We comment on a recent Optica paper [1], in which an amplified femtosecond (fs) laser combined with a two-photon microscopy system was used to ablate targeted individual cells and monitor neuronal network changes in living mouse cortex in real time. We first report the laser parameters used by the paper’s authors, and derive the actual pulse energy and irradiance reached at laser focus, then clarify the underlying mechanism of cell ablation by single pulses and elucidate the origin of collateral damage produced by pulse series from a fs laser oscillator. We come to the conclusion that collateral damage cannot be attributed to thermal damage as hypothesized by the authors but is caused by nonlinear chemistry and free-electron-mediated biomolecular modifications.

The amplified fs system delivered microjoule laser pulses with pulse duration ${\tau _L} = {{35}}\;{\rm{fs}}$ and wavelength $\lambda = {{800}}\;{\rm{nm}}$. Single pulses with energies ${E_L}$ ranging from 0.13 to 10 µJ were used to ablate neuron cells located at depth $z = {{20}}$ to 450 µm. A fs oscillator emitting pulses with ${\tau _L} = {{140}}\;{\rm{fs}}$, $\lambda = {{930}}\;{\rm{nm}}$, and 80 MHz repetition rate was used for two-photon imaging. Pulse series from the imaging laser with about five times elevated average power were employed to produce targeted cell damage. To reach sub-micrometer resolution, the pulses were tightly focused at a numerical aperture ${\rm NA} = {1.1}$.

Figure 2 in Ref. [1] demonstrates that deep-tissue cell ablation is feasible using a single pulse from the amplified fs laser system, and shows that the ablation threshold energy ${E_{\rm{th}}}$ increases with increasing ablation depth $z$. This trend is due to scattering and linear absorption of the brain tissue, which follows Beer–Lambert’s law ${E_{\rm{th}}}(z)\;\propto \;{\exp}({\mu _{\rm{ext}}}(\lambda)z)$. Through fitting the ${E_{\rm{th}}}(z)$ data by ${E_{\rm{th}}} = a \times {\exp}({\mu _{\rm{ext}}}\;z)$, one obtains the attenuation coefficient ${\mu _{\rm{ext}}} = {0.0115}\;{\unicode{x00B5}} {{\rm{m}}^{- 1}}$, which corresponds to an attenuation length of 87 µm. This value agrees well with previous data for brain tissue at 800 nm of $\approx \;{{100}}\;\unicode{x00B5}{\rm m}$ [2].

Using this ${\mu _{\rm{ext}}}$ value, we can link the laser pulse energy ${E_L}$ incident on the tissue to the pulse energy ${E_z}$ reaching the focus at depth $z$ by [3]

The corresponding peak irradiance ${I_z}$ at focus is [3]

Here, $A$ denotes the area of the central spot of the Airy pattern with $A = \pi \;{({M^{\:2}} \times d/{{2}})^2}$, where ${M^{2}}$ denotes the beam quality parameter, and $d = {1.22}\;\lambda /{\rm NA}$ is the diffraction-limited diameter of the Airy pattern arising from focusing a beam with a top-hat profile.

 

Fig. 1. Nonlinear energy deposition by one laser pulse with ${I_z} = {2.61} \times {{10}^{12}}\;{\rm{W}}/{\rm{cm}}^2$, and heat accumulation by laser pulse series. (a) Temporal evolution of electron number density ${n_e}$ and average kinetic energy ${\varepsilon _{\rm{avg}}}$; (b) heat accumulation by pulse series for ${\rm NA} = {1.1}$ at 80 MHz; (c) electron energy spectrum at the end of pulse. Simulation parameters follow experimental conditions in Figs. 3(i) and 3(j) in Ref. [1]. The simulations in (a) and (c) were performed using the MRE(2) model in Ref. [9], and the data in (b) were obtained with Eq. (19) in Ref. [4].

Download Full Size | PDF

According to Fig. 3(h) in Ref. [1], a targeted neuron cell at $z = {{80}}\;\unicode{x00B5}{\rm m}$ can be ablated by a single pulse with ${E_L} = {0.14}\;{\unicode{x00B5}} {\rm{J}}$ without influencing the neighboring cells. The pulse energy reaching the focus is ${E_z} = {{56}}\;{\rm{nJ}}$, and the peak irradiance at the focal spot is ${I_z}\; \gt \;{{10}^{14}}\;{\rm{W}}/{\rm{cm}}^2$, well above the optical breakdown threshold at which a bubble can be created by a single laser pulse [4]. According to Fig. 2 in Ref. [5], the maximum bubble radius obtained at ${E_z} = {{56}}\;{\rm{nJ}}$ in free liquid is larger than 10 µm. A smaller value is expected in tissue due to the influence of the viscoelastic tissue matrix [6], but the initial expansion velocities are similar. In both free liquid and tissue, laser-induced bubble formation is always accompanied by shockwave emission. Thus, we conclude that mechanical forces arising from bubble formation and shockwave emission lead to the targeted cell ablation.

Although the maximum bubble radius is several micrometers large, disruptive damage is confined to a small region around the laser-induced plasma, in which the pressure of the laser-induced shockwave is high and the velocity of the expanded bubble is large [7,8]. Therefore, no collateral damage was observed experimentally.

By comparison, as seen in Fig. 3(j) of Ref. [1], generation of targeted cell damage by pulse series from the fs oscillator led to collateral damage reaching as far as tens of micrometers. Here, fs pulse series with single pulse energy ${E_L} = {2.2}\;{\rm{nJ}}$ were used to scan a region of interest at $z = {{70}}\;\unicode{x00B5}{\rm m}$ with an area of ${\rm{1 - 5}}\;{\unicode{x00B5}} {{\rm{m}}^2}$ in 2–5 s. The average power was more than 5 times higher than used for two-photon imaging. According to Eqs. (1) and (2), ${E_z} = {0.99}\;{\rm{nJ}}$ and ${I_z} = {2.61} \times {{10}^{12}}\;{\rm{W}}/{\rm{cm}}^2$. Thus, the actual pulse energy at focus was 56 times smaller, and the focal peak irradiance ${I_z}$ was more than 200 times weaker than for the single-pulse case. Collateral damage around the ablated spot was characterized by a decrease in fluorescence intensity of yellow fluorescence protein (YFP). The radius of the collateral damage zone depended on the total number ${N_{\rm{pulse}}}$ of applied laser pulses, which relates to dwell time ${t_{\rm{dwell}}}$ by ${N_{\rm{pulse}}} = {f_{\rm{PRF}}} \times {t_{\rm{dwell}}}$. Using fixed-spot irradiation, no collateral damage was observed for ${t_{\rm{dwell}}}\; \le \;{0.3}\;{\rm{s}}$; and a damage zone with radii ${R_{\rm{dz}}}\; \approx \;{{25}}\;\unicode{x00B5}{\rm m}$ and ${R_{\rm{dz}}}\; \approx \;{{45}}\;\unicode{x00B5}{\rm m}$ was observed for ${t_{\rm{dwell}}} = \;{0.7}\;{\rm{s}}$ and ${t_{\rm{dwell}}} = {1.0}\;{\rm{s}}$, respectively. These features are indicative of cumulative damage involving diffusion of the damaging “agents” from the laser focal spot to surroundings. The authors attribute it to cumulative heating and heat diffusion. However, we will show in the following that the temperature and the temperature distribution evolving during laser irradiation cannot explain the observed damage pattern.

We simulated the plasma dynamics and electron energy spectrum using a recently established multi-rate-equation (MRE) model [9]. We found that one laser pulse produces an electron number density ${n_e} = {3.4} \times {{10}^{16}}\;{{\rm{cm}}^{- 3}}$ [Fig. 1(a)], which is about four orders of magnitude smaller than the density at the optical breakdown threshold ${n_{\rm{th}}}\; \approx \;{2.0} \times {{10}^{20}}\;{{\rm{cm}}^{- 3}}$ [3]. The average kinetic energy of free electrons is ${\varepsilon _{\rm{avg}}}\; \approx \;{{7}}\;{\rm{eV}}$, and their total energy is given by the sum of kinetic energy and bandgap energy, which for water is ${E_{\rm{gap}}} = {9.5}\;{\rm{eV}}$ [3]. This leads to a temperature rise by a single pulse of $\Delta {T_{\rm{\rm single}}} = {0.022}\;{\rm{K}}$, given by $\Delta {T_{{\rm{\rm single}}}} = {n_e} \times ({\varepsilon _{{\rm{avg}}\:}} + {E_{{\rm{gap}})}}/({\rho _0}{C_p})$ [3,4]. Here, ${C_p}$ is the heat capacity and ${\rho _0}$ the mass density of water. In the above estimate, we have considered only heating by nonlinear absorption; linear absorption in tissue water is negligibly small at $\lambda = {{930}}\;{\rm{nm}}$ [10].

For fs laser pulse series, heat accumulates if the heat diffusion time is longer than the inter-pulse interval. For very large NAs, one can assume an approximately spherical shape of the heat source, and the characteristic heat diffusion time is ${t_d} = d_s^2/({{8}}\kappa)$ [4,10], where $\kappa = {1.38} \times {{10}^{- 7}}\;{{\rm{m}}^2}\;{{\rm{s}}^{- 1}}$ is the thermal diffusivity of water at room temperature, and ${d_s}$ is the source diameter. For ${\rm NA} = {1.1}$ and $\lambda = {{930}}\;{\rm{nm}}$, the equivalent spherical focus diameter is ${d_s} = {0.53}\;\unicode{x00B5}{\rm m}$, and we get ${t_d} = {{63}}\;{\rm{ns}}$ [see also the dashed line in Fig. 1(b)]. For a pulse repetition frequency of 80 MHz, the inter-pulse interval of 12.5 ns is considerably shorter than the heat dissipation time, and heat accumulates. As seen in Fig. 1(b), the peak temperature induced by pulse series is about 15 times higher than that by a single pulse. Thermal equilibrium is reached in $\approx \;{{100}}\;{\unicode{x00B5}} {\rm{s}}$, when heat diffusion is compensated for by energy deposition from subsequent laser pulses. For dwell times longer than a few µs, the tissue is heated as if by a cw laser. Taking heat accumulation into account, the temperature rise is $\Delta {T_{\rm{series}}} = {0.3}\;{\rm{K}}$. This value is far too low to cause thermal denaturation of YFP proteins or cell death [11]. Furthermore, thermal effects can be excluded because the temperature distribution around the focus drops to 1/10 of the peak value already within the first 2 µm (see Fig. 11 in Ref. [4]), which is inconsistent with the observed extent of the damage zone. Only in pigmented tissue, such as retina and skin, thermal effects can cause collateral damage in nonlinear microscopy, and special strategies must be pursued to mitigate heat accumulation [12,13]. Therefore, we conclude that the collateral damage observed in Ref. [1] is not caused by heat accumulation.

Instead, our simulations provide strong evidence that the observed damage is caused by plasma-mediated chemical effects and nonlinear photochemistry. The number of free electrons produced by each laser pulse can be estimated by multiplying the free electron density with the plasma volume, which is through the order of multiphoton absorption linked to the focal volume as described in Ref. [4]. For ${\rm NA} = {1.1}$ and $\lambda = {{930}}\;{\rm{nm}}$, about 6000 electrons are produced in the plasma volume, with an average kinetic energy of $\approx \;{{7}}\;{\rm{eV}}$, and an energy spectrum extending from $\approx \;{{2}}$ to $\approx \;{{14}}\;{\rm{eV}}$ [Fig. 1(c)]. Since tens of millions of laser pulses are applied during the dwell time, the accumulated total number of free electrons produced in the focal volume is enormous. This can lead to direct plasma-mediated effects from the interaction of free electrons with biomolecules as well as to the generation of reactive oxygen-species (ROS) through plasma-mediated water dissociation [14–16]. While laser-produced free electrons have lifetimes of the order of picoseconds and a mean traveling distance in the nanometer range [17,18], the lifetimes of some ROS (such as hydrogen peroxide) are considerably longer [14], and because of their small sizes, the diffusivity is high [19,20]. Within 0.1–1 s, they can travel over a distance of 25 µm.

Free electrons can cause bond breaks in biomolecules via dissociative electron attachment [4,21–23]. The interaction starts with the capture of free electrons into an antibonding molecular orbital to form a transient molecular anion. The anion state decays either by electron autodetachment leaving a vibrationally excited molecule, or by dissociation of one or several bonds [4,21]. In proteins, electrons can be captured by amino acids and dissociate peptide chains [23–25]. Free-electron-produced ROS may also induce molecular fragmentation and cell damage [26,27]. Photoproducts generated by the primary reactions with electrons and ROS will likely include free radicals with lifetimes of seconds or minutes that can diffuse far enough to create damage in a zone with tens of micrometer radius [20]. All these reactions can likely destroy the chemical structure of YFP and quench fluorescence.

Laser-induced plasma-mediated effects are dose dependent because chemical reactions are rate processes that depend on the concentration of reactants and time. The cumulated irradiation dose $H$ producing the reacting electrons can be expressed by the radiant exposure:

For ${t_{\rm{dwell}}} = {0.3}$ to 0.7 s, the irradiance dose is $H = {3.2} \times {{10}^6}\;{\rm{J}}/{\rm{cm}}^2$ to ${7.4} \times {{10}^6}\;{\rm{J}}/{\rm{cm}}^2$ using parameters from Fig. 1. These values are two to three orders of magnitude larger than the radiant exposure leading to the onset of photodamage in multiphoton microscopy, which is characterized by hyperfluorescence of endogenous biomolecules in the irradiated area [28–30]. It is attributed to nonlinear photochemistry and low-density-plasma-mediated effects [31–33]. At much higher doses, as those applied in Ref. [1] for intentional cell destruction, the concentration of damaging agents at the site of laser generation is large enough to produce damage even in a fairly large region around the irradiated area.

In summary, we support the conclusion by Cheng et al. that plasma-mediated targeted cell ablation by single fs laser pulses is a versatile tool for studying cellular function with little collateral damage, which is backed also by previous research [4,34,35]. However, we want to point out that collateral damage by fs pulse series is due to free-electron-mediated biomolecular damage rather than thermal effects as proposed by the authors. Exploration of the underlying interaction pathways and kinetics is an important field of research that deserves more attention.