Evaluating the degree of molecular contact between cellulose fiber surfaces using FRET microscopy

The following brief introduction to Förster Theory was taken from the book by B.W. Van der Meer (Meer 2013) and is based partly on the original papers by Förster. The efficiency of the energy transfer between donor and acceptor dye depends on the distance between the two molecules, and is in principle determined by the Förster Radius (\(\text {R}_0\)). The Förster Radius can be calculated for every combination of donor and acceptor molecules as can be seen in the supplementary information (Table S1). \(\text {R}_0\) is specific for every donor-acceptor pair in a particular system. The efficiency of the energy transfer (\(\eta _{eff}\)) between donor and acceptor is given by Eq. 1:where r stands for the distance between the donor and acceptor and \(\text {R}_0\) stands for the Förster radius of the donor-acceptor pair. Practically, the FRET efficiency can be measured by different methods which can be implemented in either microscopy setups or measured by spectrophotometry, both of which were used in this paper.

Regardless of the measurement method, a FRET process can be detected by different aspects of the effect. FRET changes the Donor emission intensity (Donor quenching), the acceptor emission intensity (Acceptor sensitation), the lifetime of donor fluorescence (Lifetime measurements) and the polarization of the light (Anisotropy measurements). Here we used two of the more common methods to measure FRET, namely donor quenching (DQ) and acceptor sensitation (AS). Donor quenching measures the decrease of the donor fluorescence due to FRET. Acceptor sensitation measures the increase of the acceptor fluorescence due to FRET. While DQ is an indication for FRET, one cannot be certain of it, as there are other mechanisms such as concentration quenching that can deactivate the excited donor. AS on the other hand provides conclusive evidence for FRET as the acceptor fluorescence can only be increased by some sort of energy transfer (Meer 2013). An excellent review about implementation and the pitfalls of the technique is provided by Broussard et al. (2013). One crucial part of using FRET that should be mentioned here is to meticulously check if the signal is indeed valid especially when spectroscopic data is not available. This can be done by using negative controls, dynamic FRET or applying more than one FRET technique.

The dyes 7-(diethylamino)coumarin-3-carbohydrazide (DCCH, Purity \(95{\%}\), CAS: 100343-98-4) and fluorescein-5-thiosemicarbazide (FTSC, Purity \(99{\%}\), CAS: 76863-28-0) were bought from Santacruz Biotechnology. Solvents N,N-dimethylformamide (DMF), Tetrahydrofuran (THF), and Acetonitril were purchased from VWR. All chemicals were used without further purification. Poly(2-hydroxyethyl methacrylate) (pHEMA) for manufacturing of thin films was purchased from Sigma Aldrich. The model films were deposited on PET foils. The pulp used was an unbleached softwood Kraft pulp from industrial production.

For investigating pulp fibers [bleached Kraft wood, ratio spruce/pine (90/10)], the samples were prepared as described by Thomson et al. (2007). Fibers were dyed in 15 mL of a solution of either dye in DMF (1 mmol/L and 0.2 mmol/L) overnight at a pH of 5, adjusted by adding HCl. The dyes form a meta-stable bond with the reducing end of the cellulose molecule and by adding HCl this equilibrium is shifted towards the meta-stable bond species side. After the dyeing, sodium borohydrate \(\hbox {NaBH}_4\) (0.02 mmol \(\text {NaBH}_4\) per 0.5 g fibers) was added immediately to the solution and allowed to react for 1 h. The sodium borohydrate leads to a reductive amination of the dyes with the cellulose which results in a covalent bond between the reagents. After this reductive amination the samples were first washed in DMF and again washed by Soxhlet extraction with acetonitrile for at least 4 h to remove any non-covalently bonded dye. Subsequently the dyed fibers were stored in an aqueous pH 9 buffer solution using 0.4 g/L borax. The fiber bonds were prepared by crossing one DCCH dyed and one FTSC dyed fiber in a droplet of water (pH 9) and subsequently drying them in a Rapid Köthen sheet former. For microscopy the fiber bonds were then transferred to microscopy slides. Fiber Crossings (negative control for FRET) were prepared by simply crossing two fibers and covering them with a microscopy cover slip.

The model films were prepared by doctorblading \(500\, {\upmu }\hbox {L}\) of a \(10{\%}\) pHema in a 95/5 EtOH/H2O solution mixed with dyes dissolved in THF onto PET foils. This resulted in an approx. \(1.5\,{\upmu }\hbox {m}\) thick film. To ensure alkaline conditions a volume of \(12\,{\upmu }\hbox {L}\) triethylamine was added to the solution. The PET foils were cleaned with acetone before the doctorblading. FRET was investigated in two different model systems. In the first trial (Mixture sample) the dyes were added to the film as follows. Donor in pHema, Acceptor in pHema and Donor + Acceptor in pHema. The interaction of molecules within this film was investigated. In a second step (Bonds and Crossings) two films (Donor in pHema, Acceptor in pHema) were prepared, stacked together and left at \(1\,\hbox {bar}\) pressure at \(93\,^\circ \hbox {C}\) for \(3\,\hbox {h}\) to form contact. The concentrations and applied measurement methods for the samples can be seen in Table 1.

Fluorescence spectra were recorded on a RF-5301PC, spectrofluorophotometer from Shimadzu. FRET experiments were performed with concentrations of molecules seen in Table 1. Also care was taken to minimize the exposure of any of the samples to ambient light to avoid photo bleaching. The model films were on the one hand investigated using this fluorometry setup and on the other hand they were investigated using a wide field microscope with the filter sets seen in Table 2. The spectra of the pHema films were recorded using the setup seen in Fig. 3. The samples were investigated under an angle that avoids the detection of the directly reflected light and only measures the fluorescence coming from the sample.

The efficiency of the FRET energy transfer is calculated from the measured spectra and is being explained in Fig. 4. Due to the energy transfer the donor emission is quenched (\(\text {I}_{DA}\)) with respect to the normal emission of the donor (\(\text {I}_{D}\)). The energy not emitted by the donor is then transferred to the acceptor and increases (or sensitizes) the emission (\(\text {I}_{AD}\)) compared to the normal acceptor sensitation (\(\text {I}_{A}\)). The spectral overlap seen in Fig. 4 is a necessary requirement for FRET and determines to a large extent the distance range of the effect as can be seen in Eq. 2 .\(\text {R}_0\) is the Förster radius, k is the orientation factor, n the refractive index, (\(\text {Q}_{D}\)) the quantum yield of the donor and J the overlap integral. The Förster radius determines the distance of the energy transfer as it is known that quantification of FRET is only possible within distances ranging from \((\frac{1}{2} - 2) \text {R}_0\) as also depicted in Fig. 2. A detailed description on how to determine the Förster radius can be found in FRET—Förster Resonance Energy Transfer by Medintz and Hildebrandt (Meer 2013).

The fluorescence emission spectra of DCCH and FTSC overlap to a certain extent as seen in Fig. 7. Therefore, to get an accurate measurement of the FRET efficiency the spectra need to be spectrally unmixed. To do this in the spectrometer we fitted the emission peaks and used the area under the peaks as the intensity of the fitted signal. Practically this was done by fitting Gaussian peaks to the recorded emission spectra of a pure donor sample (resulted in \(\text {I}_{D}\)), a pure acceptor sample (resulted in \(\text {I}_{A}\)) and a sample of interest (resulted in \(\text {I}_{DA}\) and \(\text {I}_{AD}\)) such as in Fig. 11. The full fitting parameters can be found in the supplementary information (Figure S3 and Table S2). The resulting intensity information was used with Eqs. 3 and 4 to get the FRET efficiency. For the donor quenching (Eq. 3) and acceptor sensitation (Eq. 4) the following equations were used.\(\epsilon _{A}\) and \(\epsilon _{D}\) are the extinction coefficients of the acceptor and donor at the excitation wavelength (Meer 2013; Clegg 1992, 1995). The ratio of \(\frac{\epsilon _{A}}{\epsilon _{D}}\) was determined to be 0.02 and can be found in the supplementary information (Table S1). Acceptor sensitation and donor emission are sketched in Fig. 4.

The paper fibers were investigated using a wide field microscopy (WM) setup equipped with the filter sets seen in Table 2. The WM was operated with a 50 W halogen lamp and a CMOS detector from QI Imaging (Optimos).

Both the intensity of the lamp and the detector sensitivity show a dependency of the wavelength and were corrected by calculating correction factors from the lamps emission spectrum folded with the excitation filters and the extinction coefficient and the detector sensitivity folded with the emission filters and the emission spectra. These correction factors were applied to the recorded images to get a true representation of the measured intensity.

A schematic of the dyed fibers and pHema films can be seen in Fig. 5. The images are false colored to emphasize the differences in the fibers/films. The samples were investigated using a \(400\times\) magnification in the case of fibers and \(50\times\) magnification for the case of films. To minimize background noise, the microscope was used in a box surrounded by black fabric.

When using the microscope one trades the spectral for spatial information and one loses the advantage to apply this simple fitting method described above. Here, a sophisticated algorithm that results in a fully corrected measure of the FRET efficiency has been developed by Gordon et al. (1998). The method makes use of images recorded with the three different filter sets from Table 2 of three different samples. Usually this would result in a total of nine images but with the arrangement seen in Fig. 5 it is possible to get the same information from only three images instead as in every image there is automatically a pure donor, a pure acceptor and the FRET area included. For a detailed description of the algorithm please refer to the original paper (Gordon et al. 1998). In brief this method calculates the FRET intensity corrected for all possible spectral bleed-through scenarios according to the following equations.The equations consist of variables with 2 letters. The first letter stands for the used filter set as seen in Table 2, the second letter for the investigated sample (d = donor only, a = acceptor only, f = FRET area). E.g. Af therefore stands for the FRET region (bonded area) investigated with the acceptor filter set. The variables represent the measured light intensities from the aforementioned microscope images recorded as 16 bit gray values. \({\overline{Afa}}\) refers to the acceptor signal that would have been if no donor were present and therefore no FRET occurred. Similarly, \({\overline{Dfd}}\) refers to the donor signal that would have been if no acceptor were present and therefore no FRET occurred. Regions of interest were chosen in the images as depicted in Fig. 5b. The evaluation was performed pixelwise. G is a factor relating the loss of the donor signal to the increase of the acceptor signal (see Supplementary information Table S3). In this work we used the slightly different algorithm by Xia et al. which differs solely in the normalization of the \(\text {N}_{\text {FRET}}\) value but is otherwise equivalent to the Gordon algorithm (Xia and Liu 2001).

For being able to reproduce the area in molecular contact a method was developed to always choose the appropriate area. The method consisted of manually drawing an ROI (region of interest) capturing the “optically-bonded” area in Fig. 6. Subsequently we applied image erosion to avoid the edges and thus obtain the Eroded ROI (Fig. 6, bottom images) which was then used for evaluation of FRET intensity. Avoiding the edges was necessary because regions of extreme intensity appear to give a false positive FRET signal. Additionally, false positives were also detected on areas that are impossible to exhibit FRET such as the area seen in Fig. 6a that is not in the bonded region.

It should be mentioned here that other FRET experiments were also employed. A confocal laser scanning microscope (CLSM, Leica TCS SPE) with a photomultiplier tube from Hamamatsu was used to investigate acceptor photobleaching and sensitized emission of the acceptor in fiber–fiber bonds. The idea was that due to the superior setup a better signal to noise ratio will be achieved.

To check that the combination of dyes can be used as a FRET system on the chosen substrates and the chosen pH value the excitation and emission spectra of the dyes on both substrates at a pH of 9 were investigated. An optimal setup would have a significant spectral overlap between the donor emission spectrum and the acceptor excitation spectrum (Overlap Integral, necessary condition for FRET) while having as little spectral bleed-through as possible. The resulting spectra can be seen in Fig. 7. As can be seen the donor emission and the acceptor excitation spectra show significant overlap. However also there is partial spectral overlap of the donor emission and the acceptor emission and vice versa. Also the acceptor excitation spectrum overlaps with the donor excitation spectrum. This spectral bleed through in the fluorescence microscopy is the reason for the necessity of the Xia algorithm in the optical microscopy and the spectral unmixing in the spectrophotometry to correct the signal. When comparing the dyes in the different substrates one can see some changes. First one can see that the excitation spectra of both dyes change at approx. \(400\,\hbox {nm}\) when comparing the fibers and pHema. Additionally, the FTSC spectra slightly shift in wavelength. These changes are attributed to the covalent linking and therefore chemical modification of the dye molecules. That the covalent bonding has an effect on the dyes can also be seen in the quantum yield reported in the supplementary information (Table S1). The changes are low enough to be able to compare the two systems with each other.

Paper fiber bonds and crossings were investigated to see if it is possible to measure the area in molecular contact between two paper fibers using FRET fluorescence microscopy. When investigating fiber bonds one would expect a FRET signal between the fiber surfaces because the two surfaces and therefore also the dye molecules come very close. To check the plausibility of the signal, unbonded fiber crossings were investigated as a reference. As can be seen in Fig. 8 for both dye concentrations (0.2 mmol/L and 1 mmol/L) it is not possible to detect a significant difference between the bond and the crossing. For illustration also refer to Fig. 6b where a considerable false FRET signal can be seen in the overlapping (but unbonded) regions of the fiber crossing. A possible concentration dependence of the FRET signal was also investigated but did not show any improvement. These results show that it is not possible to measure the interaction of two paper fiber surfaces by using the employed method.

One reason for not being able to detect a difference in the fibers could be a low signal to noise ratio. If the dyes stay only at the surface of the fibers the signals can only originate from a thin layer of interaction. If, however the complete bulk is dyed it would likely lead to a disadvantageous signal to noise ratio. To investigate this the cross sections of the fibers were looked at by using a confocal laser scanning microscope. The results of this investigation seen in Fig. 9 show that the complete bulk of the paper fibers with a thickness of up to \(10\,{\upmu }\hbox {m}\) was dyed by the employed method. This continuous dyeing of a \(10\,{\upmu }\hbox {m}\) thick bulk will lead to a very low signal to noise ratio as the following argumentation shows. The energy transfer due to FRET is only possible up to 2 times the Förster distance. In our case the Förster distance was measured to be about \(6\,\hbox {nm}\). (supplementary information Table S1) This means that a minimum distance of \(12\,\hbox {nm}\) is needed to get a FRET signal. In other words this means that when our paper fibers are bonded a thickness layer of approx. \(10\,\hbox {nm}\) is interacting. Comparing this to a fiber thickness of up to \(10\,{\upmu }\hbox {m}\) this results in a signal to noise ratio of 1:1000. Therefore, if there is even only little spectral bleed-through it will strongly interfere with the FRET signal. This argumentation shows that if one wants to measure a FRET signal between two surfaces it is necessary to maximize the signal to noise ratio either by using thinner substrates or by confining the dyes to the region of interaction.

Another way to see whether a FRET experiment is successful or not is to investigate the dynamic development of the FRET signal. In our case we tried to replicate the experiments done by Thomson et al. (2007) in which the drying of a fiber bond was monitored. During paper drying the capillary forces of the retracting water are pressing the fiber surfaces together, thus creating the molecular contact area that facilitates adhesion between the fiber surfaces (Persson et al. 2012). In this investigation 6 fiber bonds were investigated that all show a similar behavior as to the one seen in Fig. 10. The collected measurements can be found in the supplementary information (Figure S2). As can be seen in Fig. 10a the FRET intensity is largest below the “dry point” . The dry point is the time at which no water could be detected in the microscopy images. After this point the \(\text {N}_{\text {FRET}}\) drops to almost zero and stays there. Hence, we were observing the opposite of what one would expect. In the wet state where the fibers are certainly not bonded the FRET signal is the strongest and decreases to zero during the drying where the signal should increase. In Fig. 10b one can see that the spectroscopic bleed-through between the filter sets changes over the drying period which is a partial reason for the unexpected development of the \(\text {N}_{\text {FRET}}\) characteristic. Since the bleed-through values are a measure of the spectroscopic properties of the dyes this means that as the water is reduced the excitation and emission spectra of the dyes bound to the fibers appear to be changing. In other words, this means that during drying the properties of the dyes are changing as well which makes a comparable analysis between measurements difficult. The reason for this change could be the change in pH when the amount of water is going down and thus the chemicals concentration is changing. Another reason for the trend could be the worse signal to noise ratio when investigating structures in water due to the stronger scattering of the light.

In summary both experiments aiming to validate the FRET system at hand did not show an increase in FRET signal between bonded and unbonded fibers. Thus the investigated FRET technique seems not suited to provide qualitative or quantitative information on the molecular contact area related to the adhesion between the fibers. The possible reasons for this is firstly and most likely the poor signal to noise ratio but also reflection and scattering of light on the fibers proves to be a difficult obstacle to overcome as it apparently leads to artifacts in the FRET signal.

Our next step was to see whether it is possible to measure a FRET signal using the two dyes mixed together within one film. This measurement serves as a proof of concept that the employed dye system is in fact applicable as a FRET pair. The results are presented in Fig. 11a. There one can see the clear behavior of a FRET system. In the presence of the acceptor the donor decreases (donor quenching \(\text {I}_{D}\) to \(\text {I}_{DA}\)) in magnitude and transfers its energy to the acceptor (acceptor sensitation \(\text {I}_{A}\) to \(\text {I}_{AD}\)) which shows a strong increase. The curves in Fig. 11 were fitted with Gaussian peaks and the values for \(\text {I}_{D}\), \(\text {I}_{A}\), \(\text {I}_{DA}\) and \(\text {I}_{AD}\) were extracted and used for the calculation of the FRET efficiencies, Eqs. 3 and 4, results are seen in Table 3, line 1. As presented a clear FRET response was measured using the spectrofluorometry.

In another experiment we wanted to see if we can measure a FRET interaction between two initially separated films as depicted in Fig. 2c. To achieve this, pHema films dyed with either FTSC or DCCH were produced. In a next step these films were subjected to the bonding procedure described in the experimental section. Similar to the fibers we produced bonds and crossings as a reference. If the system works the bonds should exhibit a significantly higher FRET signal than the crossings but should not exceed the mixture as this is the maximum signal that one should be able to achieve at a constant concentration of dyes in the films. In Fig. 11b one can see the results for the bonded films. As expected, the FRET signal is stronger than the FRET response in Fig. 11c but weaker than in the case of the mixed film. In Fig. 11c one can see the spectroscopy measurement of the crossed but not bonded thin films. This measurement serves as a reference for unbonded surfaces as in this case the dyes are not capable of interacting with each other. This result is a good example on why it is not sufficient and reliable to only use donor quenching as an indication for FRET. As can be seen in Fig. 11d the result from the donor quenching indicates a higher FRET signal for the crossings while the FRET efficiency from the acceptor sensitation correctly shows a lower signal than the bonded films. The reason for the lower reliability of donor quenching is because a reduction in the donor signal can come from other reasons than energy transfer such as environmental quenching (Meer 2013). Here the lower donor signal is likely due to the gap of air between the thin films that only exists in the crossed but unbonded samples. Although measured with the same instrument settings, the overall intensity of the crossed films is much lower than in the within and bonded samples as seen in Fig. 11. Therefore, it is necessary to use the acceptor sensitation as a more reliable measure for FRET.

One reason why the pHema system works could be the increased concentration of dye molecules in the films compared to the fibers. The concentrations of dye molecules in the pHema film can be seen in Table 1. The concentration of the fibers was estimated by measuring the intensity of the dyed fibers and comparing it to the thin films with a known concentration. Using the measured quantum yield and thickness of the systems the concentrations in Table 2 were calculated. As can be seen the concentrations differs by a factor of approx. 15. An increased concentration means that the potential of dyes to be much closer increases and also the amount of interacting dyes increases. Both leads to an increase in the FRET efficiency.

Another reason could be the increased signal to noise ratio. First the films (\(1.5\,{\upmu }\hbox {m}\)) are thinner than the paper fibers (\(10\,{\upmu }\hbox {m}\)). But also due to the temperature initiated bonding procedure very close to the melting point of pHema it is likely that the dyes within the films were diffusing into the other film. Both the decreased thickness and the diffusion of the dyes lead to an increased signal to noise ratio.

In summary the experiments with the pHema thin films permit two important conclusions. With the film with the mixed dyes it is demonstrated that the dyes DCCH and FTSC provide a working FRET system. The experiments with the bonded and unbonded films demonstrate that this FRET system, while it has not been able to indicate bonding between fiber surfaces, is able to indicate the bonding between the pHema thin films.

In this final experiment we wanted to see if it is possible to replicate the FRET results obtained by the photo spectroscopy setup using widefield fluorescence microscopy. The reason why this is more difficult is that for widefield microscopy the camera only provides intensity data instead of the full spectral information. Of course on the other hand one gains spatial information on where in the sample the signal stems from and one might even be able to detect local intensity differences and differentiate between the strength of interaction. Therefore, we again produced pHema bonds (11 bonds) and crossings (10 crossings) and also investigated a film with a mixture of dyes in it. Figure 12 shows the same kind of samples as investigated by the spectrofluorometer. In Fig. 13 the average values of the \(\text {N}_{\text {FRET}}\) intensity can be seen and show the same trend as in Fig. 11d. The \(\text {N}_{\text {FRET}}\) of the mixture is the highest as there, almost all the DCCH and FTSC molecules interact due to the homogeneous distribution of molecules. But also the bonds exhibit a 3 times higher FRET signal than the crossings as seen in Fig. 13 (subfigure). These results highlight that it is possible to measure a FRET signal between substrates using widefield microscopy.

We also looked at the images in more detail to see if it is possible to see local variations in the FRET signals of the bonded film. Figure 14 shows the enlarged and rescaled pHema bond from Fig. 12. The higher the \(\text {N}_{\text {FRET}}\) value the stronger is the interaction. In the image there are regions with considerably higher FRET signals than in others. Since the dyes are not covalently linked and the bonds were formed using increased temperature it is not possible at the moment to say whether this increased signal comes from a stronger bonding (as molecules get closer the \(\text {N}_{\text {FRET}}\) increases) or interdiffusion (as more molecules interact with each other the \(\text {N}_{\text {FRET}}\) increases). However, in either case it is a sign for local variations in bonding intensity between the surfaces.

As briefly mentioned in the experimental Sect. 2.6 we also tried acceptor photobleaching and sensitized emission, using a CLSM. From literature we found that acceptor photobleaching promises to be a rather simple but reliable FRET experiment (Jares-Erijman and Jovin 2003) and by using a CLSM we were hoping that due to the improved microscopy setup with greater focus and sensitivity a solution to the problems stated above would be found. However also with the CSLM it was not possible to detect a FRET signal when investigating the fiber-fiber bonds.