed.

pgs. 36, 38, 40.

Supplemental figures 2, 3, 4: what do the R1, R2 etc denote?

Data availability and statistical methodology:

It was not possible to comment on the statistical methods used as there was insufficient detail provided in the manuscript. It was also not possible to discern the exact details for how the acquisition of FLIM data was performed including duration of scanning for each cell and whether a cell was examined more than once.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "in vivo fluorescence lifetime imaging of macrophage intracellular metabolism during wound responses in zebrafish" for further consideration by eLife. Your revised article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Didier Stainier as the Senior Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below. In particular, 2 issues raised by the Reviewers which require further revision of the text/analysis of existing data are:

1)The authors provided a technical information in the rebuttal on how images were segmented in the TNFa expression data (previous Figure 3 current Figure 2). However I am not sure they answered my question. My point was that in the fluorescence images they show dotted outlines of macrophages which are supposed to be derived from segmentation based on cherry intensity. However, the cherry intensity within those outlines is not uniformly higher than outside the outlines – there are dark/low-intensity areas within the outlines. Conversely there are areas of high-intensity which are not included in the outlines. So it still puzzles me how these outlines are deduced because if segmentation is threshold-based the areas inside the outlines should all have intensities above a certain common threshold. Perhaps the authors need to change the brightness/contrast of the image and/or show us how they threshold the image, to validate how these cellular outlines are deduced.

2)The authors need to acknowledge that they are looking at both NADPH/NADP AND also NADH/NAD with the methods they have used. These two redox molecule pairs have quite different functions in the cytoplasm and mitochondria so it is important to recognise this point. Saying that, their methodology for discriminating oxidative vs reduced status in a cell is OK despite this but it makes it difficult to be specific about which metabolic process is likely to be occurring based on these measures. Perhaps they could update the manuscript to acknowledge/ discuss.

Reviewer #1 (Recommendations for the authors):

The authors have delivered a significantly revised manuscript in response to reviewer comments, where new results strongly support the use of fluorescence lifetime imaging to characterise macrophage metabolism in situ. Many interesting future directions emerge from this innovative work.

I am highly positive about this revised manuscript.

Reviewer #2 (Recommendations for the authors):

The authors have performed a substantial revision in response to the reviews. In response to my comments they:

– Added data that indicate the biological significance of the optical redox measurements. Using STAT6 depletion or Metformin treatment they shifted the macrophage differentiation status towards M1 or M2 and correlated this with changes in optical redox ratio measurements. The treatments led to changes in tissue repair process, indicating that the optical redox ratio measurements may have predictive value in regeneration outcome after different treatments.

– Added new references to explain limitations in choice of fluorescent reporters that can be used for this method.

– Added a comment in the paper regarding discrepancies between in vitro and in vivo data. Given this was raised by two reviewers, they could provide a stronger commentary, for example by indicating other relevant examples in the literature where in vitro and in vivo observations on cell phenotypes don’t match.

– Provided potential reasons in the rebuttal for the low magnitude of the metabolic changes observed in vitro with their system. These reasons could be also included in the manuscript in brief form.

– Provided a technical information in the rebuttal on how images were segmented in the TNFa expression data (previous Figure 3 current Figure 2). Here I am not sure they answered my question. My point was that in the fluorescence images they show dotted outlines of macrophages which are supposed to be derived from segmentation based on cherry intensity. However, the cherry intensity within those outlines is not uniformly higher than outside the outlines – there are dark/low-intensity areas within the outlines. Conversely there are areas of high-intensity which are not included in the outlines. So it still puzzles me how these outlines are deduced because if segmentation is threshold-based the areas inside the outlines should all have intensities above a certain common threshold. Perhaps the authors need to change the brightness/contrast of the image and/or show us how they threshold the image, to explain more clearly how these cellular outlines are deduced.

– Added a new Table 1 with better explanation of the interpretations of the metrics.

-explained choice of time point for measurements.

– Improved the abstract.

– Improved the discussion. I still found paragraph in lines 349-368 difficult to follow. I am not really sure what the reader should take from it. This paragraph should probably move to supplementary information, as the role seems to be to provide more technical information for some specialists.

Altogether they have generally addressed my concerns but I would just recommend some straightforward edits as indicated in the specific points above.

Reviewer #3 (Recommendations for the authors):

I thought the authors have addressed most of the points raised well and thoughtfully. The manuscript is much more readable and results are now clear. I think this is a valuable and helpful addition to the field and recommend publication.

I have a few additional comments/ questions that I think would strengthen the manuscript further:

1. I wondered about the declaration on line 277 that 'The modest changes in NAD(P)H lifetime endpoints are likely attributed to the modest decrease in TNFα expression in macrophages upon Metformin treatment'.

This implies that TNF-a regulates the metabolic status of the cell. Is this intended? It would be helpful to therefore clarify the potential relationship between NAD(P)H t1 and t2 and TNF-a expression.

2. how does their study correspond to the descriptions of the metabolic profiling of macrophages during regeneration described by Ratnayake et al., 2019? In this work the authors describe a role for the NAD+ regenerating enzyme Nampt in promoting regeneration through regulating NAD+ replenishment in macrophages. By measuring the NADH/ NAD+ ratios by FLIM they argue the pro-regenerative macrophages display a glycolytic phenotype. It would be helpful to compare results from this work to this study.

3. The major thrust of this work is that it is appropriate to use the ratio of NAD(P)H and FAD to approximate the relative oxidative state of a cell. The two methods used to measure NAD(P)H – intensity measures of autofluorescence and relative changes to fluorescence lifetime – do discriminate between NAD(P)H and NAD(P). However the emission wavelength of the equally important redox molecule NADH is identical to NAD(P)H so any intensity measure will include both. It is important to consider this caveat as changes to NADH levels will affect any such intensity based measures. I wondered therefore if the changes to the optical ratio in various manipulations might not always be correlated to changes of NAD(P)H tau1 and NAD(P)H tau2 due to this confounding factor? It would be greatly helpful to introduce the importance of NADH/ NAD in metabolic signalling and discuss whether the optical ratio utilised in this study would also provide an insight to changes of these critical molecules.

4. I commented on the caveat that the lifetime for bound NAD(P)H can change depending on the protein it is bound to. The authors have made the data for all measured lifetimes available and discussed it in the Discussion section (line 377), but after reading this I was left wondering what the reader could definitively determine from this measure. The last two sentences of this section in the Discussion are intriguing as they imply changes to tau1 and tau2 could be used to infer changes to different metabolic processes. Is it possible to present this with clearer predictions for what changes to tau1 and tau2 might mean in relation to the metabolism of the cells?

https://doi.org/10.7554/eLife.66080.sa1

Author response

Reviewer #1 (Recommendations for the authors):

I am enthusiastic about the imaging approach and its potential, but not yet convinced fluorescence lifetime imaging of macrophages in situ is reliable.

We include new data (Figure 6) and analysis that provide more convincing evidence that FLIM is a reliable way to image macrophages in situ. 1. It is crucial to better understand why results in vitro do not mimic results in vivo. In the case of Listeria infection, controls can be performed. For example:

We agree with the reviewer that is an unexpected observation. Our approach in the revision was to provide even more convincing evidence that our in situ results accurately report metabolic activity of macrophages. We have moved the in vitro results (which displayed a minor phenotype) to the supplemental data (Figure S3I-K). We include robust controls of the in vivo analysis where we include 2-DG, metformin and depletion of STAT6 that impact metabolism and are consistent with our findings with TNFa+ macrophages at the infected tail wound (Figure 2, 3) and burn wound (Figure 4). We also include new functional data on the impact of these changes on TNFa+ macrophages and regeneration of the fin. Since our submission, there is new publication (PMID: 34518542) that support our findings in zebrafish larvae.

– Confirm that Listeria is intracellular during fluorescence lifetime imaging of macrophages (in vitro and in vivo).

We provide confirmation that Listeria is intracellular during in vitro infection. For the in vitro experiments Listeria was labeled with mCherry, see Figure S3I. We also mention in the text that the mCherry label allowed us to subtract Listeria from the images during image analysis, so that we measure and report macrophage metabolism only. For in vivo Listeria infection, we also have evidence that Listeria is intracellular in zebrafish larvae; please see Figure 3—figure supplement 2B in our previous publication (PMID: 31259685) showing that Listeria co-localize with macrophages in an infected wound. We also include image set in Figure S2E of this current manuscript. Regarding labeling Listeria for the in vivo FLIM experiments, we were limited by the colors available for imaging and it was not feasible. There are 3 PMTs available for detection in this system, so at most we can only detect in 3 channels; one for macrophages, one for FAD and one for NAD(P)H. For the in vivo experiments we needed a label for macrophages so that we could segment them from the field of view during data analysis (this is not an issue for in vitro experiments).

– Listeria is well known to invade epithelial cells. Could this influence results described in vivo?

Yes, it is a possible reason why in vitro results are different from in vivo. in vitro measurements where macrophages are in isolation do not take into the account the input from multiplex soluble signals and other interacting cells in the microenvironment, which in vivo measurements do. This underscores the reason why we set out to test FLIM as a tool to study single cell-based intracellular metabolism in vivo, to be able to address these fundamental differences between in vitro and in vivo settings. Therefore, in the revision we improved the analysis and treatments to detect how changes in metabolism alter macrophage FLIM in tissues. We think our findings are robust and not surprisingly suggest that metabolism of macrophages is different in vitro and in vivo.

– If to perform fluorescence lifetime imaging of Listeria infected HeLa cells, would these results mimic BMDM infection in vitro or infection in vivo?

That is an interesting biological question, but this does not address the in vitro versus in vivo problem and is beyond the scope of the current paper.

2. Metabolomics of whole tail fin tissue (not macrophages alone)

It is surprising that mass spec of whole tail fin tissue gives similar result as for imaging of single macrophages. Considering these results suggest a whole tissue response (dominated by non-macrophages), what is the significance of monitoring macrophages only (if they behave the same as the rest of the tissue)?

We were also surprised that the tail fin tissue metabolomics was consistent with imaging single macrophages. However, we think this finding is informative and could not have been discovered without the single cell in vivo analysis. In addition, mass spec analysis destroys the samples and is not compatible with obtaining temporal and spatial information in the same animal. Using FLIM, we can measure live samples, in situ, and obtain temporal and spatial information. Furthermore, FLIM also provides single cell-based information, which can then be used to analyze heterogeneity in the measured population. The Skala lab is developing methods to analyze population heterogeneity (PMID: 31078067, PMID: 32719514, PMID: 31737571), and we plan to apply these tools in future studies.

3. Measure / manipulate classic markers of inflammation, eg mROS and cytokines (TNF), in parallel to fluorescence lifetime measurements to help build confidence in these new results.

We imaged TNF reporter and NAD(P)H lifetime in the same cells (Figure 2A). We also treated zebrafish with Metformin to reduce the proportion of TNFa+ macrophages at the burn wound (Figure 5). Metformin is known to promote M2 polarization, and we showed previously that it reduces the number of TNFa+ macrophages (PMID: 30572006). As expected, we observed an increase in redox ratio in the Metformin treated larvae. We also observed improved tissue repair (Figure 6G, H). To further validate the correlation between macrophage population and FLIM endpoints, we did the opposite experiment, where we attempted to increase the proportion of TNFa+ macrophages. We used the recently published Stat6 zebrafish mutant (PMID: 33761328). Stat6 is required for M2 macrophage polarization (PMID: 27813830), so we expected that deletion of Stat6 would result in an increase in the proportion of TNFa+ macrophages at the burn wound. We confirmed this in revised Figure 6A, B. Based on the increase in TNFa+ macrophages, we predicted that wound healing would be worse compared to wild-type, and indeed we found a decrease in wound healing (revised Figure 6I, J). We also observed a decrease in redox ratio and mean lifetime of NAD(P)H (Figure 6D, F).

4. Are fluorescence lifetime measurements for non-motile cells directly comparable to those from motile cells?

This is also an interesting biological question. However, we did not test it. M1 and M2 macrophage have different metabolism and different motility characteristics (PMID: 28458726), M1 macrophages are less motile compared to M2 cells. Therefore, we expect that non-motile cells would have different intracellular metabolism compared to motile cells.

5. Line 162. 'We found that GFP is suitable to image in conjunction with NAD(P)H, but it excludes the acquisition of FAD as they have overlapping spectra. However, we found mCherry to be compatible for simultaneous imaging with NAD(P)H and FAD. We also optimized imaging on live larvae.'

-Please show these data for readers to follow how conclusions are made.

Overlap between GFP and FAD is well documented in the literature, see review (PMID: 32406215), as is separation between NAD(P)H and GFP (PMID: 34321477). Our prior published studies and those of other groups have also shown that mCherry fluorescence can be separated from NAD(P)H and FAD autofluorescence (PMID: 33959597; PMID: 32509583). We have added these citations so that these methods can be replicated.

6. Lines 227-229. In the previous work Miskolci et al. 2019, macrophage recruitment, TNFa+ cells and Listeria-infected cells peak between 72 and 96hpw. However in the present work, authors only analysed 48hpw. Why?

We picked this time point based on the proportion of TNFa+ macrophages at the infected wound. We wanted to pick a time point where the macrophage populations as measured by TNFa expression would be maximally different between simple transection and the infected wound. At the infected wound, the number of macrophages peak around 72-96 hpw, however the proportion of TNFa+ macrophages stays relatively similar past 48 hpw, ~75-80%. Based on this, we selected 48 hpw. At 48 hpw, most macrophages at the simple transection wound are TNFa-, and most macrophages at the infected wound are TNFa+, so we felt this time point represented the maximal differences in macrophage populations between the 2 wound types, sufficient enough that we should detect differences by FLIM.

7. Figures

– For all images: Additionally provide inset image to zoom in on a single cell to better illustrate changes with this new technique.

We opted not to include zoomed insets as FLIM image resolution is low (256x256), and the zoomed-in images would be pixelated.

– In the case of Figure 1, metabolomic changes of bacteria were excluded – why is this not done for all figures involving infection?

See answer to question 1; we cannot image 4 channels/PMTs and also image FAD and NAD(P)H. This is why we used a sterile burn wound without infection. However, given that macrophages at the burn wound had reduced redox ratio and the endpoints of NAD(P)H trended similarly as at the infected wound (see Table 2), the burn wound provides a simpler inflammatory damage model for FLIM.

– Statistical information can be consistently presented across all figures (eg Figure 1 direclty on plots; Figure 2 presented in Table).

We have revised the statistical information to address these concerns.

To reduce confusion, we removed the tables from the Figures, and created a excel file (Table 2S) summarizing the mean+/-95%CI for each endpoint, overall effects (fold change or simple difference) between groups and corresponding p values for every Figure.

8. Figures 3,4,5: Understandably the authors have counted cells in the same number of embryos. However, would it be more informative to count similar number of cells for the different conditions?

The differences in the number of cells are due to the nature of the wound; the simple transection is the least inflammatory compared to the infected or burn wounds, so that wound recruits fewer cells. We based the analysis on the advice of our statistician.

Reviewer #3 (Recommendations for the authors):

The approach used is novel and could potentially be helpful for researchers using zebrafish to measure cell metabolism. My major criticism is that the interpretation of the intensity and fluorescence lifetime data is simplistic and does not account for the complexity of NADH/NADPH bioavailability in cells. It is also very difficult to relate such measures to cellular metabolism without some additional method for confirming this.

We agree with the reviewer. In this manuscript we focused on correlating the changes in the redox ratio and lifetimes with the macrophage population; this is also why we did the Metformin experiment in Figure 5, where we modulated the macrophage population and expected changes in the redox ratio and lifetimes based on the trends established at the infected and burn wounds (Table 2). But yes, we have not correlated the changes with a specific metabolic pathway; correlating changes in autofluorescence measurements with specific metabolic pathway is a current challenge and the focus of the field itself. However, the 2DG experiment in Figure 1 provides some hints; inhibiting glycolysis has the same effect on the redox ratio and NAD(P)H lifetime endpoints (tm, t1, t2 and a1) in macrophages as a pro-inflammatory M1 polarization (more TNFa+ cells), as summarized in Table 2. This suggests that M1-like (TNFa+) macrophages have reduced glycolysis. However, making correlations to specific metabolic changes will be the goal of future studies. We are also developing additional reporter lines, so we can also better characterize macrophage population and correlate them with FLIM measurements. Currently there is only M1 reporter line available (TNF reporter) for live imaging, and we are developing M2 reporter lines. We could potentially perform immunostaining for M2 markers, however we are performing the FLIM on live zebrafish.

The optical redox ratio measures the oxidation-reduction state of the cell, while the NAD(P)H and FAD mean lifetimes report on protein-binding activities, proximity to quenchers, and microenvironmental factors [PMID: 32406215]. These autofluorescence endpoints have been correlated with traditional measures of metabolism including oxygen consumption [PMID: 27300321] and metabolite levels [PMID: 24305550] but provide unique sources of contrast. Unlike traditional measurements, these autofluorescence imaging endpoints can monitor single cells over time within intact samples, which provides important context for in vivo metabolic changes. This non-destructive single-cell insight is advantageous even with the caveat of multiple sources of contrast. Future studies will continue to define these sources of contrast as the technology develops.

The paper was somewhat hard to read, with technical details insufficiently explained and results only cursorily described or described using indirect language without specifically saying what was observed (e.g. a similar trend was observed as seen previously…). The manuscript would have benefited from much clearer descriptions of the actual data obtained and brief but to the point interpretations of these results.

We have revised the manuscript as requested by the reviewer.

On a technical note, I was not quite clear about how long it took to acquire sufficient photon counts from the zebrafish larvae – I note in the methods (pg. 17, line 562) that images acquired were 256 x 256 pixels and acquisition speed was 4.6 µs per pixel with 256 time bins. As macrophages are relatively motile I wondered how precise this measure of fluorescence lifetime was for each macrophage in these experiments and would like to have a clearer description of the acquisition process in the text to clarify this point.

We have clarified this issue in the text. The duration of integration time was 60 seconds and was selected based on the minimum time needed to collect sufficient number of photons to allow for the reliable fitting of the resulting exponential decay to a two-component model in SPCImage, taking into account that the cells are live and mobile and that we wanted to minimize photobleaching. This acquisition time should not be affected by macrophage motility.

I was also quite surprised that there was no mention of other methods used to measure NADH/ NADPH activity in zebrafish such as the iNap biosensor described by Tao et al. (doi: 10.1038/nmeth.4306). How do the measures of NADH/NADPH by intensity measures compare to such biosensors? It would be important to know how comparable FLIM measures of NAD(P)H are to such a biosensor to enable other researchers to make an informed choice as to whether FLIM is sensitive enough to resolve biologically meaningful differences in NAD(P)H bioavailability.

Thank you for pointing out this biosensor. We agree with the reviewer it would be very useful to make these comparisons. We plan to do this in future, but because of technical issues it is beyond the scope of the current manuscript. We would need a BFP-tagged macrophage reporter line to be compatible with the iNap biosensor. However, we have been unable to generate such line after multiple attempts; we believe it is too dim when driven from the macrophage promoter. We are in the process of obtaining mTagBFP2, that is approximately 5 times brighter compared to its parental mTagBFP. Alternatively, far red options can be pursued, however success with BFP is more likely.

Specific comments are as follows:

pg. 2: the abstract is rather opaque and would benefit from some rewriting to clarify certain statements e.g. line 46 'FLIM also resolved temporal changes in the optical redox ratio and lifetime variables of NAD(P)H..'. There is no explanation of what these lifetime variables are making this sentence hard to understand.

Lifetime variables refer to the individual endpoints we calculate and then use to derive the mean lifetime (tm) of the NAD(P)H and FAD; the individual endpoints are t1, t2, α1 and α2 and tm = t1α1 + t1α2. We simplified this sentence to make it less confusing to the reader.

pg. 5 (line 143): it would be helpful to clarify that initially the authors have obtained an intensity based measure of oxidation/ reduction to measure the optical redox ratio and they also obtain lifetime measures for the different molecules. The introduction implies they are only aiming to measure lifetime so it is not obvious this is not lifetime that is initially described.

We improved the text to clarify this issue.

pg. 5 (line 148) and pg. 26 (line 853): When describing calculations for the optical redox ratio in mouse macrophage infected by Listeria the authors appear to have adjusted this relative to expression of mCherry in bacteria: 'mCherry expression in bacteria was used to subtract from lifetime images in order to exclude metabolic changes in the pathogen from that of macrophages.' I assume this means they used the mCherry as a mask to remove the lifetime signals from bacteria expressing mCherry but it hard to understand and needs to be clarified and explained.

Yes, this assumption is correct, and we clarified it in the text.

pg. 5 (line 151): I could not find a description for FLIM measures for the in vitro macrophage culture experiments and noticed the authors only briefly commented on the changes to fluorescence lifetime that were observed. The emphasis of the paper is that FLIM is a powerful method to identify changes to the redox state of the cell so I would consider it important to address this in the text and describe what the changes were.

The optical redox ratio measures the oxidation-reduction state of the cell, while lifetime primarily reflects protein-binding activity of the coenzymes. We minimized the discussion of the changes in lifetime measurements per se for the in vitro data, because we felt that the only data in this set of experiments that we could compare to the mass spec results of the cited study (Gillmaier 2012, PMID: 23285016) was the optical redox ratio. Gillmaier et al. showed that Listeria infection of murine bone marrow derived macrophages was associated with increased glycolytic activity. We found a slight, but significant increase in the redox ratio, which is consistent with an increase in glycolysis.

pg. 5 (line 163): the authors cite unpublished data that argues mCherry emissions are compatible with FLIM imaging of NAD(P)H and FAD+ but do not show any evidence for this. This statement should be justified.

Several prior studies have shown that mCherry fluorescence does not corrupt the emission from NAD(P)H and FAD. We have cited two of these papers for reference [PMID: 33959597; PMID: 32509583]. We provide a representative phasor plot (Author response image 1), where the mCherry lifetimes are plotted in red and are clearly separated from the NAD(P)H (blue) and FAD (green) fluorescence lifetimes.

Author response image 1

pg. 6 (line 175): Calculations to measure changes to NAD(P)H fluorescence lifetime are performed by using an average of the short and long fluorescence lifetime signals NAD(P)H. This does not account for the known differences of the long lifetime (tau2) component of NADH when it is bound to different proteins. It would be helpful to know how the authors have accounted for this potential source of error when calculating the proposed FLIM redox ratio using a mean fluorescence lifetime value of NAD(P)H.

Our optical redox ratio is calculated from the intensity of NAD(P)H divided by the sum of intensities from NAD(P)H and FAD. This optical redox ratio is calculated from the sum of all photons in the lifetime decay, so it is not a FLIM redox ratio but the traditional intensity redox ratio measured in photon counting mode. The mean lifetime (τm) is the weighted average of the short and long lifetimes of NAD(P)H (τm = α1τ1 + α2τ2). This mean lifetime is reported separately from the optical redox ratio, which is an intensity measurement that is not calculated from the mean lifetime of NAD(P)H. We also report individual lifetime components in the supplemental graphs for both NAD(P)H and FAD (α1, τ1, τ2). Therefore, all the data is available for interpretation, so that changes in the long lifetime can be evaluated separately from changes in the mean lifetime and intensity-based optical redox ratio. We have clarified this in the text and in a new Table 1.

pg. 6 (line 177): how many macrophages were measured in total to obtain the relative alpha and tau values for fluorescence lifetime shown in figure 2? And were the same macrophages measured at multiple times or are the datapoint shown in Figure 2 representative of different macrophages? These results should be more fully described to allow the reader to understand how much sampling was performed when measuring FLIM.

For each data set we report the number of cells and zebrafish larvae in the figure legends and the supplemental figures. Each data point is a different macrophage. We included a statement in the methods section that images were acquired in a single plane and then we moved to a different plane to avoid measuring the same cell twice.

pg. 7 (line 223): when measuring FAD+ and NAD(P)H fluorescence lifetime the authors describe a difference in the relative proportion of signals in the alpha1 fractional component of free NAD(P)H between infected and uninfected animals (shown in Figure 4) compared to the alpha1 fractional component measured when measuring NAD(P)H alone in animals expressing GFP (shown in Figure 3). It is difficult to discriminate this difference when comparing Figure 3 and 4 supplements as the significance of the difference between control and infected animals for the NAD(P)H alpha1 fraction is only shown in Figure 4 supplement. Comparable data should be shown for both sets of data to allow a comparison of the alpha1 component between them.

We revised the text to address this concern.

pg. 8 (line 250): it is difficult to discern from the text which measures show significant differences in NAD(P)H and FAD+ between burn compared to transection injury models. It would be helpful in this instance to state whether the lifetime measures are increased or decreased and how these differences reflect the redox state. As an example, the authors state that 'The trends for the differences in the mean lifetime and individual lifetime components of NAD(P)H and FAD between the burn wound and simple transection at 24 hpw were comparable to the observed differences between the infected and simple transection (Figure 5C, D, Figure S5A-F).' This does not provide the reader with detail as to what is actually different and should clearly state what such differences are.

We improved the text throughout to make the results more clear.

pg. 8 (line 254): The authors state that there is a significant difference for the NAD(P)H lifetime in macrophages in the tail of burn compared to transection injury models. This is calculated from the relative contributions of tau1 and tau2 to the mean fluorescence lifetime of NAD(P)H (Figure 5 Supplement). What was not discussed is that tau1 for NAD(P)H clearly shows a difference between sterile injuries (306, 95% CI: 289-325) and burn injuries (274, 95% CI: 264-284) at 24 hpi but tau2 is almost identical between these two models. Is this change to the tau1 component important for understanding differences in the mean NAD(P)H lifetime between the two models at 24 hpi and not the tau2 component?

The mean fluorescence lifetime is a weighted average of the short and long lifetime components (τm = α1τ1 + α2τ2). Therefore, data in the supplemental figures can be used to understand changes in this mean lifetime. In this case, the decrease in NAD(P)H τm for burn compared to transection wound at 24 hours is driven by a decrease in τ1 and an increase in α1. This can be interpreted as an increase in the pool of free NADH(P)H for the burn compared to transection wound at 24 hours post wound. This has been clarified in the revised text.

pg. 9 (line 303): the authors describe changes to NAD(P)H lifetimes in animals treated with Metformin and state 'the changes in NAD(P)H tau1 and tau2 also trended as expected'. It is not clear what should be expected and how these trended. The authors should clarify this statement to state what the results actually showed.

We revised the manuscript as requested by the reviewer.

pg. 19 (line 624): the authors state that a general linear model was used to test for significant changes. How was the data fitted to the models, what were the models used, with which assumptions were the models generated and where are the results from the models? I would expect this to be made more explicit so this work could be examined critically, but the level of detail of the statistical approaches used is not sufficient to enable this and should be addressed.

We provided more detail on the statistical methods.

pgs. 36, 38, 40.Supplemental figures 2, 3, 4: what do the R1, R2 etc denote?

R stands for repeat and has been clarified in the text.

Data availability and statistical methodology:

It was not possible to comment on the statistical methods used as there was insufficient detail provided in the manuscript.

We provided more detail on the statistical methods in the revised manuscript.

It was also not possible to discern the exact details for how the acquisition of FLIM data was performed including duration of scanning for each cell and whether a cell was examined more than once.

We have revised the text as requested by the reviewer.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #2 (Recommendations for the authors):

The authors have performed a substantial revision in response to the reviews. In response to my comments they:

– Added a comment in the paper regarding discrepancies between in vitro and in vivo data. Given this was raised by two reviewers, they could provide a stronger commentary, for example by indicating other relevant examples in the literature where in vitro and in vivo observations on cell phenotypes don’t match.

We edited as suggested; see lines 329-333.

– Provided potential reasons in the rebuttal for the low magnitude of the metabolic changes observed in vitro with their system. These reasons could be also included in the manuscript in brief form.

We included this information in the figure legend; lines 830-840.

– Provided a technical information in the rebuttal on how images were segmented in the TNFa expression data (previous Figure 3 current Figure 2). Here I am not sure they answered my question. My point was that in the fluorescence images they show dotted outlines of macrophages which are supposed to be derived from segmentation based on cherry intensity. However, the cherry intensity within those outlines is not uniformly higher than outside the outlines – there are dark/low-intensity areas within the outlines. Conversely there are areas of high-intensity which are not included in the outlines. So it still puzzles me how these outlines are deduced because if segmentation is threshold-based the areas inside the outlines should all have intensities above a certain common threshold. Perhaps the authors need to change the brightness/contrast of the image and/or show us how they threshold the image, to explain more clearly how these cellular outlines are deduced.

We apologize for the confusion. The source of confusion is that macrophages are labeled with membrane-localized mCherry, so there won’t be a uniform signal throughout the cell, and the outlines were placed over the cell membrane. We adjusted the contrast and placed the outlined around the cell. In Author response image 2 we included original and brightness auto-adjusted (in Fiji) images without outlines. In the figure we outline only a few as examples, which is stated in the legend. The mCherry image in Figure 2 was replaced by this auto-adjusted version shown in Author response image 2. The corresponding GFP image was also auto-adjusted in Fiji and replaced in Figure 2.

Author response image 2

– Added a new Table 1 with better explanation of the interpretations of the metrics.

-explained choice of time point for measurements.

– Improved the abstract.

– Improved the discussion. I still found paragraph in lines 349-368 difficult to follow. I am not really sure what the reader should take from it. This paragraph should probably move to supplementary information, as the role seems to be to provide more technical information for some specialists.

We clarified this paragraph; lines 355-366. Our intention was to contextualize changes in the mean lifetime of NAD(P)H that were observed with M1-like (TNFα+) macrophage populations. We wanted to clarify that the NAD(P)H lifetime does not provide information about any specific pathway because NAD(P)H has over 300 binding partners. Additionally, changes in the fluorescence lifetime of NAD(P)H are multi-faceted, and can be attributed to changes in binding activity, preferred binding partners, and microenvironmental factors (e.g., pH). We hope that the edits improve clarity. Altogether they have generally addressed my concerns but I would just recommend some straightforward edits as indicated in the specific points above.Reviewer #3 (Recommendations for the authors):

I thought the authors have addressed most of the points raised well and thoughtfully. The manuscript is much more readable and results are now clear. I think this is a valuable and helpful addition to the field and recommend publication.

I have a few additional comments/ questions that I think would strengthen the manuscript further:

1. I wondered about the declaration on line 277 that 'The modest changes in NAD(P)H lifetime endpoints are likely attributed to the modest decrease in TNFα expression in macrophages upon Metformin treatment'.

This implies that TNF-a regulates the metabolic status of the cell. Is this intended? It would be helpful to therefore clarify the potential relationship between NAD(P)H t1 and t2 and TNF-a expression.

We agree that this statement is confusing. We use TNFa expression strictly as a surrogate measure of macrophage polarization towards pro-inflammatory M1-like subsets, as TNFa is a well-established marker of M1 macrophages in humans, mice and zebrafish. What was meant here is that the magnitude by which we were able to modulate macrophage polarization, as measured by changes in the fraction of TNFa+ macrophages, was modest (~15% decrease by Metformin and ~20% increase by Stat6 depletion). What we meant to imply is that if we were able to cause a larger change in macrophage polarization, let’s say a decrease or increase by 50%, the differences in FLIM readouts would have been larger also. We see this correlation when we compare macrophages at the different wound models. The percentage of TNFa+ macrophages at the infected wound is ~75%, at the burn wound it’s ~50-60% and the simple transection ~10% (as characterized in Miskolci et al. eLife 2019). The differences in FLIM readouts were larger when comparing infected wound versus simple transection (Figure 3) than when comparing burn wound versus simple transection (Figure 4). Even though we used TNFa expression as a marker of M1 polarization, it is possible that TNFa plays a role in macrophage metabolism. This could be tested by TNFa inhibitors for instance. However, we did not intend to speculate on this possibility at this point, therefore we clarified this sentence; lines 277-279.2. How does their study correspond to the descriptions of the metabolic profiling of macrophages during regeneration described by Ratnayake et al., 2019? In this work the authors describe a role for the NAD+ regenerating enzyme Nampt in promoting regeneration through regulating NAD+ replenishment in macrophages. By measuring the NADH/ NAD+ ratios by FLIM they argue the pro-regenerative macrophages display a glycolytic phenotype. It would be helpful to compare results from this work to this study.

Thank you for mentioning this work. We included this reference in the discussion; line 399-401. We believe our conclusions that a reduced intracellular metabolism support regeneration (as indicated by the time-related shift toward a higher optical redox ratio, which means a more reduced intracellular redox state (Figure 4B) and our Metformin data that shows that shifting the intracellular redox state to a more reduced environment (Figure 5D) is associated with better wound healing (Figure 6H)) is consistent with the observations of Ratnayake et al. (PMID: 33568815), as a more reduced intracellular redox state is likely caused by an increase in glycolysis.

3. The major thrust of this work is that it is appropriate to use the ratio of NAD(P)H and FAD to approximate the relative oxidative state of a cell. The two methods used to measure NAD(P)H – intensity measures of autofluorescence and relative changes to fluorescence lifetime – do discriminate between NAD(P)H and NAD(P). However the emission wavelength of the equally important redox molecule NADH is identical to NAD(P)H so any intensity measure will include both. It is important to consider this caveat as changes to NADH levels will affect any such intensity based measures. I wondered therefore if the changes to the optical ratio in various manipulations might not always be correlated to changes of NAD(P)H tau1 and NAD(P)H tau2 due to this confounding factor? It would be greatly helpful to introduce the importance of NADH/ NAD in metabolic signalling and discuss whether the optical ratio utilised in this study would also provide an insight to changes of these critical molecules.

We included a statement in the discussion about the caveat that intensity measurements do not distinguish between NADPH and NADH due to overlapping spectra; lines 346-348.

4. I commented on the caveat that the lifetime for bound NAD(P)H can change depending on the protein it is bound to. The authors have made the data for all measured lifetimes available and discussed it in the Discussion section (line 377), but after reading this I was left wondering what the reader could definitively determine from this measure. The last two sentences of this section in the Discussion are intriguing as they imply changes to tau1 and tau2 could be used to infer changes to different metabolic processes. Is it possible to present this with clearer predictions for what changes to tau1 and tau2 might mean in relation to the metabolism of the cells?

We clarified this paragraph. Please see our response to Reviewer #2 comment “improved the discussion. I still found paragraph in lines 349-368 difficult to follow…”.

https://doi.org/10.7554/eLife.66080.sa2

Article and author information

Author details

  1. Veronika Miskolci

    Department of Medical Microbiology and Immunology, University of Wisconsin-Madison, Madison, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7900-4626
  2. Kelsey E Tweed

    1. Morgridge Institute for Research, Madison, United States
    2. Department of Biomedical Engineering, University of Wisconsin-Madison, Madison, United States
    Contribution
    Data curation, Formal analysis, Investigation, Software
    Competing interests
    No competing interests declared
  3. Michael R Lasarev

    Department of Biostatistics & Medical Informatics, University of Wisconsin-Madison, Madison, United States
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1896-2705
  4. Emily C Britt

    1. Morgridge Institute for Research, Madison, United States
    2. Department of Nutritional Sciences, University of Wisconsin-Madison, Madison, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  5. Alex J Walsh

    Morgridge Institute for Research, Madison, United States
    Present address
    Department of Biomedical Engineering, Texas A&M University, College Station, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3832-8207
  6. Landon J Zimmerman

    Department of Medical Microbiology and Immunology, University of Wisconsin-Madison, Madison, United States
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared
  7. Courtney E McDougal

    Department of Medical Microbiology and Immunology, University of Wisconsin-Madison, Madison, United States
    Contribution
    Resources
    Competing interests
    No competing interests declared
  8. Mark R Cronan

    Department of Molecular Genetics and Microbiology, Duke University School of Medicine, Durham, United States
    Present address
    In Vivo Cell Biology of Infection Unit, Max Planck Institute for Infection Biology, Berlin, Germany
    Contribution
    Resources
    Competing interests
    No competing interests declared
  9. Jing Fan

    1. Morgridge Institute for Research, Madison, United States
    2. Department of Nutritional Sciences, University of Wisconsin-Madison, Madison, United States
    Contribution
    Methodology, Resources
    Competing interests
    No competing interests declared
  10. John-Demian Sauer

    Department of Medical Microbiology and Immunology, University of Wisconsin-Madison, Madison, United States
    Contribution
    Methodology, Resources
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9367-794X
  11. Melissa C Skala

    1. Morgridge Institute for Research, Madison, United States
    2. Department of Biomedical Engineering, University of Wisconsin-Madison, Madison, United States
    Contribution
    Conceptualization, Funding acquisition, Methodology, Resources, Supervision, Writing – review and editing
    For correspondence
    mcskala@wisc.edu
    Competing interests
    No competing interests declared
  12. Anna Huttenlocher

    1. Department of Medical Microbiology and Immunology, University of Wisconsin-Madison, Madison, United States
    2. Department of Pediatrics, University of Wisconsin-Madison, Madison, United States
    Contribution
    Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing – original draft
    For correspondence
    huttenlocher@wisc.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7940-6254

Funding

National Institutes of Health (R35 GM118027)

  • Anna Huttenlocher

National Institutes of Health (R01 CA205101)

  • Melissa C Skala

National Institutes of Health (K99 GM138699)

  • Veronika Miskolci

American Heart Association (17POST33410970)

  • Veronika Miskolci

National Institutes of Health (R21 AI159312)

  • Anna Huttenlocher

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank members of the Huttenlocher and Skala laboratories, notably Elizabeth S Berge, Steve Trier, Kayvan Samimi, Peter R Rehani, Emmanuel Contreras Guzman, Tiffany M Heaster and Amani Gillette, and our collaborators from the Laboratory for Optical and Computational Instrumentation (LOCI), Jayne M Squirrell and Kevin W Eliceiri, for technical assistance and valuable discussions.

Ethics

Animal care and use was approved by the Institutional Animal Care and Use Committee of University of Wisconsin and strictly followed guidelines set by the federal Health Research Extension Act and the Public Health Service Policy on the Humane Care and Use of Laboratory Animal, administered by the National Institute of Health Office of Laboratory Animal Welfare. All protocols using zebrafish and mouse in this study have been approved by the University of Wisconsin-Madison Research Animals Resource Center (protocols M005405-A02/zebrafish, M005916/mouse).

Senior Editor

  1. Didier YR Stainier, Max Planck Institute for Heart and Lung Research, Germany

Reviewing Editor

  1. Serge Mostowy, London School of Hygiene & Tropical Medicine, United Kingdom

Reviewer

  1. Robert Knight, Kings College London, United Kingdom

Publication history

  1. Preprint posted: June 16, 2020 (view preprint)
  2. Received: December 29, 2020
  3. Accepted: February 2, 2022
  4. Version of Record published: February 24, 2022 (version 1)

Copyright

© 2022, Miskolci et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,365
    Page views
  • 201
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Veronika Miskolci
  2. Kelsey E Tweed
  3. Michael R Lasarev
  4. Emily C Britt
  5. Alex J Walsh
  6. Landon J Zimmerman
  7. Courtney E McDougal
  8. Mark R Cronan
  9. Jing Fan
  10. John-Demian Sauer
  11. Melissa C Skala
  12. Anna Huttenlocher
(2022)
In vivo fluorescence lifetime imaging of macrophage intracellular metabolism during wound responses in zebrafish
eLife 11:e66080.
https://doi.org/10.7554/eLife.66080

Further reading

    1. Cell Biology
    Meng Zhao, Niels Banhos Dannieskiold-Samsøe ... Katrin J Svensson
    Research Article

    The secreted protein Isthmin-1 (Ism1) mitigates diabetes by increasing adipocyte and skeletal muscle glucose uptake by activating the PI3K-Akt pathway. However, while both Ism1 and insulin converge on these common targets, Ism1 has distinct cellular actions suggesting divergence in downstream intracellular signaling pathways. To understand the biological complexity of Ism1 signaling, we performed phosphoproteomic analysis after acute exposure, revealing overlapping and distinct pathways of Ism1 and insulin. We identify a 53 % overlap between Ism1 and insulin signaling and Ism1-mediated phosphoproteome-wide alterations in ~ 450 proteins that are not shared with insulin. Interestingly, we find several unknown phosphorylation sites on proteins related to protein translation, mTOR pathway and, unexpectedly, muscle function in the Ism1 signaling network. Physiologically, Ism1 ablation in mice results in altered proteostasis, including lower muscle protein levels under fed and fasted conditions, reduced amino acid incorporation into proteins, and reduced phosphorylation of the key protein synthesis effectors Akt and downstream mTORC1 targets. As metabolic disorders such as diabetes are associated with accelerated loss of skeletal muscle protein content, these studies define a non-canonical mechanism by which this anti-diabetic circulating protein controls muscle biology.

    1. Cell Biology
    Jia Chen, Daniel St Johnston
    Research Article

    In the adult Drosophila midgut, basal intestinal stem cells give rise to enteroblasts that integrate into the epithelium as they differentiate into enterocytes. Integrating enteroblasts must generate a new apical domain and break through the septate junctions between neighbouring enterocytes, while maintaining barrier function. We observe that enteroblasts form an apical membrane initiation site (AMIS) when they reach the septate junction between the enterocytes. Cadherin clears from the apical surface and an apical space appears between above the enteroblast. New septate junctions then form laterally with the enterocytes and the AMIS develops into an apical domain below the enterocyte septate junction. The enteroblast therefore forms a pre-assembled apical compartment before it has a free apical surface in contact with the gut lumen. Finally, the enterocyte septate junction disassembles and the enteroblast/pre-enterocyte reaches the gut lumen with a fully-formed brush border. The process of enteroblast integration resembles lumen formation in mammalian epithelial cysts, highlighting the similarities between the fly midgut and mammalian epithelia.