2.1.

Cell Culture

Chinese hamster ovary (CHO) cells were cultured in flasks containing Dulbecco’s modified Eagle’s medium (Invitrogen, Breda, The Netherlands) containing 7.5% fetal calf serum (Invitrogen), 2 mM L-glutamine (Invitrogen), and 2% antibiotic mix (Invitrogen) at pH 7.4. Cultures were maintained at 37°C in an incubator aerated with 5% ${\mathrm{CO}}_2$.

For experiments, cells were trypsinized and 3 ml of the cell suspension was transferred into sterile petri dishes of 35 mm diameter with an inbuilt glass cover slip of 0.16 to 0.19 mm thickness and 20 mm diameter (MatTek, USA), located in the petri dish center for optimal laser irradiation/imaging. Grids were drawn on the petri dish bottom in order to image known groups of cells with varying laser parameters. To obtain a confluent monolayer of cells the cell chambers were additionally incubated for another day under the above conditions.

2.2.

Laser Irradiation

The cells were imaged with a laser scanning unit (C1, Nikon, Japan) mounted on an inverted microscope (TE2000, Nikon, Japan) coupled to a mode-locked titanium sapphire femtosecond laser (Chameleon model, Coherent, USA) with 80-MHz pulse repetition rate, and 130-fs pulse width. Excitation intensity was controlled with a combination of a half-waveplate and a polarizing cube (Thorlabs, USA). Pulse width was adjusted with a group velocity dispersion compensator consisting of a pair of gratings (Thorlabs, USA) and measured at the focal plane with an autocorellator (APE Gmbh, Germany). All experiments were performed over a range of excitation wavelengths between 695 and 810 nm with varied intensity and pulse width using a water immersion $40\times 0.8\mathrm{N.A.}$ objective (Nikon, Japan) with autofluorescence signal acquired in a nondescanned configuration.

Confluent monolayers of CHO cells were exposed to NIR irradiation doses under varying imaging conditions. The acquired xyz optical stacks contained between 10 and 12 images of $512\times 512\mathrm{pixel}$ ($170\times 170\mathrm{um}$) size with axial intervals of 1 um. The cells in the nonirradiated regions served as an internal negative control. UV-irradiation was performed in a UV cabinet (Chromato Vue CC-20, USA) over the whole surface of the petri dishes at two individual wavelengths: 254 nm corresponding to UVC band, and 365 nm corresponding to UVA band.

2.3.

Immuno-Fluorescence Assay for CPDs

After the NIR irradiation the cells were fixed for 10 min with 4% formalin (Sigma Aldrich) in phosphate buffered saline (PBS). Washing the cells with 2 mL PBS five times followed this step and every subsequent one. Fixed cells were incubated for 5 min on ice with 0.5% Triton X-100 in PBS in order to permeabilize the cell membranes for antibody penetration. The cellular DNA was then denatured by treating the samples with 2N HCL at room temperature for 30 min. Primary monoclonal antibodies specific for CPDs (Cosmobio, Japan) were diluted $1:500$ in 5% bovine serum albumin in PBS and incubated with the cells for 1 h at room temperature. For the secondary antibodies, we used goat anti-mouse IgG conjugated with AlexaFluor-594 (Invitrogen, Germany) diluted $1:100$ in 5% bovine serum albumin in PBS, where incubation was performed for 30 min at room temperature. Finally, the cells were stained with $0.05\mu \text{g}/\mathrm{mL}$ DAPI in PBS to facilitate visual localization of nuclei during the analysis. The samples were dried and mounted with Vectashield antifade medium, and the petri dishes were closed and sealed with paraffin tape.

The one-photon CPD immuno-fluorescence from the stained cells was recorded using an EMCCD camera (Cascade model, Photometrics, USA) coupled to a fluorescence microscope (TE2000, Nikon, Japan) with a $20\times 0.75\mathrm{NA}$ PlanApo air objective (Nikon, Japan). Since formalin fixation is known to flatten cells, acquisition of a single optical section from the axial center of cells is a sufficient indicator of the total fluorescence. Signal intensity was quantified using ImageJ software ( http://rsbweb.nih.gov/ij/), where pixel intensity values were obtained from individual nuclei, and then averaged between all the cells irradiated under the same conditions. All samples were imaged using the same acquisition parameters with minimal and similar levels of photobleaching. Therefore the quantified fluorescence intensity serves as a valid metric for the amount of induced CPD lesions. Every set of simultaneously processed samples had one UVC-irradiated sample that served as positive control and provided a normalization factor for comparison between different sets to account for unavoidable fluctuations in staining efficiency and inhomogeneity.

3.1.

Intensity Versus Damage

Experimentally measured dependence of CPD damage on peak intensity for three different wavelengths at fixed pulse width of 175 fs (measured at the sample) is plotted in Fig. 2. We note there is apparent minimal peak intensity around $0.35\mathrm{TW}/{\mathrm{cm}}^2$ that produces detectable damage just above the background level. However, this is indicative of the overall staining method sensitivity at lower levels of CPDs rather than a threshold behavior. Nevertheless, these laser intensities are lower than required for tissue imaging (0.5 to $0.8\mathrm{TW}/{\mathrm{cm}}^2$), and the damage over this practical range of intensities is investigated here. As evident from the data, the minimal peak intensity that produces detectable damage decreases, while the level of CPD damage increases for shorter wavelengths.

Fig. 2

Logarithmic plot of CPD immunofluorescence signal dependence on peak intensity of laser. The slopes of the linear fits are $3.40\pm 0.33$, $2.94\pm 0.18$, and $2.81\pm 0.16$ for 780, 750, and 711 nm, respectively. Scan speed: 30 μs, pulsewidth at the focal plane: 164 fs.

Any third-order process would be governed by cubic power dependence. While accounting for the background level (parameter $A_0$, Table 1) of CPD immuno-fluorescence from nonirradiated cells, linear fits of the experimental data on the log-log scale yielded slopes of $3.40\pm 0.33$, $2.94\pm 0.18$, and $2.81\pm 0.16$ for 780, 750, and 711 nm, respectively. Clearly, these results show within the experimental error a three-photon nature of CPD damage at longer wavelengths. However, the decline in slope value of power dependence with shorter wavelengths points to the apparent sub-third-order behavior at 711 nm. In fact, if shorter wavelengths give rise to two-photon absorption corresponding to UVA absorption band, the total CPD damage would result from a superposition of the second and the third-order events.

Table 1

Results of fitting the experimental data of CPD damage dependence on peak intensity.

In order to independently verify our finding that CPD lesions are induced by simultaneous combination of two- and three-photon absorption below 780 nm we investigated dependence of CPD damage on pulse width of excitation laser at the wavelength of 750 nm and constant intensity.

For n-photon absorption process, the number of photons absorbed per molecule ($n_a$) is given by⁴⁶

The pulse widths used for imaging are normally above 100 fs, as shorter pulses get severely broadened by the microscope optics through group velocity dispersion. In our experiment, CPD damage was recorded at 750 nm with constant intensity of 14 mW while varying pulse width at the focal plane from 164 to 425 fs with a grating pair, and the results are plotted in Fig. 3. While the intensity of 14 mW is on the high end of the levels expected for cell imaging, it provided the necessary dynamic range for excitation with 425-fs pulse width. As described above, several spots in the same cell dish were imaged with different pulse widths in order to have the same staining conditions within the data series. Using, Eq. (1), the fit of the experimental data yielded the exponential value ($n^{-1}$) of $1.76\pm 0.18$ confirming our previous finding that damage arises from a mixture of two- and three-photon absorption. This dependence suggests that at longer pulse widths contribution from two-photon absorption increases as reduced photon density makes three-photon absorption less efficient. Indeed, a rapid decrease of induced CPD damage is evident between 164 and 200 fs, while this trend slows down as excitation pulse is broadened further.

Fig. 3

Pulsewidth dependence of CPD damage recorded at 750 nm with 14 mW average intensity and 30 μs pixel dwell time.

Earlier, König et al. reported the evidence of a strong effect of excitation pulse width at 780 nm on CPD damage.⁴⁷ However, the damage was shown as purely two-photon following the ($P^2/\tau$) relation. The discrepancy with our finding is likely due to the different methods of damage assessment. While our data accounts for total induced CPD damage based on immediate immunofluorescent assay, Konig et al. estimated residual damage by monitoring cell-cloning efficiency for 48 h after the exposure to laser. The latter approach obviously includes not only the direct damage, but also the subsequent cellular response (repair and apoptosis). Therefore it cannot be indicative of the actual order of absorption and the physics involved. On the other hand, if we consider the two findings together, second- and third-order dependence of direct CPD damage and second-order dependence of cloning efficiency on laser intensity at 780 nm, one can speculate that stronger initial damage has sublinear effect on cell viability.

The pulse width dependence investigated here has implications for in vivo imaging with ultrashort pulses, such as two-photon microscopy with pulse widths as low as 12 femtoseconds.⁴⁸ On one hand, decreasing the pulse width translates into linear increase of two-photon excited fluorescence intensity. At the same time, as we demonstrated here, the CPD damage increases as ${\tau }^{1.76}$ (on average), and proportionally to the subcubic peak intensity under the typically used imaging conditions. For much shorter pulses, the damage will be caused by almost 100% third-order absorption overshadowing the gain in fluorescence efficiency. The increased damage can only be offset by the corresponding decrease in the average excitation intensity in order to maintain the same peak intensity of the pulses. However, this damage cannot be compensated for without loss of intensity and, consequently, image quality.

Having demonstrated the superposition of competing two- and three-photon processes, we can decouple their relative contributions by refitting the data from Fig. 2 with the following equation:

Fig. 4

CPD immunofluorescence signal dependence on peak intensity of laser. The fits to Eq. (2) $\mathrm{CPD}\sim A_0+A_2{I}^2+A_3{I}^3$ are used to decouple the relative contributions of two- and three-photon absorption to CPD damage. Pixel dwell time: 30 μs, pulsewidth at the focal plane: 164 fs.

3.2.

Wavelength Dependence

To further investigate the relative contributions of two- and three-photon absorption, we looked at spectral response of CPD damage. A strong dependence of induced CPD damage on excitation wavelength was found, as can be seen on Fig. 5. For this experiment, only the wavelength was varied, while intensity and pulse width were maintained constant. Although the damage level is low for wavelengths above 780 nm, where the absorption is purely three-photon according to the intensity dependence discussed above, a moderate increase in damage with shorter wavelengths is observed for the 780- to 750-nm range, followed by a dramatic threefold rise of CDP formation over 750- to 710-nm range. The factors that must be considered for the explanation of this trend act oppositely. On one hand, one-photon DNA absorption is considerably stronger at UVC range (peaking between 255 and 260 nm) than at UVA. This translates into the eightfold higher CPD damage produced by UVC lamp as compared with that from UVA lamp in our positive control experiment. On the other hand, compared with three-photon absorption, two-photon process has at least an order of magnitude higher probability and occurs over a larger focal volume. At longer wavelengths, the absorption bands of endogenous cellular fluorophores are excited primarily. However, as the excitation wavelength becomes short enough ($<750\mathrm{nm}$) to access the low UVA band, the two-photon absorption quickly becomes an important and efficient mechanism of CPD damage, although still much weaker than the three-photon effect.

Fig. 5

Spectral dependence of CPD damage production. Pixel dwell time: 30 μs, pulsewidth at the focal plane: 164 fs.

We note the excessive CPD damage at 695 nm, where the data has the largest error bars. In fact, we have experienced problems with keeping cells still attached to the cell dish between exposure at 695 nm and fixation with formaldehyde. The cell loss was so high that measurement of excitation power dependence at this wavelength was not feasible, as it requires successive irradiation of several spots with 695-nm wavelength. Here the laser easily excites UVA and UVC absorption bands with high efficiency. We speculate that cumulative absorption at 695 nm is so strong that cell necrosis and/or apoptosis mechanisms are triggered immediately. Fortunately, for in vivo imaging, all endogenous fluorophores can still be excited above 750 nm, so the extreme levels of CPD damage observed at shorter wavelengths do not present additional practical concerns.

Our results suggest that imaging with wavelengths above 1000 nm would dramatically, if not completely, reduce the CPD carcinogenic risk since only the long-wavelength edge of UVA absorption band might be excited with three-photon absorption.

3.3.

Effect of Radiation Dose

A linear dependence of CPD damage on scan speed is observed in our experiments, where we maintained constant peak intensity at 750 nm (Fig. 6). As expected, pixel dwell time is directly proportional to the number of multiphoton absorption events in DNA. We note the apparent damage saturation at exposure time of 40 μs. This could be due to complete DNA dimerization and/or acute effects that cause cells detachment and loss prior to fixation and staining.

Fig. 6

Linear effect of pixel dwell time on induction of CPD lesions, pulsewidth at the focal plane: 164 fs.

3.4.

Considerations for Tissue Imaging

The main finding of this study was that the CPD damage is induced by a combination of two- and three-photon absorption processes, where the relative contributions are dependent on imaging parameters. Our conclusions are made based on DNA absorption properties within one-pixel resolution. When translating these findings into the highly scattering environment of biological tissues, attenuation of excitation as well as defocussing must be considered. The attenuation will reduce average excitation intensity, while defocussing will further decrease the peak excitation intensity at the focal volume. Additionally, the relative volumes that are subject to significant two- and three-photon fluorescence (based on $1/{\mathrm{e}}^2$ intensity profile) will change. However, as can be estimated⁴⁹ with focused Gaussian beam approximation, the volume changes alone does not significantly alter the total CPD production over 200 μm depth in human skin. Finally, we expect the scattering effectively affects these imaging parameters resulting in the corresponding combination of two- and three-photon CPD damage within the exposed focal volume.