Non-invasive classification of macrophage polarisation by 2P-FLIM and machine learning
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
Neto et al set out to use Two-Photon FLIM and machine learning to classify macrophages that are polarised along the M1/M2 axis and then subjected to different metabolic stresses classically used to distinguish metabolic strategies of different cell states. Additional information is sought regarding the photophysics of the measurements and if there are an adequate number of photons to fairly compare the three conditions. The work will be of interest to immunologists, physiologists interested in metabolism and engineers looking to translate the findings.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their names with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Bioengineering (eLife)
Abstract
In this study, we utilise fluorescence lifetime imaging of NAD(P)H-based cellular autofluorescence as a non-invasive modality to classify two contrasting states of human macrophages by proxy of their governing metabolic state. Macrophages derived from human blood-circulating monocytes were polarised using established protocols and metabolically challenged using small molecules to validate their responding metabolic actions in extracellular acidification and oxygen consumption. Large field-of-view images of individual polarised macrophages were obtained using fluorescence lifetime imaging microscopy (FLIM). These were challenged in real time with small-molecule perturbations of metabolism during imaging. We uncovered FLIM parameters that are pronounced under the action of carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), which strongly stratifies the phenotype of polarised human macrophages; however, this performance is impacted by donor variability when analysing the data at a single-cell level. The stratification and parameters emanating from a full field-of-view and single-cell FLIM approach serve as the basis for machine learning models. Applying a random forests model, we identify three strongly governing FLIM parameters, achieving an area under the receiver operating characteristics curve (ROC-AUC) value of 0.944 and out-of-bag (OBB) error rate of 16.67% when classifying human macrophages in a full field-of-view image. To conclude, 2P-FLIM with the integration of machine learning models is showed to be a powerful technique for analysis of both human macrophage metabolism and polarisation at full FoV and single-cell level.
Article activity feed
-
-
Evaluation Summary:
Neto et al set out to use Two-Photon FLIM and machine learning to classify macrophages that are polarised along the M1/M2 axis and then subjected to different metabolic stresses classically used to distinguish metabolic strategies of different cell states. Additional information is sought regarding the photophysics of the measurements and if there are an adequate number of photons to fairly compare the three conditions. The work will be of interest to immunologists, physiologists interested in metabolism and engineers looking to translate the findings.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their names with the authors.)
-
Reviewer #1 (Public Review):
This study investigates the use of two-photon fluorescence lifetime microscopy (FLIM) with single-cell resolution for endogenous FAD+/NAD(P)H+ signals in polarized human macrophages in vitro as an approach to classification through machine learning. They perform parallel conventional analysis of secreted proteins and metabolomics to complement the FLIM measurements. They find that fluorescence lifetime parameters can be used to distinguish M1 and M2 macrophages in a population and particularly the response of the different populations to the metabolic inhibitor FCCP. The results are put in context with earlier work on the role of fluorescent metabolite environment and protein binding in the context of the metabolic data on FLIM parameters.
-
Reviewer #2 (Public Review):
Neto et al. use advanced two-photon FLIM to investigate human macrophages based on their autofluorescence (mainly coming from NAD(P)H. The main hypothesis is that different metabolic activities coming from human macrophages will aid in their classification in primary cells.
It is nice that the authors explain why they use a double exponential approach. I wonder however if the limited signal to noise and therefore high background allows for such a model. One should perhaps consider a triple exponential that also accounts for the noise? How do the authors deal with this limitation? Other labs have used in the past non-fitting approaches (i.e. Phasor plot). Have you considered employing this approach? If not, why? It seems that in this case, it would be a good implementation provided that the background and …
Reviewer #2 (Public Review):
Neto et al. use advanced two-photon FLIM to investigate human macrophages based on their autofluorescence (mainly coming from NAD(P)H. The main hypothesis is that different metabolic activities coming from human macrophages will aid in their classification in primary cells.
It is nice that the authors explain why they use a double exponential approach. I wonder however if the limited signal to noise and therefore high background allows for such a model. One should perhaps consider a triple exponential that also accounts for the noise? How do the authors deal with this limitation? Other labs have used in the past non-fitting approaches (i.e. Phasor plot). Have you considered employing this approach? If not, why? It seems that in this case, it would be a good implementation provided that the background and scattered photons might have different lifetimes as compared to the ones coming from NAD(p)H bound/unbound.
It will be of interest to check all parameters provided (both lifetimes and the normalised pre-exponential factors) versus the number of photons. Also, plot these in a pixel by pixel basis and present the corresponding histograms. According to the few images presented it seems clear to me that measuring autofluorescence could be a big problem, especially using such low magnification (see Fig. 3A and 4A) - By the way - Fig 4 is named Fig. 3 in the text. In these figures, one cannot see any changes when inspecting the pixel by pixel data. The changes in ps are minimal (also check the resolution of your figures, it is very difficult to check the legends and values).
It is interesting that one can employ Maniche learning to analyse the FLIM data, but one should be able to see clear differences in the average lifetimes when comparing different macrophage populations and the authors should show these pictures with the corresponding histograms and the number of photons.
When measuring autofluorescence (and noise) one needs to be very accurate and I commend the authors to produce all 2D plots showing the number of photons versus their lifetimes to check if there is no correlation between these two parameters. If this is the case (positive correlation) then all data is not correct.
-
