Living cells rely on transport and interaction of biomolecules to perform their diverse functions. To investigate the underlying molecular processes in the native cellular environment, minimally invasive techniques are needed. Fluorescence fluctuation spectroscopy (FFS) approaches provide a powerful toolbox that fulfills this aim (Jameson et al., 2009; Weidemann et al., 2014; Petazzi et al., 2020). FFS takes advantage of inherent molecular dynamics present in biological systems, for example, diffusion, to obtain molecular parameters from fluctuations of the signal emitted by an ensemble of fluorescent molecules. More in detail, the temporal evolution of such fluctuations allows the quantification of intracellular dynamics. In addition, concentration and oligomerization state of molecular complexes can be determined by analyzing the magnitude of fluctuations. Finally, hetero-interactions of different molecular species can be detected by cross-correlation analysis of fluctuations emitted by spectrally separated fluorophores (Schwille et al., 1997). Over the last two decades, several experimental FFS schemes such as raster image (cross-) correlation spectroscopy (RI(C)CS) (Digman et al., 2005; Digman et al., 2009b), (cross-correlation) Number&Brightness analysis (Digman et al., 2008; Digman et al., 2009a), and imaging FCS (Krieger et al., 2015) have been developed, extending the concept of traditional single-point fluorescence (cross-) correlation spectroscopy (F(C)CS) (Magde et al., 1972). A further interesting example of FFS analysis relevant in the field of cell biology is represented by scanning F(C)CS (SF(C)CS). Using a scan path perpendicular to the plasma membrane (PM), this technique provides enhanced stability and the ability to probe slow membrane dynamics (Ries and Schwille, 2006), protein interactions (Ries et al., 2009b; Dunsing et al., 2017), and oligomerization (Dunsing et al., 2018) at the PM of cells.

FFS studies are conventionally limited to the analysis of two spectrally distinguished species due to (i) broad emission spectra of fluorophores with consequent cross-talk artifacts and (ii) limited overlap of detection/excitation geometries for labels with large spectral separation. Generally, only a few fluorescence-based methods are available to detect ternary or higher order interactions of proteins (Galperin et al., 2004; Sun et al., 2010; Hur et al., 2016). First in vitro approaches to perform FCS on more than two species exploited quantum dots (Burkhardt et al., 2005) or fluorescent dyes with different Stokes shifts excited with a single laser line in one- (Hwang et al., 2006) or two-photon excitation (Heinze et al., 2004; Ridgeway et al., 2012a), coupled with detection on two or more single photon counting detectors. Following an alternative conceptual approach, it was shown in vitro that two spectrally strongly overlapping fluorophore species can be discriminated in FCS by applying statistical filtering of detected photons based on spectrally resolved (fluorescence spectral correlation spectroscopy [FSCS]; Benda et al., 2014) or fluorescence lifetime (fluorescence lifetime correlation spectroscopy [FLCS]; Böhmer et al., 2002; Kapusta et al., 2007; Ghosh et al., 2018) detection. Such framework allows the minimization of cross-talk artifacts in FCCS measurements performed in living cells (Padilla-Parra et al., 2011). Recently, three-species implementations of RICCS and FCCS were successfully demonstrated for the first time in living cells. Schrimpf et al. presented raster spectral image correlation spectroscopy (RSICS), a powerful combination of RICS with spectral detection and statistical filtering based on the emission spectra of mEGFP, mVenus, and mCherry fluorophores (Schrimpf et al., 2018). Stefl et al. developed single-color fluorescence lifetime cross-correlation spectroscopy (sc-FLCCS), taking advantage of several GFP variants characterized by short or long fluorescence lifetimes (Štefl et al., 2020). Using this elegant approach, three-species FCCS measurements could be performed in yeast cells, with just two excitation lines.

Here, we explore the full potential of FSCS and RSICS. In particular, we present scanning fluorescence spectral correlation spectroscopy (SFSCS), combining SFCS and FSCS. We show that SFSCS enables cross-talk-free SFCCS measurements of two protein species at the PM of living cells tagged with strongly overlapping fluorophores in the green or red regions of the visible spectrum, excited with a single excitation line. This approach results in correct estimates of protein diffusion dynamics, oligomerization, and interactions between both species. Further, we extend our approach to the analysis of three or four interacting partners: by performing cross-correlation measurements on different fluorescent protein (FP) hetero-oligomers, we demonstrate that up to four FP species can be simultaneously analyzed. We then apply this scheme to simultaneously investigate the interaction of influenza A virus (IAV) matrix protein 2 (M2) with two cellular host factors, the tetraspanin CD9 and the autophagosome protein LC3, co-expressed in the same cell. Finally, we extend RSICS for the detection of four molecular species and quantify, for the first time directly in living cells, the complete stoichiometry of ternary IAV polymerase complexes assembling in the nucleus, using three-species fluorescence correlation and brightness analysis.

In this work, we combine FFS techniques with spectral detection to perform multicolor studies of protein interactions and dynamics in living cells. In particular, we present SFSCS, a combination of FSCS (Benda et al., 2014) and lateral scanning FCS (Ries and Schwille, 2006). We show that SFSCS allows cross-talk-free measurements of protein interactions and diffusion dynamics at the PM of cells and demonstrate that it is capable of detecting three or four species simultaneously. Furthermore, we extend RSICS (Schrimpf et al., 2018) to investigate four fluorophore species and apply this approach to determine the stoichiometry of higher order protein complexes assembling in the cell nucleus. Notably, the technical approaches can be carried out on a standard confocal microscope, equipped with a spectral photon counting detector system.

In the first part, we present two-species SFSCS using a single excitation wavelength and strongly overlapping fluorophores. Compared to the conventional implementation of FCCS with two excitation lasers and two detectors, two-species SFSCS has substantial advantages, similar to the recently presented sc-FLCCS (Štefl et al., 2020). Since it requires a single excitation line and is compatible with spectrally strongly overlapping FPs, it circumvents optical limitations such as imperfect overlap of the observation volumes. This is evident from higher rel.cc. values of 70–80% measured for mEGFP and mEYFP coupled in FP hetero-oligomers compared to 45–60% observed for mEGFP and mCherry2. Rel.cc. values around 70% are to be expected for the examined FP tandems even in the case of single-wavelength excitation, given that the p_f for such fluorophores is indeed around 0.7 (Foo et al., 2012; Dunsing et al., 2018) (see also SI, paragraph 1). On the other hand, in three- and four species measurements discussed below, FP pairs requiring two excitation wavelengths display the typical reduction of the rel.cc. due to imperfect optical volume overlap. For combinations of green and red FPs, rel.cc. values below 60% were also observed with single-wavelength excitation (Foo et al., 2012; Shi et al., 2009), indicating that overlap of both excitation and detection volumes (the latter requiring FPs with similar emission spectra) is required to maximize the achievable cross-correlation (Foo et al., 2012). Notably, two-species SFSCS can not only successfully discriminate between mEGFP and mEYFP, but is also applicable when using the red FPs mApple and mCherry2. These two FPs were successfully used in several FFS studies (Dunsing et al., 2018; Foust et al., 2019; Sankaran et al., 2021), providing the best compromise between brightness, maturation, and photostability among available red FPs, which generally suffer from reduced SNR compared to FPs emitting in the green or yellow part of the optical spectrum (Dunsing et al., 2017; Dunsing et al., 2018; Cranfill et al., 2016).

In comparison to sc-FLCCS, it may be more robust to discriminate fluorophores based on spectra rather than lifetimes, which can be strongly affected by FRET (Štefl et al., 2020). The emission spectra of the FPs utilized in this study did not depend on cell lines or subcellular localization (Figure 5—figure supplement 1) and showed no (mEGFP, mEYFP) or little (mApple, mCherry2) variation with pH over a range of 5.0–9.2 (Figure 5—figure supplement 2). For red FPs, specifically mApple, a red shift appeared at more acidic pH, in agreement with previous studies (Hendrix et al., 2008). This aspect should be considered for specific applications, for example, RSICS in the cytoplasm containing acidic compartments such as lysosomes. Generally, spectral approaches require accurate detection of photons in each spectral bin. A previous study using the same detection system reported intrinsic cross-talk between adjacent spectral bins (Foust et al., 2019). However, since the methodology presented here is based on temporal (SFSCS) or spatial (RSICS) correlation (both excluding the correlation at zero time or spatial lag), this issue can be neglected in our analysis.

A major limitation of SFSCS is the reduced SNR of the CFs (see Figure 1, Figure 3—figure supplement 3) caused by the statistical filtering of the signal emitted by spectrally overlapping fluorophore species (see, e.g., Figure 4—figure supplement 1). This limitation applies to all FFS methods that discriminate different fluorophore species based on spectral (e.g., FSCS [Benda et al., 2014], RSICS [Schrimpf et al., 2018]) or lifetime patterns (e.g., sc-FLCCS [Štefl et al., 2020]). The increase in noise depends on the spectral (or lifetime) overlap of different species and is more prominent for species that completely lack ‘pure’ channels, that is, detection channels in which the majority of photons can be univocally assigned to a single species (Schrimpf et al., 2018). In sc-FLCCS, this issue particularly compromises the SNR of short lifetime species (Štefl et al., 2020) since photons of longer lifetime species are detected in all ‘short lifetime’ channels at substantial relative numbers. In these conditions, sc-FLCCS could not provide reliable results with sixfold (or higher) difference in relative protein abundance, even though the lower abundant protein was tagged with the brighter, longer lifetime FP (Štefl et al., 2020). Similarly in SFSCS, CFs corresponding to mEYFP or mCherry2 were most prone to noise (Figure 1C and F) since all channels that contain, for example, mEYFP signal also contain mEGFP signal (Figure 1—figure supplement 1). In our experiments, cross-talk-free SFSCS analysis with two species excited with a single excitation wavelength could be performed for relative intensity levels as low as 1:10 (mEGFP/mEYFP) or 1:5 (mApple/mCherry2). In this range, SFSCS not only enabled the quantification of protein interactions via cross-correlation analysis, but also yielded correct estimates of protein diffusion dynamics and oligomerization at the PM. An improvement of the allowed relative concentration range can be achieved by using brighter or more photostable fluorophores, for example, organic dyes, compensating for reduced SNR due to statistical filtering. Alternatively, FP tags could be selected based on proteins’ oligomerization state. For example, monomeric proteins exhibiting low molecular brightness should be tagged with fluorophores that are less prone to noise. It should be noted that the limitation of reduced SNR due to excess signal from another species also applies to conventional dual-color FCCS: bleed-through from green to red channels can be corrected on average, but reduces the SNR in red channels (Bacia et al., 2012), unless more sophisticated schemes such as pulsed interleaved excitation (Müller et al., 2005; Hendrix et al., 2013) are applied.

Having demonstrated that two-species SFSCS is feasible with a single excitation wavelength in the green (mEGFP, mEYFP) or red (mApple, mCherry2) part of the visible spectrum, we finally implemented three- and four-species SFSCS as well as four-species RSICS. These extensions do not further compromise the SNR of CFs detected for mEGFP and mEYFP (see Figure 3—figure supplement 3A,B), but may additionally reduce the SNR of CFs corresponding to red FPs (in particular when mEGFP and/or mEYFP concentration is much higher than that of red FPs, Figure 3—figure supplement 3C). For this reason, three- and four-species analysis was restricted to cells with relative average intensity levels of 1:5 or less between species with adjacent emission spectra. In this range, the increase in noise due to statistical filtering was moderate, benefitting from the fairly large spectral separation of green/yellow and red emission (Figure 3—figure supplement 3). In addition, the higher molecular brightness of mApple (compared to mCherry2) compensated for the larger overlap of this FP with the tail of mEYFP emission. The excitation power for red FPs was generally limited by the lower photostability of mApple, which could be responsible for consistently lower rel.cc. values of mEGFP or mEYFP with mApple than with mCherry2. Nevertheless, four-species SFSCS and RSICS could successfully resolve different combinations of strongly overlapping FP hetero-oligomers, for example, a mixture of mEGFP-mCherry2 and mEYFP-mApple heterodimers, at the PM or in the cytoplasm of cells. To explore the interaction of four different FP-tagged proteins, four-species FFS may substantially reduce the experimental effort because all pair-wise interactions can be quantified in a single measurement (instead of six separate conventional two-species FCCS measurements). Yet, weak interaction of proteins, that is, a low amount of hetero-complexes compared to a high amount of unbound proteins, may not be detectable due to the large noise of the CCF in this case. The SNR might be further compromised by slow FP maturation or dark FP states, limiting the amount of complexes that simultaneously emit fluorescence of all bound FP species (Dunsing et al., 2018). Ultimately, the mentioned limitations currently restrict SFSCS and RSICS to four FP species. The approaches would thus strongly benefit from a multiparametric analysis. For instance, combining spectral and lifetime detection schemes would provide additional contrast for photons detected in the same spectral bin. This improvement could expand the range of detectable relative concentrations or might allow further multiplexing of FFS.

Conventional two-color scanning FCCS has been previously applied to quantify receptor-ligand interactions in living zebrafish embryos (Ries et al., 2009b) and CRISPR/Cas9 edited cell lines to study such interactions at endogenous protein level (Eckert et al., 2020). SFSCS is thus directly applicable in the complex environment of living multicellular organisms. In this context, spectral information could be further exploited to separate low signal levels of endogenously expressed, fluorescently tagged proteins from autofluorescence background.

As a first biological application of SFSCS, we investigated the interaction of IAV matrix protein M2 with two cellular host factors: the tetraspanin CD9 and the autophagosome protein LC3. We observed strong association of LC3 with M2, and consequent recruitment of LC3 to the PM (Figure 3—figure supplement 4), in agreement with previous in vitro and localization studies (Beale et al., 2014). Interestingly, molecular brightness analysis reported oligomerization (dimers to tetramers) of M2, but indicated a monomeric state of LC3 at the PM, that is, binding of LC3 to M2 in an apparent stoichiometry of 1:2 to 1:4. However, each M2 monomer provides a binding site for LC3 in the cytoplasmic tail (Claridge et al., 2020). A more detailed analysis of our data showed that in the analyzed cells (i.e., cells showing clear membrane recruitment of LC3, Figure 3—figure supplement 4A,B), the PM concentration of LC3 was on average only 30% compared to that of M2 (Figure 3—figure supplement 4C), although both proteins were expressed in comparable amounts in the sample in general. This suggests that not all potential binding sites in the cytoplasmic tail of M2 may be available to fluorescently tagged LC3, either due to binding of endogenous LC3, other cellular host factors, or steric hindrance. In contrast to the case of LC3, we did not detect significant binding of M2 with the tetraspanin CD9, a protein that was previously shown to be incorporated into IAV virions and supposedly plays a functional role during the infection process (Shaw et al., 2008; Hutchinson, 2014). Of note, we cannot exclude the possibility that the FP tag at the C-terminus of CD9 might hamper interactions with M2, in the specific case of M2-CD9 interaction being mediated by the C-terminal cytoplasmic tails of the two proteins. In future studies, the approach presented here may be used to further elucidate the complex interaction network of viral proteins, for example, matrix protein 1 (M1) (Hilsch et al., 2014), M2, HA, and neuraminidase, cellular host factors, and PM lipids (Bobone et al., 2017) during the assembly process of IAV at the PM of living cells (Rossman and Lamb, 2011).

Finally, we demonstrated that RSICS allows the quantification of the stoichiometry of higher order molecular complexes, based on molecular brightness analysis for each FP species. As example of an application in a biological context, we determined the stoichiometry of the IAV PC. Our data provide strong evidence for a 2:2:2 stoichiometry of the PC subunits PA, PB1, and PB2, that is, dimerization of heterotrimeric PCs. Such interactions were previously proposed based on experiments in solution using X-ray crystallography and cryo-electron microscopy (Fan et al., 2019), co-immunoprecipitation assays (Jorba et al., 2008; Nilsson-Payant et al., 2018), as well as single-channel brightness analysis of FCCS data (for the PA subunit) (Huet et al., 2010). Intermolecular interactions in the PC are hypothesized to be required for the initiation of vRNA synthesis during replication of the viral genome (Fan et al., 2019; Chen et al., 2019). The results presented here provide the first quantification of these interactions in living cells and a direct estimate of the stoichiometry of PCs in the cell nucleus. The formation of ternary PC complexes in these samples could be extrapolated from the observed high rel.cc. values for all three pair combinations, indicating very low amounts of unbound PA, PB1, or PB2 and higher order interactions (see Appendix 1, Section A1.1 for additional details). Furthermore, this observation could also be directly confirmed by performing, for the first time in living cells, a triple-correlation analysis (TRICS), indicating the presence of a considerable amount of PA-PB1-PB2 complexes. It is worth noting though that the detection of coincident triple fluctuations is prone to considerable noise and thus still limited to molecular complexes present at low concentration and characterized by high molecular brightness for each fluorophore species (Ridgeway et al., 2012a; Ridgeway et al., 2012b).

Of note, the RSICS approach presented here provides for the first time simultaneous information on molecular interactions, molecular brightness (and thus stoichiometry), diffusion dynamics, and concentration for all three complex subunits. This specific feature opens the possibility of a more in-depth analysis. For example, it is possible to quantify the relative cross-correlation of two subunits, e.g. PA and PB2, as a function of the relative concentration of the third subunit, for example, PB1 (Figure 6—figure supplement 1A). Similarly, molecular brightness and diffusion coefficients can be analyzed as a function of the abundance of each subunit (Figure 6—figure supplement 1B,C). With this approach, it is therefore possible to distinguish specific molecular mechanisms, such as inefficient PA-PB2 interactions in the presence of low PB1 concentration or efficient heterotrimer dimerization when all subunits are present at similar concentrations. The employed experimental scheme offers a powerful tool for future studies, exploring, for example, interaction of the PC with cellular host factors or the development of inhibitors that could interfere with the assembly process of the complex, as a promising therapeutic target for antiviral drugs (Massari et al., 2021).

All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for Figures 1-7. The analysis software is freely available on GitHub: https://github.com/VaDu8989/SpectralFFS.

1. 1. Bacia K
   2. Petrášek Z
   3. Schwille P

   (2012) Correcting for spectral cross-talk in dual-color fluorescence cross-correlation spectroscopy

   Chemphyschem 13:1221–1231.

   https://doi.org/10.1002/cphc.201100801

   • PubMed
   • Google Scholar
2. 1. Beale R
   2. Wise H
   3. Stuart A
   4. Ravenhill BJ
   5. Digard P
   6. Randow F

   (2014) A LC3-interacting motif in the influenza A virus M2 protein is required to subvert autophagy and maintain virion stability

   Cell Host & Microbe 15:239–247.

   https://doi.org/10.1016/j.chom.2014.01.006

   • PubMed
   • Google Scholar
3. 1. Benda A
   2. Kapusta P
   3. Hof M
   4. Gaus K

   (2014) Fluorescence spectral correlation spectroscopy (FSCS) for probes with highly overlapping emission spectra

   Optics Express 22:2973–2988.

   https://doi.org/10.1364/OE.22.002973

   • PubMed
   • Google Scholar
4. 1. Bobone S
   2. Hilsch M
   3. Storm J
   4. Dunsing V
   5. Herrmann A
   6. Chiantia S

   (2017) Phosphatidylserine lateral organization influences the interaction of influenza virus matrix protein 1 with lipid membranes

   Journal of Virology 91:e00267-17.

   https://doi.org/10.1128/JVI.00267-17

   • PubMed
   • Google Scholar
5. 1. Böhmer M
   2. Wahl M
   3. Rahn HJ
   4. Erdmann R
   5. Enderlein J

   (2002) Time-resolved fluorescence correlation spectroscopy

   Chemical Physics Letters 353:439–445.

   https://doi.org/10.1016/S0009-2614(02)00044-1

   • Google Scholar
6. 1. Burkhardt M
   2. Heinze KG
   3. Schwille P

   (2005) Four-color fluorescence correlation spectroscopy realized in a grating-based detection platform

   Optics Letters 30:2266–2268.

   https://doi.org/10.1364/ol.30.002266

   • PubMed
   • Google Scholar
7. 1. Chen K-Y
   2. Santos Afonso ED
   3. Enouf V
   4. Isel C
   5. Naffakh N

   (2019) Influenza virus polymerase subunits co-evolve to ensure proper levels of dimerization of the heterotrimer

   PLOS Pathogens 15:e1008034.

   https://doi.org/10.1371/journal.ppat.1008034

   • PubMed
   • Google Scholar
8. 1. Ciccotosto GD
   2. Kozer N
   3. Chow TTY
   4. Chon JWM
   5. Clayton AHA

   (2013) Aggregation distributions on cells determined by photobleaching image correlation spectroscopy

   Biophysical Journal 104:1056–1064.

   https://doi.org/10.1016/j.bpj.2013.01.009

   • PubMed
   • Google Scholar
9. 1. Claridge JK
   2. Mohd-Kipli F
   3. Florea A
   4. Gate T
   5. Schnell JR

   (2020) Ph-dependent secondary structure propensity of the influenza A virus m2 cytoplasmic tail

   Biomolecular NMR Assignments 14:157–161.

   https://doi.org/10.1007/s12104-020-09937-8

   • PubMed
   • Google Scholar
10. 1. Cranfill PJ
    2. Sell BR
    3. Baird MA
    4. Allen JR
    5. Lavagnino Z
    6. de Gruiter HM
    7. Kremers GJ
    8. Davidson MW
    9. Ustione A
    10. Piston DW

    (2016) Quantitative assessment of fluorescent proteins

    Nature Methods 13:557–562.

    https://doi.org/10.1038/nmeth.3891

    • PubMed
    • Google Scholar
11. 1. Dahmane S
    2. Doucet C
    3. Le Gall A
    4. Chamontin C
    5. Dosset P
    6. Murcy F
    7. Fernandez L
    8. Salas D
    9. Rubinstein E
    10. Mougel M
    11. Nollmann M
    12. Milhiet PE

    (2019) Nanoscale organization of tetraspanins during HIV-1 budding by correlative DSTORM/AFM

    Nanoscale 11:6036–6044.

    https://doi.org/10.1039/c8nr07269h

    • PubMed
    • Google Scholar
12. 1. Digman MA
    2. Brown CM
    3. Sengupta P
    4. Wiseman PW
    5. Horwitz AR
    6. Gratton E

    (2005) Measuring fast dynamics in solutions and cells with a laser scanning microscope

    Biophysical Journal 89:1317–1327.

    https://doi.org/10.1529/biophysj.105.062836

    • PubMed
    • Google Scholar
13. 1. Digman MA
    2. Dalal R
    3. Horwitz AF
    4. Gratton E

    (2008) Mapping the number of molecules and brightness in the laser scanning microscope

    Biophysical Journal 94:2320–2332.

    https://doi.org/10.1529/biophysj.107.114645

    • PubMed
    • Google Scholar
14. 1. Digman MA
    2. Wiseman PW
    3. Choi C
    4. Horwitz AR
    5. Gratton E

    (2009a) Stoichiometry of molecular complexes at adhesions in living cells

    PNAS 106:2170–2175.

    https://doi.org/10.1073/pnas.0806036106

    • PubMed
    • Google Scholar
15. 1. Digman MA
    2. Wiseman PW
    3. Horwitz AR
    4. Gratton E

    (2009b) Detecting protein complexes in living cells from laser scanning confocal image sequences by the Cross correlation raster image spectroscopy method

    Biophysical Journal 96:707–716.

    https://doi.org/10.1016/j.bpj.2008.09.051

    • PubMed
    • Google Scholar
16. 1. Dunsing V
    2. Mayer M
    3. Liebsch F
    4. Multhaup G
    5. Chiantia S

    (2017) Direct evidence of amyloid precursor-like protein 1 trans interactions in cell-cell adhesion platforms investigated via fluorescence fluctuation spectroscopy

    Molecular Biology of the Cell 28:3609–3620.

    https://doi.org/10.1091/mbc.E17-07-0459

    • PubMed
    • Google Scholar
17. 1. Dunsing V
    2. Chiantia S

    (2018) A fluorescence fluctuation spectroscopy assay of protein-protein interactions at cell-cell contacts

    Journal of Visualized Experiments 1:e58582.

    https://doi.org/10.3791/58582

    • Google Scholar
18. 1. Dunsing V
    2. Luckner M
    3. Zühlke B
    4. Petazzi RA
    5. Herrmann A
    6. Chiantia S

    (2018) Optimal fluorescent protein tags for quantifying protein oligomerization in living cells

    Scientific Reports 8:10634.

    https://doi.org/10.1038/s41598-018-28858-0

    • PubMed
    • Google Scholar
19. Software

    1. Dunsing V
    2. Chiantia S

    (2021) SpectralFFS, version 8942980

    Github.

    https://github.com/VaDu8989/SpectralFFS
20. 1. Eckert AF
    2. Gao P
    3. Wesslowski J
    4. Wang X
    5. Rath J
    6. Nienhaus K
    7. Davidson G
    8. Nienhaus GU

    (2020) Measuring ligand-cell surface receptor affinities with axial line-scanning fluorescence correlation spectroscopy

    eLife 9:e55286.

    https://doi.org/10.7554/eLife.55286

    • PubMed
    • Google Scholar
21. 1. Fan H
    2. Walker AP
    3. Carrique L
    4. Keown JR
    5. Serna Martin I
    6. Karia D
    7. Sharps J
    8. Hengrung N
    9. Pardon E
    10. Steyaert J
    11. Grimes JM
    12. Fodor E

    (2019) Structures of Influenza a virus RNA polymerase offer insight into viral genome replication

    Nature 573:287–290.

    https://doi.org/10.1038/s41586-019-1530-7

    • PubMed
    • Google Scholar
22. 1. Florin L
    2. Lang T

    (2018) Tetraspanin assemblies in virus infection

    Frontiers in Immunology 9:1140.

    https://doi.org/10.3389/fimmu.2018.01140

    • PubMed
    • Google Scholar
23. 1. Foo YH
    2. Naredi-Rainer N
    3. Lamb DC
    4. Ahmed S
    5. Wohland T

    (2012) Factors affecting the quantification of biomolecular interactions by fluorescence cross-correlation spectroscopy

    Biophysical Journal 102:1174–1183.

    https://doi.org/10.1016/j.bpj.2012.01.040

    • PubMed
    • Google Scholar
24. 1. Foust DJ
    2. Godin AG
    3. Ustione A
    4. Wiseman PW
    5. Piston DW

    (2019) Two-color spatial cumulant analysis detects heteromeric interactions between membrane proteins

    Biophysical Journal 117:1764–1777.

    https://doi.org/10.1016/j.bpj.2019.09.028

    • PubMed
    • Google Scholar
25. 1. Galperin E
    2. Verkhusha VV
    3. Sorkin A

    (2004) Three-chromophore fret Microscopy to analyze multiprotein interactions in living cells

    Nature Methods 1:209–217.

    https://doi.org/10.1038/nmeth720

    • PubMed
    • Google Scholar
26. 1. Ghosh A
    2. Karedla N
    3. Thiele JC
    4. Gregor I
    5. Enderlein J

    (2018) Fluorescence Lifetime Correlation Spectroscopy: Basics and applications

    Methods 140–141:32–39.

    https://doi.org/10.1016/j.ymeth.2018.02.009

    • PubMed
    • Google Scholar
27. 1. Hantak MP
    2. Qing E
    3. Earnest JT
    4. Gallagher T

    (2019) Tetraspanins: Architects of viral entry and exit platforms

    Journal of Virology 93:e01429-17.

    https://doi.org/10.1128/JVI.01429-17

    • PubMed
    • Google Scholar
28. 1. Heinze KG
    2. Jahnz M
    3. Schwille P

    (2004) Triple-color coincidence analysis: One step further in following higher order molecular complex formation

    Biophysical Journal 86:506–516.

    https://doi.org/10.1016/S0006-3495(04)74129-6

    • PubMed
    • Google Scholar
29. 1. Hendrix J
    2. Flors C
    3. Dedecker P
    4. Hofkens J
    5. Engelborghs Y

    (2008) Dark states in monomeric red fluorescent proteins studied by fluorescence correlation and single molecule spectroscopy

    Biophysical Journal 94:4103–4113.

    https://doi.org/10.1529/biophysj.107.123596

    • PubMed
    • Google Scholar
30. 1. Hendrix J
    2. Schrimpf W
    3. Höller M
    4. Lamb DC

    (2013) Pulsed interleaved excitation fluctuation imaging

    Biophysical Journal 105:848–861.

    https://doi.org/10.1016/j.bpj.2013.05.059

    • PubMed
    • Google Scholar
31. 1. Hendrix J
    2. Dekens T
    3. Schrimpf W
    4. Lamb DC

    (2016) Arbitrary-region raster image correlation spectroscopy

    Biophysical Journal 111:1785–1796.

    https://doi.org/10.1016/j.bpj.2016.09.012

    • PubMed
    • Google Scholar
32. 1. Hilsch M
    2. Goldenbogen B
    3. Sieben C
    4. Höfer CT
    5. Rabe JP
    6. Klipp E
    7. Herrmann A
    8. Chiantia S

    (2014) Influenza A matrix protein m1 multimerizes upon binding to lipid membranes

    Biophysical Journal 107:912–923.

    https://doi.org/10.1016/j.bpj.2014.06.042

    • PubMed
    • Google Scholar
33. 1. Huang L
    2. Pike D
    3. Sleat DE
    4. Nanda V
    5. Lobel P

    (2014) Potential pitfalls and solutions for use of fluorescent fusion proteins to study the lysosome

    PLOS ONE 9:e88893.

    https://doi.org/10.1371/journal.pone.0088893

    • PubMed
    • Google Scholar
34. 1. Huet S
    2. Avilov SV
    3. Ferbitz L
    4. Daigle N
    5. Cusack S
    6. Ellenberg J

    (2010) Nuclear import and assembly of influenza a virus RNA polymerase studied in live cells by fluorescence cross-correlation spectroscopy

    Journal of Virology 84:1254–1264.

    https://doi.org/10.1128/JVI.01533-09

    • PubMed
    • Google Scholar
35. 1. Hur KH
    2. Chen Y
    3. Mueller JD

    (2016) Characterization of ternary protein systems in vivo with tricolor heterospecies partition analysis

    Biophysical Journal 110:1158–1167.

    https://doi.org/10.1016/j.bpj.2016.01.009

    • PubMed
    • Google Scholar
36. 1. Hutchinson EC

    (2014) Conserved and host-specific features of influenza virIon architecture

    Nature Communications 5:1–11.

    https://doi.org/10.1038/ncomms7446

    • Google Scholar
37. 1. Hwang LC
    2. Gösch M
    3. Lasser T
    4. Wohland T

    (2006) Simultaneous multicolor fluorescence cross-correlation spectroscopy to detect higher order molecular interactions using single wavelength laser excitation

    Biophysical Journal 91:715–727.

    https://doi.org/10.1529/biophysj.105.074120

    • PubMed
    • Google Scholar
38. 1. Işbilir A
    2. Serfling R
    3. Möller J
    4. Thomas R
    5. De Faveri C
    6. Zabel U
    7. Scarselli M
    8. Beck-Sickinger AG
    9. Bock A
    10. Coin I
    11. Lohse MJ
    12. Annibale P

    (2021) Determination of G-protein-coupled receptor oligomerization by molecular brightness analyses in single cells

    Nature Protocols 16:1419–1451.

    https://doi.org/10.1038/s41596-020-00458-1

    • PubMed
    • Google Scholar
39. 1. Jameson DM
    2. Ross JA
    3. Albanesi JP

    (2009) Fluorescence fluctuation spectroscopy: Ushering in a new age of enlightenment for cellular dynamics

    Biophysical Reviews 1:105–118.

    https://doi.org/10.1007/s12551-009-0013-8

    • PubMed
    • Google Scholar
40. 1. Jorba N
    2. Area E
    3. Ortín J

    (2008) Oligomerization of the influenza virus polymerase complex in vivo

    The Journal of General Virology 89:520–524.

    https://doi.org/10.1099/vir.0.83387-0

    • PubMed
    • Google Scholar
41. 1. Kaliszewski MJ
    2. Shi X
    3. Hou Y
    4. Lingerak R
    5. Kim S
    6. Mallory P
    7. Smith AW

    (2018) Quantifying membrane protein oligomerization with fluorescence cross-correlation spectroscopy

    Methods 140:40–51.

    https://doi.org/10.1016/j.ymeth.2018.02.002

    • PubMed
    • Google Scholar
42. 1. Kapusta P
    2. Wahl M
    3. Benda A
    4. Hof M
    5. Enderlein J

    (2007) Fluorescence lifetime correlation spectroscopy

    Journal of Fluorescence 17:43–48.

    https://doi.org/10.1007/s10895-006-0145-1

    • PubMed
    • Google Scholar
43. 1. Kim SA
    2. Heinze KG
    3. Bacia K
    4. Waxham MN
    5. Schwille P

    (2005) Two-photon cross-correlation analysis of intracellular reactions with variable stoichiometry

    Biophysical Journal 88:4319–4336.

    https://doi.org/10.1529/biophysj.104.055319

    • PubMed
    • Google Scholar
44. 1. Kimura S
    2. Noda T
    3. Yoshimori T

    (2007) Dissection of the autophagosome maturation process by a novel reporter protein, tandem fluorescent-tagged LC3

    Autophagy 3:452–460.

    https://doi.org/10.4161/auto.4451

    • PubMed
    • Google Scholar
45. 1. Krieger JW
    2. Singh AP
    3. Bag N
    4. Garbe CS
    5. Saunders TE
    6. Langowski J
    7. Wohland T

    (2015) Imaging Fluorescence (cross-) correlation spectroscopy in live cells and organisms

    Nature Protocols 10:1948–1974.

    https://doi.org/10.1038/nprot.2015.100

    • PubMed
    • Google Scholar
46. 1. Linkert M
    2. Rueden CT
    3. Allan C
    4. Burel J-M
    5. Moore W
    6. Patterson A
    7. Loranger B
    8. Moore J
    9. Neves C
    10. Macdonald D
    11. Tarkowska A
    12. Sticco C
    13. Hill E
    14. Rossner M
    15. Eliceiri KW
    16. Swedlow JR

    (2010) Metadata matters: Access to image data in the real world

    The Journal of Cell Biology 189:777–782.

    https://doi.org/10.1083/jcb.201004104

    • PubMed
    • Google Scholar
47. 1. Magde D
    2. Elson E
    3. Webb WW

    (1972) Thermodynamic Fluctuations in a Reacting System—Measurement by Fluorescence Correlation Spectroscopy

    Physical Review Letters 29:705–708.

    https://doi.org/10.1103/PhysRevLett.29.705

    • Google Scholar
48. 1. Massari S
    2. Desantis J
    3. Nizi MG
    4. Cecchetti V
    5. Tabarrini O

    (2021) Inhibition of Influenza Virus Polymerase by Interfering with Its Protein-Protein Interactions

    ACS Infectious Diseases 7:1332–1350.

    https://doi.org/10.1021/acsinfecdis.0c00552

    • PubMed
    • Google Scholar
49. 1. Müller BK
    2. Zaychikov E
    3. Bräuchle C
    4. Lamb DC

    (2005) Pulsed interleaved excitation

    Biophysical Journal 89:3508–3522.

    https://doi.org/10.1529/biophysj.105.064766

    • PubMed
    • Google Scholar
50. 1. Nilsson-Payant BE
    2. Sharps J
    3. Hengrung N
    4. Fodor E

    (2018) The Surface-Exposed PA51-72-Loop of the Influenza A Virus Polymerase Is Required for Viral Genome Replication

    Journal of Virology 92:e00687-18.

    https://doi.org/10.1128/JVI.00687-18

    • PubMed
    • Google Scholar
51. 1. Padilla-Parra S
    2. Audugé N
    3. Coppey-Moisan M
    4. Tramier M

    (2011) Dual-color fluorescence lifetime correlation spectroscopy to quantify protein-protein interactions in live cell

    Microscopy Research and Technique 74:788–793.

    https://doi.org/10.1002/jemt.21015

    • PubMed
    • Google Scholar
52. 1. Palmer AG
    2. Thompson NL

    (1987) Molecular aggregation characterized by high order autocorrelation in fluorescence correlation spectroscopy

    Biophysical Journal 52:257–270.

    https://doi.org/10.1016/S0006-3495(87)83213-7

    • PubMed
    • Google Scholar
53. 1. Petazzi RA
    2. Aji AK
    3. Chiantia S

    (2020) Fluorescence microscopy methods for the study of protein oligomerization

    Progress in Molecular Biology and Translational Science 169:1–41.

    https://doi.org/10.1016/bs.pmbts.2019.12.001

    • PubMed
    • Google Scholar
54. 1. Ridgeway WK
    2. Millar DP
    3. Williamson JR

    (2012a) Quantitation of ten 30S ribosomal assembly intermediates using fluorescence triple correlation spectroscopy

    PNAS 109:13614–13619.

    https://doi.org/10.1073/pnas.1204620109

    • PubMed
    • Google Scholar
55. 1. Ridgeway WK
    2. Millar DP
    3. Williamson JR

    (2012b) The Spectroscopic Basis of Fluorescence Triple Correlation Spectroscopy

    The Journal of Physical Chemistry. B 116:1908–1919.

    https://doi.org/10.1021/jp208605z

    • PubMed
    • Google Scholar
56. 1. Ries J
    2. Schwille P

    (2006) Studying slow membrane dynamics with continuous wave scanning fluorescence correlation spectroscopy

    Biophysical Journal 91:1915–1924.

    https://doi.org/10.1529/biophysj.106.082297

    • PubMed
    • Google Scholar
57. 1. Ries J
    2. Chiantia S
    3. Schwille P

    (2009a) Accurate determination of membrane dynamics with line-scan FCS

    Biophysical Journal 96:1999–2008.

    https://doi.org/10.1016/j.bpj.2008.12.3888

    • PubMed
    • Google Scholar
58. 1. Ries J
    2. Yu SR
    3. Burkhardt M
    4. Brand M
    5. Schwille P

    (2009b) Modular scanning FCS quantifies receptor-ligand interactions in living multicellular organisms

    Nature Methods 6:643–645.

    https://doi.org/10.1038/nmeth.1355

    • PubMed
    • Google Scholar
59. 1. Rossman JS
    2. Lamb RA

    (2011) Influenza virus assembly and budding

    Virology 411:229–236.

    https://doi.org/10.1016/j.virol.2010.12.003

    • PubMed
    • Google Scholar
60. 1. Sankaran J
    2. Balasubramanian H
    3. Tang WH
    4. Ng XW
    5. Röllin A
    6. Wohland T

    (2021) Simultaneous Spatiotemporal Super-Resolution and Multi-Parametric Fluorescence Microscopy

    Nature Communications 12:1748.

    https://doi.org/10.1038/s41467-021-22002-9

    • PubMed
    • Google Scholar
61. 1. Schrimpf W
    2. Lemmens V
    3. Smisdom N
    4. Ameloot M
    5. Lamb DC
    6. Hendrix J

    (2018) Crosstalk-free multicolor rics using spectral weighting

    Methods 140–141:97–111.

    https://doi.org/10.1016/j.ymeth.2018.01.022

    • PubMed
    • Google Scholar
62. 1. Schwille P
    2. Meyer-Almes FJ
    3. Rigler R

    (1997) Dual-color fluorescence cross-correlation spectroscopy for multicomponent diffusional analysis in solution

    Biophysical Journal 72:1878–1886.

    https://doi.org/10.1016/S0006-3495(97)78833-7

    • PubMed
    • Google Scholar
63. 1. Shaw ML
    2. Stone KL
    3. Colangelo CM
    4. Gulcicek EE
    5. Palese P

    (2008) Cellular proteins in influenza virus particles

    PLOS Pathogens 4:e1000085.

    https://doi.org/10.1371/journal.ppat.1000085

    • PubMed
    • Google Scholar
64. 1. Shi X
    2. Foo YH
    3. Sudhaharan T
    4. Chong SW
    5. Korzh V
    6. Ahmed S
    7. Wohland T

    (2009) Determination of dissociation constants in living zebrafish embryos with single wavelength fluorescence cross-correlation spectroscopy

    Biophysical Journal 97:678–686.

    https://doi.org/10.1016/j.bpj.2009.05.006

    • PubMed
    • Google Scholar
65. 1. Štefl M
    2. Herbst K
    3. Rübsam M
    4. Benda A
    5. Knop M

    (2020) Single-Color Fluorescence Lifetime Cross-Correlation Spectroscopy In Vivo

    Biophysical Journal 119:1359–1370.

    https://doi.org/10.1016/j.bpj.2020.06.039

    • PubMed
    • Google Scholar
66. 1. Sun Y
    2. Wallrabe H
    3. Booker CF
    4. Day RN
    5. Periasamy A

    (2010) Three-Color Spectral FRET Microscopy Localizes Three Interacting Proteins in Living Cells

    Biophysical Journal 99:1274–1283.

    https://doi.org/10.1016/j.bpj.2010.06.004

    • PubMed
    • Google Scholar
67. 1. Weidemann T
    2. Mücksch J
    3. Schwille P

    (2014) Fluorescence fluctuation microscopy: a diversified arsenal of methods to investigate molecular dynamics inside cells

    Current Opinion in Structural Biology 28:69–76.

    https://doi.org/10.1016/j.sbi.2014.07.008

    • PubMed
    • Google Scholar
68. 1. Wohland T
    2. Rigler R
    3. Vogel H

    (2001) The standard deviation in fluorescence correlation spectroscopy

    Biophysical Journal 80:2987–2999.

    https://doi.org/10.1016/S0006-3495(01)76264-9

    • PubMed
    • Google Scholar
69. Preprint

    1. Ziegler M
    2. Yserentant K
    3. Middel V
    4. Dunsing V
    5. Gralak A
    6. Pakari K
    7. Bargstedt J
    8. Kern C
    9. Chiantia S
    10. Strähle U
    11. Herten DP
    12. Wombacher R

    (2020) A Chemical Strategy to Control Protein Networks in Vivo

    bioRxiv.

    https://doi.org/10.1101/2020.04.08.031427

    • Google Scholar

Acceptance summary:

This manuscript provides a new tool to study multi-protein interaction in living cells based on fluorescence fluctuation spectroscopy. The paper includes a reliable validation and description of the method as well as a proof of principle application and assessment of potential limitations.

Decision letter after peer review:

Thank you for submitting your article "Multi-color fluorescence fluctuation spectroscopy in living cells via spectral detection" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewers, and the evaluation has been overseen by Anna Akhmanova as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Erdinc Sezgin (Reviewer #2); Thorsten Wohland (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:

The manuscript is well written and the data very diligently analysed. There remain mainly technical questions and some points that should be discussed to ensure consistency of the results.

1. Figure 1: It seems like the time resolution of the measurements are on the order of several milliseconds. But the diffusion time measured are on the order of ~9 ms. With the time resolution relatively close to the measured diffusion time, did the authors check whether diffusion times are biased?

2. Page 11: Why are correlation amplitudes limited to positive values? The authors mention this also in the supplement but don't explain why that is necessary. For true correlations, only positive values should be obtained. And I assume that is so in their case, if not that should be addressed. Any negative values would be a clear indication that the correlation just indicates noise. This could also be compared with the actual correlation times measured. For correlations of noise, it is not only the amplitude that varies strongly but also the correlation times typically vary widely and mostly do no coincide with expected or reasonable values. The authors might be able to use that criterion to identify non-correlated data and thus avoid the positive amplitudes which are artifacts of the restrictions of the fitting parameters.

3. Page 11 and 12: It would be interesting if the authors could determine whether the observed correlation amplitudes are consistent with the probability of the different FPs (mEGFP, mEYFP, mCherry2 with rel cc amplitudes ranging from 0.45-0.79). If FRET influences these amplitudes, the authors might be even able to extract FRET efficiencies and demonstrate FRET. Another possibility would be to measure fluorescence lifetimes to determine FRET?

4. Page 19/20: The diffusion coefficient for the hetero-tetramer is only about half of the one determined for hetero-dimers. As these are membrane probes with the fluorescent proteins not expected to interact with the membrane, is there an explanation for the factor 2?

5. Did the authors determine the various observation volumes, given the concentrations they measure and the probability of fluorescence of the fluorescent proteins they determined? And is this consistent with the difference in diffusion coefficient they see for the hetero-oligomers?

6. Page 22: The authors indicate that they normalize their RSICS brightness data to measurements of monomers on the same day. It would be interesting if the authors could comment on the day-to-day variability of their calibrations.

7. Figure 3. M2 protein oligomerize but its interacting partner LC3 does not, does that mean each single LC3 associates with multiple M2 protein? Can authors confirm this by looking at the diffusion coefficients from the FCs curves?

8. The authors provided a validation of the method in HEK cells expressing the three Fluorescent Proteins in the plasma membrane in different oligomerization states (Figure 3C). I'm wondering how relatively changes in concentration/expression i.e. fluctuations within the 3 species of probes would influence the observable. Several applications might aim to study interaction within proteins of different abundance, it is important to understand the relative concentration range where the method can be used and provide reliable results. I suggest the authors to provide further experiment and eventually simulation to characterize such dynamic range.

9. The authors applied SFSCS to study how the Influenza A virus matrix protein 2 interact with the autophagy protein LC3 and the tetraspanin CD9. They found that IAV preferentially interact LC3 but not with CD9 (Figure 3). How the position of the label influences the cross-correlation studies?

10. For this technique to be applicable by other researchers, the data analysis tool should be openly available to others. Can authors put their software in an open repository?

11. Since it is a new technique all the limitations of the technique can be discussed in a "limitations" subsection. That would give the readers a clear picture what can and cannot be done with this new technique.

Reviewer #1 (Recommendations for the authors):

Multiplexing methods focusing on dynamic studies are challenging yet very important to understand cellular mechanism at the molecular level. Usually such studies are complex and done on specialized system, this work presents a method easily translatable to commercial system and compatible with standard fluorophores with a well-described analysis pipeline.

Therefore, it can open up several new dynamic studies with 3-4 fluorophores simultaneous read-out.

The choice of fluorophores, its dimerization tendency and the relative labeling densities of each species might influence the cross-correlation observable so a careful validation should be considered and discussed to validate the general applicability of the methodology in various biological applications.

To the author:

1 – The authors provided a validation of the method in HEK cells expressing the three Fluorescent Proteins in the plasma membrane in different oligomerization sates (Figure 3C). I'm wondering how relatively changes in concentration/expression i.e. fluctuations within the 3 species of probes would influence the observable. Several applications might aim to study interaction within proteins of different abundance, it is important to understand the relative concentration range where the method can be used and provide reliable results. I suggest the authors to provide further experiment and eventually simulation to characterize such dynamic range.

2 –The authors applied SFSCS to study how the Influenza A virus matrix protein 2 interact with the autophagy protein LC3 and the tetraspanin CD9. They found that IAV preferentially interact LC3 but not with CD9 (Figure 3). How the position of the label influences the cross-correlation studies?

Reviewer #2 (Recommendations for the authors):

In this paper, the authors developed a new modality of multi-color FCS using spectral detection to investigate stoichiometry of multi-component complexes in live cells. They showed the proof-of-principle with tandem proteins. Furthermore, they showed the biological application of their methodology by investigating Influenza virus components. This will be a useful addition to live cell spectroscopy tools to study protein-protein interactions. The authors' claims are supported by the data throughput the manuscript. The data is analyzed carefully, and results were reported clearly. This method is likely to be used by cell biologists to determine the stoichiometry of multi-protein complexes.

– Figure 3. M2 protein oligomerize but its interacting partner LC3 does not, does that mean each single LC3 associates with multiple M2 protein? Can authors confirm this by looking at the diffusion coefficients from the FCs curves?

– For this technique to be applicable by other researchers, the data analysis tool should be openly available to others. Can authors put their software in an open repository?

– Since it is a new technique all the limitations of the technique can be discussed in a "limitations" subsection. That would give the readers a clear picture what can and cannot be done with this new technique.

Reviewer #3 (Recommendations for the authors):

This manuscript is a carefully conducted study of multi-color fluorescence fluctuation spectroscopy as applied to plasma membranes. By using two-wavelength excitation and spectral detection in a scanning mode the authors show that they can determine the cross-correlation between up to four different probes in a single measurement. In addition, they build on their earlier work and show that the collected data can be analysed by Number and Brightness analysis, providing access to biomolecular stoichiometry. The authors use a range of protein constructs that include between one to four fluorescent proteins, which they use in different compositions to demonstrate that they can analyse all possible interactions of four probes in a single measurement. They then apply the technique to the interaction of influenza A proteins. They show that the influenza A virus (IAV) matrix protein 2 (M2) interacts more strongly with LC3 compared to CD9, both host cell factors. As they measure multiple probes simultaneously, the authors can go beyond binary correlations. Using triple correlations the authors show that they can detect the interaction of the proteins PA, PB1, and PB2 of the IAV polymerase complex. The extension of FCS to four probes, the demonstration of triple correlations and the application to a biological context provides important progress in fluorescence fluctuation spectroscopy and will allow the measurements of complex interactions at the cellular membranes.

The manuscript is well written and the data very diligently analysed. There remain mainly technical questions and some points that should be discussed to ensure consistency of the results.

1. Figure 1: It seems like the time resolution of the measurements are on the order of several milliseconds. But the diffusion time measured are on the order of ~9 ms. With the time resolution relatively close to the measured diffusion time, did the authors check whether diffusion times are biased?

2. Page 11: Why are correlation amplitudes limited to positive values? The authors mention this also in the supplement but don't explain why that is necessary. For true correlations, only positive values should be obtained. And I assume that is so in their case, if not that should be addressed. Any negative values would be a clear indication that the correlation just indicates noise. This could also be compared with the actual correlation times measured. For correlations of noise, it is not only the amplitude that varies strongly but also the correlation times typically vary widely and mostly do no coincide with expected or reasonable values. The authors might be able to use that criterion to identify non-correlated data and thus avoid the positive amplitudes whioch are artefacts of the restrictions of the fitting parameters.

3. Page 11 and 12: It would be interesting if the authors could determine whether the observed correlation amplitudes are consistent with the probability of the different FPs (mEGFP, mEYFP, mCherry2 with rel cc amplitudes ranging from 0.45-0.79). If FRET influences these amplitudes, the authors might be even able to extract FRET efficiencies and demonstrate FRET. Another possibility would be to measure fluorescence lifetimes to determine FRET?

4. Page 19/20: The diffusion coefficient for the hetero-tetramer is only about half of the one determined for hetero-dimers. As these are membrane probes with the fluorescent proteins not expected to interact with the membrane, is there an explanation for the factor 2?

5. Did the authors determine the various observation volumes, given the concentrations they measure and the probability of fluorescence of the fluorescent proteins they determined? And is this consistent with the difference in diffusion coefficient they see for the hetero-oligomers?

6. Page 22: The authors indicate that they normalize their RSICS brightness data to measurements of monomers on the same day. It would be interesting if the authors could comment on the day-to-day variability of their calibrations.

Essential revisions:

The manuscript is well written and the data very diligently analysed. There remain mainly technical questions and some points that should be discussed to ensure consistency of the results.

1. Figure 1: It seems like the time resolution of the measurements are on the order of several milliseconds. But the diffusion time measured are on the order of ~9 ms. With the time resolution relatively close to the measured diffusion time, did the authors check whether diffusion times are biased?

The time resolution of the SFSCS measurements was ca. 0.8 ms, since the line-scan time was ca. 0.4 ms and a 2-line binning was performed. To assess whether this time resolution is sufficient to detect the diffusion dynamics observed in the experiments, we re-analyzed a three-species SFSCS measurement acquired on cells co-expressing membrane-anchored FPs (mp-1x-G + mp-1x-Y + mp-1x-Ch2) using a variable time binning (i.e., effective time resolution varying here from ca. 0.4 ms to 6 ms, see Author response image 1). The obtained diffusion times (between 6 and 9 ms at high time resolution, panel A) and particle numbers (panel B) are very similar for the first three binning modalities and start to decrease above 2 ms binning, indicating that a resolution lower than this value might induce biased estimates. Hence, we conclude that the utilized time resolution of less than 1 ms is sufficient to reliably probe the dynamics observed in our samples, and state this now in lines 981-983 of the manuscript. We have furthermore added a short comment on time resolution in the limitations section (lines 827-830).

Author response image 1

Download asset Open asset

2. Page 11: Why are correlation amplitudes limited to positive values? The authors mention this also in the supplement but don't explain why that is necessary. For true correlations, only positive values should be obtained. And I assume that is so in their case, if not that should be addressed. Any negative values would be a clear indication that the correlation just indicates noise. This could also be compared with the actual correlation times measured. For correlations of noise, it is not only the amplitude that varies strongly but also the correlation times typically vary widely and mostly do no coincide with expected or reasonable values. The authors might be able to use that criterion to identify non-correlated data and thus avoid the positive amplitudes which are artifacts of the restrictions of the fitting parameters.

We apologize for the misunderstanding: The limitation to positive values does not stem from specific boundary settings in the fitting procedure (as it might be indeed inferred by what was written in lines 854 and 248). We have now clarified these passages (lines 278-279, 1024-1029). Briefly, the presence of positive correlation values in noisy curves is a consequence of positive starting parameters for the fit routine and, probably, its converging to a local minimum (i.e., to a positive correlation value).

The (practical) reason for consistently using positive starting values for the correlation was that in the case of clear interactions (e.g. cross-correlation for the positive controls), only positive initial amplitude values resulted in robust and reproducible fitting of the CCF. Negative initial values resulted in errors in the fit routine or fit curves that clearly did not provide a reliable fit to the CFs. To perform an unbiased analysis, the same (positive) initial values were thus used for all samples, sometimes resulting in low “false-positive” amplitudes in the absence of interactions for noisy curves. This in now stated in lines 1024-1029.

We agree with the reviewer that, for the case of pure noise, a symmetric distribution of fit amplitudes around 0 is expected. We have thus implemented an alternative fit routine, which calculates the starting parameter for the amplitude as the average of the first 5 points of each CF. In the case of pure noise, this results in positive or negative fit amplitudes, as mentioned by the reviewer. In the case of interactions, the fit converges to positive values. To evaluate this fit routine, we have reanalyzed a set of three-species SFSCS measurements on cells co-expressing mp-mCherry2-mEGFP hetero-dimers and mp-mEYFP. The new fit procedure (“fit routine 2” in Author response image 2) indeed results in amplitude values that scatter around 0 for non-interacting species (e.g. G,Y; Y,Ch2). This confirms that the previously obtained low cross-correlation does not result from artefacts in the spectral filtering. On the contrary, positive values are obtained for truly correlating samples (e.g. G,Ch2) and they are similar to the results shown in the first version of the manuscript.

We furthermore agree with the reviewer that unreasonable values of the cross-correlation diffusion time can be used to identify correlation of noise. When we apply a threshold of 5 times the maximum diffusion time obtained from ACFs (i.e., considering higher values as an artefact of noise and setting the CCF fit amplitudes of these measurements to zero, for the respective FP combination), some likely false-positive correlation is removed (“fit routine 1, filtered” in Author response image 2). Nevertheless, this procedure requires to set an arbitrary threshold for filtering of the diffusion times, which may be subjective and depend on the specific sample dynamics. Overall, the differences obtained for the different fit procedures are small. For this reason and to keep the analysis comparable with our previously published work, we prefer not to modify the already proposed analysis. Nevertheless, we have added a supplementary figure (Figure 3—figure supplement 2) demonstrating the different fit options so that readers can choose the type of analysis best suited for their experiment. We refer to this figure in the results (lines 278-279). We thank the reviewer for bringing up this important issue.

Author response image 2

Download asset Open asset

3. Page 11 and 12: It would be interesting if the authors could determine whether the observed correlation amplitudes are consistent with the probability of the different FPs (mEGFP, mEYFP, mCherry2 with rel cc amplitudes ranging from 0.45-0.79). If FRET influences these amplitudes, the authors might be even able to extract FRET efficiencies and demonstrate FRET. Another possibility would be to measure fluorescence lifetimes to determine FRET?

For FPs with a probability p_f to be fluorescent, the expected rel.cc. is equal to p_f, as we discuss in the Appendix of the paper (Appendix-table 1). Since all FPs (mEGFP, mEYFP, mCherry2) exhibit similar p_f of ≈70%, the expected rel.cc. is ≈0.7 for all FP pairs. The values obtained for cross-correlation of mEGFP and mEYFP signals (with a rigid linker) are indeed close to the expected value (e.g. rel.cc.=0.72 for mEYFP-mEGFP hetero-dimers in 2-species SFSCS, Figure 1). A significantly lower rel.cc. of 0.60 was obtained with a short linker peptide between the two FPs, suggesting the presence of FRET (see Figure 1—figure supplement 3). For the other FP combinations requiring excitation with two laser lines, the obtained rel.cc. values are lower than expected (e.g. rel.cc.=0.56 and 0.57 for mEGFP/mCherry2 and mEYFP/mCherry2 in heterotrimers). In addition to non-fluorescent states, differences in observation volumes and FRET can reduce the experimentally observed rel.cc., as pointed out by the reviewer. We address the issue of imperfect overlap below, in the response to question 5. To minimize FRET artifacts, FPs in hetero-trimers and tetramers were linked by a rigid linker. Nevertheless, residual FRET may still occur. To assess FRET between FPs, we quantified the molecular brightness of FPs coupled in hetero-oligomers and compared it with the brightness obtained on the same FPs expressed as independent monomers. The brightness ratios q_i (i=G,Y,Ch2) are shown below for SFSCS measurements on cells expressing mp-mEYFP-mCherry2-mEGFP hetero-trimers (Author response image 3) or coexpressing mp-mCherry2-mEGFP hetero-dimers and mp-mEYFP (Author response image 4). FRET efficiencies (f_E) are related to the decrease in donor brightness, f_E=1-q_G (Foo et al., BJ, 2012). We observed a systematic reduction of the brightness of mEGFP, no (Author response image 4) or small (Author response image 3, hetero-trimer) changes of the mEYFP brightness, and a consistent increase of mCherry2 brightness. This indicates residual FRET between, e.g. mEGFP donor and mCherry2 acceptor. Based on the experimental q values, we can calculate the expected rel.cc. assuming p_f≈0.7 for all 3 FPs, using the formulas derived in Foo et al., BJ, 2012. We obtain rel.cc. values of 0.63 and 0.69 for the FP combination most affected by FRET (mEGFP/mCherry2), for hetero-dimers and hetero-trimers respectively. Thus, FRET has a minor effect (<10%) on the cross-correlation and alone cannot explain the reduced rel.cc. observed for pairs of green/yellow and red FPs. It should be noted that, in the case of hetero-trimers, our analysis based on the effective molecular brightness is a simplified approximation, since all three FP pairs may undergo FRET to different extent. We agree with the reviewer that it is important to keep FRET into consideration when planning experiments and constructs, in particular because hetero-oligomers are commonly used as a positive control to normalize the rel.cc. obtained on proteins of interest. We have therefore added a comment on FRET to the Limitations section (lines 864-869).

Author response image 3

Download asset Open asset

Author response image 4

Download asset Open asset

4. Page 19/20: The diffusion coefficient for the hetero-tetramer is only about half of the one determined for hetero-dimers. As these are membrane probes with the fluorescent proteins not expected to interact with the membrane, is there an explanation for the factor 2?

We think that this comment might arise from a misunderstanding. The data shown on Page 19/20 were obtained by RSICS measurements in the cytoplasm, not on the plasma membrane. Thus, diffusion coefficients reasonably depend on oligomer size. The obtained values for the diffusion of FP oligomers, e.g. diffusion of FP tetramer about 2.5 times slower than diffusion of monomers, are in good agreement with previous studies (e.g. Dross, PlosOne, 2009, Erdel Nat. Comm. 2014, Dunsing, Sci. Rep., 2018). We have clarified this issue in lines 447 and 467.

5. Did the authors determine the various observation volumes, given the concentrations they measure and the probability of fluorescence of the fluorescent proteins they determined? And is this consistent with the difference in diffusion coefficient they see for the hetero-oligomers?

We have already taken imperfect overlap of the observation volumes into account for comparison of the experimental triple correlation with simulation results (see results and Appendix for triple correlation) but have not determined the size and potential displacement of observation volumes explicitly.

As mentioned in our response to point 3, differences in observation volumes for different FP channels may cause the lower than expected rel.cc. obtained for pairs of green/yellow and red FPs, e.g. mEGFP/mCherry2. To evaluate such differences, we have analyzed the relative ACF amplitudes and diffusion times obtained from SFSCS measurements on FP hetero-oligomers, i.e. mp-mEYFP-mCherry2-mEGFP hetero-trimers and mp-mEGFP-mCherry2 hetero-dimers in the different channels (e.g. G,Y,Ch2 and G,Ch2). As shown in Author response image 5, the data spread for the diffusion times of the hetero-trimer is too large to draw definitive conclusions. Therefore, we have focused on mp-mCherry2-mEGFP hetero-dimers and started evaluating the relative ACF amplitudes. In addition to differences in observation volumes, the amplitudes are also affected by p_f (e.g. lower p_f corresponds to higher amplitude). For all FPs evaluated here, p_f values are similar (ca. 0.7, see RSICS data in Figure 6 and previously published data in Dunsing et al., Sci. Rep., 2018). Thus, if the observation volumes were the same, the relative amplitudes should be close to 1, as shown indeed in Author response image 6. This indicates that differences in observation volumes or p_f are small. More precisely, one would expect a slightly larger observation volume for mCherry2 than for mEGFP (due to the higher wavelength) and thus an amplitude ratio G_Ch2/G_G smaller than 1. However, values slightly larger than 1 are obtained. This might be due to e.g. slightly smaller p_f of mCherry2. In addition, residual FRET (see point 3) may increase the ACF amplitude of the acceptor fluorophore, here mCherry2 (see Foo et al., BJ, 2012).

Due to the difficulties in interpreting the ACF amplitudes, we have therefore directly evaluated the diffusion times of mp-mCherry2-mEGFP hetero-dimers obtained from ACFs and from CCFs. In scanning FCS, the diffusion times scale with the effective observation area A_eff=Π∙S∙w₀². The results are shown in Author response image 7. The significant increase in diffusion time for mCherry2 (taud_Ch2/taud_G=1.35) indicates a ca. 35% difference in size and a slight shift of observation volumes (since the diffusion time obtained from the CCF does not fall in between the values obtained from ACFs for taud_Ch2 and taud_G, see the discussion in Foo et al., 2012). Calculating the effective overlapping observation area from the cross-correlation diffusion time (taud_G,Ch2=1.38*taud_G), the determined FRET efficiencies (see q factors in response to point 3, q_G=0.8, q_R=1.3) and keeping the p_f values (0.7 for both) into account, we calculated an expected rel.cc. of ca. 0.50 for mp-mCherry2-mEGFP hetero-dimers, which is very close the experimental value of 0.45. Thus we conclude that the deviation from the expected rel.cc. (solely based on p_f, i.e. ca. 0.7) for FP combinations involving excitation by two laser lines in general might be caused by both, FRET, and imperfect overlap of the observation volumes.

We have added the issue of imperfect overlap, which applies to SFSCS as to standard FCCS as soon as a second excitation line is used, to the limitations section (lines 824-826) and state it more clearly also in the discussion (lines 670-674). Therein, we stress the point that for two-species measurements, the fact that only one laser line is needed provides a clear advantage of the spectral approach, resulting in higher rel.cc. (up to 0.8).

Author response image 5

Download asset Open asset

Author response image 6

Download asset Open asset

Author response image 7

Download asset Open asset

6. Page 22: The authors indicate that they normalize their RSICS brightness data to measurements of monomers on the same day. It would be interesting if the authors could comment on the day-to-day variability of their calibrations.

We thank the reviewer for this comment and agree that data on day-to-day variability are highly useful and often underreported in brightness studies. We have examined both the day-to-day variability of the relative brightness values of the homo-dimer control and of the absolute brightness of the monomeric reference on the same day. The data has been added to the SI and refer to in the results (lines 591-592).

7. Figure 3. M2 protein oligomerize but its interacting partner LC3 does not, does that mean each single LC3 associates with multiple M2 protein? Can authors confirm this by looking at the diffusion coefficients from the FCs curves?

Yes, we think that LC3 monomers bind to M2 oligomers in (apparent) average stoichiometry of 1:2 to 1:4. In the discussion of the paper, we speculate already that LC3 does not show higher oligomerization (or higher stoichiometry in the binding to M2) because endogenous LC3 or other cellular proteins might block additional binding sites in M2 oligomers, or because of steric hindrance. We have evaluated the diffusion dynamics of M2 and LC3 (see Author response image 8). Although the rel.cc. between M2 and LC3 is similar to the positive control (indicating that most LC3 is present in complexes with M2), the diffusion dynamics of LC3 are consistently faster (0.03± 0.01 ms) than the dynamics of M2 (0.06±0.01 ms). This is to be expected, since LC3 is a membrane associated protein and its apparent diffusion dynamics is likely influenced by the dynamics of association/dissociation rates (in addition to actual diffusion at the PM). Therefore, we think that the apparent diffusion dynamics of LC3 are difficult to interpret and cannot be used directly to explore M2-LC3 interactions. We have also analyzed the diffusion times obtained from the LC3-M2 CCFs. They show a fairly large variation due to noise but the majority of data points report similar diffusion dynamics (0.07±0.06 ms) as M2 oligomers, which would be expected for binding of a membrane associated to a transmembrane protein.

Author response image 8

Download asset Open asset

8. The authors provided a validation of the method in HEK cells expressing the three Fluorescent Proteins in the plasma membrane in different oligomerization sates (Figure 3C). I'm wondering how relatively changes in concentration/expression i.e. fluctuations within the 3 species of probes would influence the observable. Several applications might aim to study interaction within proteins of different abundance, it is important to understand the relatively concentration range where the method can be used and provide reliable results. I suggest the authors to provide further experiment and eventually simulation to characterize such dynamic range.

We thank the reviewer for this comment and agree that the relative concentration range that provides reliable results is important to explore. We have therefore already carefully evaluated the SNR of two-species SFSCS measurements as a function of the relative concentrations (i.e. intensities), which we show in Figure 1. In addition, we have extensively discussed the allowed relative concentration range for 2-,3-, and 4-species measurements in the discussion and methods section of the paper (e.g. 1:5 for neighboring FP species in 4-species measurements), and have compared it to the limitations of similar approaches such as fluorescence lifetime correlation spectroscopy. In order to explicitly address the question for three-species SFSCS (e.g. Figure 3), we have now analyzed the SNR of the ACFs obtained from measurements on cells co-expressing mp-mEGFP, mp-mEYFP and mp-mCherry2 (see Author response image 9). The SNR is plotted as a color code in 2-d graphs, for each species as a function of the relative intensity to the other two species. Similar to two-species SFSCS, the SNR of mEYFP (panel B) is mostly compromised, by both signal from mEGFP (horizontal axis) and mCherry2 (vertical axis), i.e. a 6-fold drop of SNR for relative signals ranging from ca. 6:1 (yellow points) to ca. 1:6 (dark blue points). On the contrary, mCherry2 (3-fold change of SNR, panel C) and mEGFP (2-fold change of SNR, panel A) are only moderately affected. As expected, the SNR ratio of mCherry2 ACFs (panel C) depends stronger on mEYFP (vertical axis) than on mEGFP signal (horizontal axis), which is caused by the higher overlap with mEYFP emission. We have added the figure as supplement (Figure 3—figure supplement 3) and refer to it now in the discussion (lines 696-697, 728, 730, 734).

Given that mEYFP is the most compromised of the three species, we have then evaluated the diffusion time and molecular brightness obtained for mEYFP in three groups corresponding to the highest, middle and lowest third of the 31 data points, sorted by SNR. The obtained results (Author response image 1) are very similar and do not show significant differences in the average values or variation of parameter estimates. This shows that the SFSCS analysis works reliably in the here explored concentration range of 1:6 to 6:1 for neighboring FP channels. As we state in the discussion, the relative concentration range may depend on the specific sample, fluorophores and detection efficiency, and thus needs to be determined carefully. We have added an additional comment on this issue to the new limitations section (lines 833-841).

Author response image 9

Download asset Open asset

Author response image 10

Download asset Open asset

9. The authors applied SFSCS to study how the Influenza A virus matrix protein 2 interact with the autophagy protein LC3 and the tetraspanin CD9. They found that IAV preferentially interact LC3 but not with CD9 (Figure 3). How the position of the label influences the cross-correlation studies?

To minimize the influence of the fluorophore tags on our interaction study, we chose a reasonable labeling approach. The M2 protein was tagged at the extracellular N-terminus. Binding with LC3 occurs via an LIR-motif in the flexible cytoplasmic tail of M2 (Beale et al., Cell Host and Microbe 2014). LC3 docks with the LIR motif via hydrophobic pockets located in the middle region of its sequence (Hamacher-Brady and Brady, Cell. and Mol. Life Sciences, 2016). We placed the mEYFP tag at the C-terminus of LC3, i.e. at some distance from the binding region. Thus, we do not expect interference of the FP tags with M2-LC3 interactions.

For CD9, both N- and C-terminus and intracellular. The mEGFP-tag is located at the C-terminus, with a 10aa linker. Having the tags on M2 and CD9 on different sides of the plasma membrane (EC/IC) should minimize potential hindrance in the interactions. Binding e.g. via the TMDs should not be affected. Given the flexibility of the cytoplasmic tail of M2, we do not think that the tag on CD9 hinders M2-CD9 interactions. Nevertheless, we cannot rule this out completely and thus added a sentence to the manuscript (lines 776-778). In addition, we have added a discussion on the general requirements of fluorophore tags and potential interference with protein-protein interactions to the Limitations section (lines 848-853).

10. For this technique to be applicable by other researchers, the data analysis tool should be openly available to others. Can authors put their software in an open repository?

We agree that it would be very beneficial to have the code openly available. We have therefore uploaded all analysis code (SFSCS, RSICS, and TRICS) and a documentation to run the main scripts on GitHub (https://github.com/VaDu8989/SpectralFFS).

11. Since it is a new technique all the limitations of the technique can be discussed in a "limitations" subsection. That would give the readers a clear picture what can and cannot be done with this new technique.

We thank the reviewer for this idea and have added a “Limitations” subsection at the end of the discussion, summarizing the main technical and conceptual limitations of the technique.

Author details

1. Valentin Dunsing

   Universität Potsdam, Institute of Biochemistry and Biology, Potsdam, Germany

   Contribution

   Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Visualization, Writing - original draft, Writing – review and editing

   For correspondence

   valentin.dunsing@gmx.de

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-2482-1498

2. Annett Petrich

   Universität Potsdam, Institute of Biochemistry and Biology, Potsdam, Germany

   Contribution

   Investigation, Resources, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-2372-826X

3. Salvatore Chiantia

   Universität Potsdam, Institute of Biochemistry and Biology, Potsdam, Germany

   Contribution

   Conceptualization, Formal analysis, Funding acquisition, Methodology, Project administration, Resources, Software, Supervision, Visualization, Writing - original draft, Writing – review and editing

   For correspondence

   chiantia@uni-potsdam.de

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-0791-967X

• 1,962

  Page views

• 279

  Downloads

• 8

  Citations

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