2.1.

Animals

C57BL/6 line mice (female and male, 4–20 weeks old) were purchased from Biochemistry Institute (Vilnius University, Lithuania) and used for the experiments. All animals were kept in the same room at $12/12\text{-}\mathrm{h}$ light/dark cycle, 20 to 22°C temperature, $55\pm 10\%$ humidity and ad libitum access to food and water. Before the experiment, the fur on the hind leg limbs of the mice was removed with a depilatory cream. Prior to plasmid intramuscular injection and subsequent EP or SP, all mice were anesthetized by intraperitoneal injection of ketamine ($100\mathrm{mg}/\mathrm{kg}$; Richter Pharma AG, Wels, Austria) and xylazine ($10\mathrm{mg}/\mathrm{kg}$; Eurovet Animal Health B.V., Bladel, Netherlands). The animals woke up within 1 h and appeared normally active. All experiments were approved by the Lithuanian Republic Alimentary and Veterinary Public Office (Nr. 0236).

2.2.

Plasmid

Plasmid pEGFP-Nuc Vector coding green fluorescent protein (GFP) mutant EGFP was used in all experiments. Mice undergoing EP received $10\mu \mathrm{g}$ of EGFP coding plasmid which was diluted in $50\mu \mathrm{l}$ of 0.9% NaCl. Mice undergoing SP received $10\mu \mathrm{g}$ of EGFP coding plasmid, which was initially prepared in $10\mu \mathrm{l}$ of 0.9% NaCl and then mixed with $40\mu \mathrm{l}$ of Sonovue (Bracco, Switzerland) MB. Sonovue MB were prepared in 0.9% NaCl according to manufacturer recommendations and counted under the hematocytometer revealing a final concentration of $\sim 2\times {10}^8\mathrm{MB}/\mathrm{ml}$. EGFP coding plasmid was injected into a tibialis cranialis muscle of both mouse legs using insulin syringe from the distal to proximal direction, $50\mu \mathrm{l}$ totally for each muscle.

2.3.

DNA Electrotransfer

Two stainless steel electrodes smeared with ECG conductive gel were applied on both sides of tibialis cranialis muscle, above the skin and 4 to 5 mm apart. Muscle EP was performed using electroporator manufactured in Kaunas University of Technology, Lithuania. Three different combinations of electric pulses were used: (1) one square-wave HV pulse with a length of $100\mu \mathrm{s}$ and amplitude of $800\mathrm{V}/\mathrm{cm}$; (2) 8 LV pulses at 1 Hz frequency with a length of 20 ms and amplitude of $200\mathrm{V}/\mathrm{cm}$; (3) combination of 1 HV + 4 LV pulses, with the LV pulse length of 100 ms and pulse amplitude reduced to $80\mathrm{V}/\mathrm{cm}$. The time delay between HV and LV pulses was 1 s.

2.4.

DNA Sonotransfer

For muscle SP, 1 MHz US at power density of $2\mathrm{W}/{\mathrm{cm}}^2$ [20% or 100% duty cycle (DC)] was delivered for 5 min through 6 mm transducer, using Sonitron 2000 (Artison Corp., OK) sonoporator. US conductive gel was used to ensure a good contact between the transducer and the skin above the tibialis cranialis muscle.

2.5.

Experimental Groups

Four to six mice were selected randomly for each of six experimental groups: (1) the control group mice receiving only EGFP coding plasmid; (2) EP group receiving EGFP coding plasmid followed by 1 HV electric pulse; (3) EP group receiving EGFP coding plasmid followed by 1 HV + 4 LV electric pulses; (4) EP group receiving EGFP coding plasmid followed by 8 LV electric pulses; (5) SP group receiving EGFP coding plasmid followed by US irradiation at 100% DC, and (6) SP group receiving EGFP coding plasmid followed by US irradiation at 20% DC. EP or SP was applied on the muscle 5 min after EGFP coding plasmid injection.

2.6.

Fluorescence Point Measurements and Multispectral Imaging

EGFP fluorescence in vivo was observed at 2, 3, 7, 14, and 30 days then 3, 6, and 12 months after electro- or sonotransfection, using optical biopsy setup depicted in Fig. 1. Fluorescence spectra were measured with a fiber optic spectrofluorimeter (AvaSpec ULS2048L, Avantes, Netherlands) which was equipped with a Y-shaped fiber bundle (FC-UV400-2-SR, Avantes, Netherlands). The central fiber of the fiber bundle was used to conduct the excitation light to the tissue, and the six circularly arranged peripheral fibers were used to collect the fluorescence signal. DPSS (473 nm) laser (5-473-DPSS-0.1, CW, max. output 100 mW, Altechna, Lithuania) was used for skin and muscle fluorescence excitation, and a long pass filter ($T_{516}=60\%$) was inserted into the detection light path to cut off the scattered excitation light. Nine spectra were measured at the skin surface perpendicularly to a muscle fiber length: three at the distal, three at the central, and three at the proximal location of muscle using 300 ms integration time. Spectra from the same muscle location were averaged and mean intensity at 512 to 515 nm was calculated. EGFP fluorescence kinetics is presented as a mean value (averaged from distal, central, and proximal locations) $\pm \text{standard error}$ of the mean.

Fig. 1

Experimental setup for fluorescence point measurements and multispectral imaging.

Mice in all experimental groups were killed 12 months after electro- or sonotransfection by using ${\mathrm{CO}}_2$ gas. Then, the skin above the muscle was cut and removed. Fluorescence spectra on the surface of the muscle were acquired as described previously, with respect to distal, central, and proximal muscle location. EGFP fluorescence spectra ex vivo were averaged and represented as mean value (at 515 nm) $\pm \text{standard error}$ of the mean.

In parallel, the multispectral imaging system depicted in Fig. 1 was employed to visualize the distribution of EGFP fluorescence on the muscle surface and on the skin above the muscle. The system consisted of a multispectral imaging camera Nuance EX (model: N-MSI-EX, CRi), a fiber-optic cable for delivering laser excitation (473 nm) to the skin surface and LP filter ($T_{516}=60\%$) blocking the scattered laser light. The images were acquired over the whole mouse leg in the wavelength range 450 to 950 nm with the scanning step of 10 nm and the filtering bandwidth of 20 nm (FWHM). The subsequent processing of the images was performed using CRi Nuance software.

2.7.

Tissue Collection, Fixation, and Processing

Immediately after the measurements ex vivo, the tibialis cranialis muscles were dissected, weighed (Kern, ABS 80-4, Germany) and snap frozen in 2-methylbutan precooled in liquid nitrogen.

2.8.

Analysis of Multispectral Images

Acquired multispectral images of muscles were processed and analyzed using the MATLAB (MathWorks, Natick, Massachusetts) software. Some of the in vivo multispectral images of a single leg of a mouse were not precisely aligned because of small leg movements [Fig. 2(a)]. Image registration was required in such cases. During registration process, all images were aligned in respect to the image of lowest wavelength of 450 nm. SURF feature detector³¹ was used to find, describe, and compare key points between images of adjacent wavelengths (e.g., 450, 460, 470,…, 950 nm). Corresponding points were matched to acquire transformation matrix that would allow registering image of higher wavelength to adjacent image of lower wavelength. Result of registration of images of Fig. 2(a) is shown in Fig. 2(b).

Fig. 2

Image processing steps: (a) initially acquired RGB image (zoomed) with unalignment between concurrent images at different wavelengths; (b) RGB image after registration; (c) manual selection of muscle region; insert: muscle blob for parameter acquisition; (d) automatic segmentation of fluorescent and nonfluorescent regions; (e) manual segmentation of fluorescent and nonfluorescent regions employing spectrum analysis for each pixel in muscle area. User defines two points: (1) in the fluorescent region, (2) in nonfluorescent region; insert: typical spectra of fluorescent and nonfluorescent points.

To acquire parameters of in vivo images, manual segmentation of muscle region was carried out. Figure 2(c) shows graphical user interface made in MATLAB. The interface allows the user to select points in the RGB image depicting the border of the muscle. The software calculates the area of selected muscle region by counting pixels inside the specified border. Using blob analysis functions of MATLAB Image Processing Toolbox, a minor axis length of muscle blob is found. This length approximately corresponds to the diameter of the muscle.

The goal of the next processing step was to find the fluorescent region in the muscle. Initial segmentation was accomplished by thresholding muscle blob for an image of 510 nm wavelength. This wavelength (green light) corresponds to the maximum of EGFP spectral band; therefore green fluorescent regions (with the presence of EGFP) were likely much more intensive than the nonfluorescent regions (NFs). Automatic threshold was found by using Otsu’s method.³² In this iterative method, pixels are divided by threshold so that the sum of the spread of foreground and background pixels is at its minimum. The result of automatic segmentation is shown in Fig. 2(d). Red border delineates the fluorescent region.

Automatic segmentation does not always return satisfactory results. If the border delineation of the fluorescent region is uncertain, the manual segmentation is possible. In this step, user defines two points (1, 2) in the image—one corresponds to fluorescent region (F), and the other to NF. Figure 2(e) shows the segmentation result after spectrum analysis for each pixel in the muscle area [same input image as in Fig 2(d)]. The $3\times 3\text{pixel}$ regions around defined points are used to find mean spectrum ($S$), $S_{\mathrm{F}}$, $S_{\mathrm{NF}}$ of both points. Similarly, spectrums are acquired for each current pixel of muscle region ($S_{\mathrm{C}}$). Spectrum of current pixel is compared with spectrums of fluorescent point and nonfluorescent point acquiring distances $\mathrm{d}(S_{\mathrm{C}},S_{\mathrm{F}})$ and $\mathrm{d}(S_{\mathrm{C}},S_{\mathrm{NF}})$:

The software combines the parameter $R$ and the result from previous automatic segmentation for final segmentation of the muscle:

When the muscle is segmented, parameters of fluorescent regions are acquired. The software calculates area of fluorescent region and muscle diameter. Algorithm also calculates mean and sum intensity values of pixels of fluorescent region in the 510 nm wavelength image.

2.9.

Statistics

Results are given as $\text{mean values}\pm \mathrm{SEM}$. The Mann–Whitney test was applied for comparison of significance between single experimental points in electroporated, sonoporated, and only pDNA injected groups.

3.1.

Noninvasive Evaluation of Enhanced Green Fluorescent Protein Transfection Efficiency in Mouse Tibialis Cranialis Muscle

Transfection efficiency of EGFP coding plasmid was assessed by two different noninvasive methods: fluorescence multispectral imaging and fluorescence spectroscopy point measurements. Briefly, the principle of multispectral imaging is to acquire concurrent monochrome images of the same object (EGFP fluorescent tibialis cranalis muscle) by using multiple optical band-pass filters [Fig. 3(a)]. The number of tibialis cranialis muscle fluorescence images, each of them taken with a different band-pass filter, forms the image cube with the fluorescence spectrum available at each pixel of an image plane. The initially acquired fluorescence image of mouse leg [Fig. 3(b)] was analyzed using 510 and 520 nm spectral bands, allowing the recovery of the spectral properties of EGFP fluorophore. As it is shown in Fig. 3(c), the muscle localized EGFP emits the specific fluorescence in 510–520 nm spectral region, making the pattern and location of EGFP clearly distinguishable from the skin autofluorescence.

Fig. 3

(a) Schematic representation of the image cube and (b) the initial fluorescence image stack of EGFP transfected tibialis cranialis muscle acquired in 450 to 950 nm spectral region. (c) The image of tibialis cranialis muscle taken with a $510\pm 20\mathrm{nm}$ (FWHM) band-pass filter used for further analysis to reveal parameters for the noninvasive quantification of EGFP transfection efficiency. The green signal within the marked area indicates the muscle fibers expressing EGFP (image scale bar: 5 mm). (d) Fluorescence spectroscopy point measurements before spectra normalization and (e) after normalization to the same intensity (1) at 613 nm; 64 fluorescence spectra from day 3 of EGFP pharmacokinetics are depicted.

Skin characteristics not to retain EGFP³³ make a superficial tibialis cranialis muscle a very suitable model for noninvasive estimation of EGFP expression levels. Under excitation at 473 nm, the autofluorescence of tibialis cranialis muscle as well as the skin above the muscle are characterized by the broad and poorly structured emission bands in 490 to 800 nm spectral region [Fig. 3(d)].

Differently from skin autofluorescence, the spectra of EGFP inside the transfected muscles possess a new fluorescence band with a peak at 512 to 515 nm which differs from tissue autofluorescence both in shape and intensity. In order to distinguish spectral differences related to EGFP emission, all fluorescence spectra were normalized to the same intensity at 613 nm [Fig. 3(e)]. Spectra normalization allowed to compensate the variation of autofluorescence intensity as well as to resolve the spectral overlap between tissue autofluorescence ($<3\mathrm{r.u}$.) and EGFP fluorescence ($>3\mathrm{r.u}$.).

Initially, we tested the correlation of EGFP fluorescence intensity between in vivo, when the detection fiber is on the skin covering the transfected muscle, and ex vivo, when skin is removed from the muscle and the detection fiber is at the same place of muscle surface. Mice for this experiment were selected randomly and independently from EP or SP protocol. The obtained data [Fig. 4(a)] showed strong linear correlation between EGFP fluorescence measured from the surface of skin and that measured at the surface of muscle ($R^2=0.94$). According to Fig. 4(a), the skin which covered the muscle from above, reduced EGFP fluorescence intensity of muscle surface by a factor of 1.4 on average.

Fig. 4

(a) Comparison between EGFP fluorescence intensity point measurements from the skin covering muscle in vivo and muscle surface ex vivo; 32 mice with both legs transfected were first measured in vivo (at the day 360) and then killed for ex vivo measurements. (b) Fluorescence data revealing the linear correlation of GFP fluorescence intensity to GFP concentration increase. (c) The linear correlation between EGFP fluorescence detected in vivo and in muscle homogenate; 4 mice were randomly selected for homogenate preparation: control group–1; SP 20% DC–1; EP 8 LV–1; EP 1 HV + 4 LV–1 mice.

As the absolute control, the homogenates of the transfected muscles were made and the fluorescence of EGFP in muscle homogenate was compared with the average EGFP fluorescence intensity measured at the surface of the same muscle. At the initial step of this investigation, the commercially available recombinant GFP (Pierce, Thermo Scientific) was dissolved in homogenate of control mice muscle (with no pDNA transfected) and used as the calibration etalon. In Fig. 4(b), the fluorescence intensity at 512 nm is plot against the etalon GFP concentration, revealing the linear dependence in 4 nM to $2.85\mu \mathrm{M}$ interval. During the next step, using the same experimental setup and the same spectra registration parameters we measured fluorescence of the muscle homogenate transfected with EGFP [Fig. 4(c)]. To match EGFP concentration with the etalon GFP concentration calibration curve, the $Y$-axis of Fig. 4(b) should be multiplied by $({\phi }^{\mathrm{EGFP}}/{\phi }^{\mathrm{et}})[(1-T^{\mathrm{EGFP}})/(1-T^{\mathrm{et}})]$, where $T^{\mathrm{et}}$ (or $T^{\mathrm{EGFP}}$) is the transmittance of etalon GFP (or EGFP) to the excitation wavelength, and ${\phi }^{\mathrm{et}}$ (or ${\phi }^{\mathrm{EGFP}}$) are the fluorescence quantum yields of the etalon GFP (or EGFP). Such correction of Y-scale for Fig. 4(b) may allow calibrating the fluorescence of EGFP in muscle homogenate with EGFP concentration.

Similarly to the EGFP fluorescence correlation obtained between muscle surface and skin surface [Fig. 4(a)], data in Fig. 4(c) show strong correlation of EGFP fluorescence intensity measured at muscle surface and in muscle homogenate ($R^2=0.9$). We conclude that the strong correlations between EGFP levels detected at the skin surface and the EGFP levels ex vivo allow the application of noninvasive fluorescence diagnostics for the in vivo evaluation of EGFP transfection efficiency in mice tibialis cranialis muscle.

3.2.

Long-Term Evaluation of Enhanced Green Fluorescent Protein Fluorescence Intensity

Further in this study we applied EP and SP using various electric field or US output parameters optimized by different groups^{3,19,26,34,35} for pDNA delivery into skeletal muscle. By using noninvasive fluorescence spectroscopy point measurements, we monitored the time dependence of EGFP fluorescence intensity in tibialis cranialis muscle for the period of 1 year. Data on day 7 and 360 were analyzed for statistical significance (Table 1).

Table 1

Data in Fig. 5(a) analyzed for significance 7 days and 360 days after transfection.

Fig. 5

Noninvasive monitoring of EGFP pharmacokinetics in vivo by using fluorescence spectroscopy point-measurements. (a) Interindividual variability of EGFP expression after muscle electro- or sonotransfection represented by the percentage of EGFP positive muscles comparing to all of EP or SP affected muscles. (b) For EP assisted EGFP coding plasmid delivery, 8 LV (closed squares), 1 HV (crossed squares) and 1 HV + 4 LV (open squares) electric pulses were used. For SP assisted delivery either 100% DC US (filled triangles) or 20% DC US (open triangles) was applied. The number of mice in experimental groups: control group–4; SP 20% DC group–6; SP 100% DC–4; EP 1 HV–6; EP 8 LV–6; EP 1 HV + 4 LV–6 mice.

No EGFP fluorescence signal was detected in the control animal group when the plasmid was injected into muscle without any further EP or SP treatment [Fig. 5(a)]. Fluorescence intensity comparison between the control group and other experimental groups revealed that EP (1 HV, 8 LV or 1 HV + 4 LV) resulted in a statistically significant ($p<0.001$) increase of EGFP fluorescence over the one year period of observation. At our experimental setup, the most efficient pDNA transfection was achieved by using 1 HV + 4 LV EP protocol. The peak of EGFP fluorescence intensity was obtained 7 days after electrotransfection, and it was significantly higher in comparison with any other EP groups. Then during the second week of observation EGFP fluorescence intensity decreased by half and remained stable throughout the next half year period. The peak of EGFP fluorescence in 1 HV group was achieved 10 days after electrotransfection but it was significantly lower in comparison to the peak fluorescence intensity in 1 HV + 4 LV group ($p<0.001$). Considering the accumulation phase of EGFP, the less intense EGFP fluorescence was detected in 8 LV group, peaking from the 1st to 4th week after electrotransfection.

Figure 5(b) represents the interindividual variability of EGFP expression. The most efficient electric pulse protocol for electrotransfection was 1 HV + 4 LV with 100% of the muscles already positive for EGFP at second day post–electric pulse delivery [Fig. 5(b)]. For 1 HV or 8 LV groups, our noninvasive observations revealed that 75% to 100% of the muscles were positive for EGFP at 7th day after electric pulse delivery [Fig. 5(b)].

Muscle transfection by using SP resulted in relatively low EGFP peak fluorescence intensity, with no statistical significance between 20% and 100% US DC applied. Moreover, using SP less than 50% of the muscles were transfected. The elevated levels of EGFP fluorescence remained approximately constant between days 10 and 30 after sonotransfection. Two weeks after SP, the mean fluorescence intensity of EGFP reached the similar values as in 1 HV or 8 LV groups. During the clearance period, the average fluorescence intensity at 512 nm gradually decreased to the value of 2.3 r.u. (which is similar to the control group autofluorescence). During the clearance stage of EGFP (day 360; 20% US DC group), 25% of sonotransfected muscles still had a few low fluorescent fibers [Fig. 6(k)] however this result did not contribute to the overall increase of muscle fluorescence intensity above the control levels.

Fig. 6

Detection of EGFP expression in muscles across the skin one year after muscle transfection. The top panel: (a–c) in vivo representative images of electrotransfected muscles by 1 HV + 4 LV, 8 LV or 1 HV pulse application or (d, e) sonotransfected muscles by applying 20% or 100% US DC. (f) Leg of mouse injected with EGFP plasmid without exposure to EP or SP is depicted. The bottom panel (g–l) depicts the same tibialis cranialis muscles after removal of the skin (image scale bar: 5 mm).

3.3.

Comparison of EGFP Fluorescence Intensity in Ex Vivo Muscles with In Vivo Results

Finally, all mice were killed 1 year after pDNA transfection and EGFP fluorescence distribution was evaluated at tibialis cranialis muscle surface and within the volume of the muscle. The multispectral images of muscle surface showed that the largest area of EGFP fluorescence was obtained when eight LV electric pulses were used for EP (Table 2). EGFP fluorescent area covered up to 68% of the muscle surface in 8 LV group [Fig. 6(h)] and this area was significantly larger compared with the EGFP fluorescent area obtained using other electric pulse protocols ($\sim 38\%$ for 1 HV + 4 LV and $\sim 28\%$ for 1 HV) or SP protocol ($\sim 8\%$) with 20% US DC [Figs. 6(g), 6(i), and 6(k)]. The remaining sonotransfected muscles (all in 100% US DC group) as well as the muscles in control group [Figs. 6(j) and 6(l)] were not fluorescent with EGFP. Similar to fluorescence spectroscopy results, the present multispectral imaging data revealed very high interindividual variability of EGFP expression after SP.

Table 2

Results of multispectral image analysis: comparison of GFP fluorescent area and GFP fluorescence intensity levels one year after applying different transfection protocols.

In contrast to the transfection area data, the highest average and integral fluorescence signals of EGFP were induced by using 1 HV + 4 LV electric pulses. Application of 8 LV or 1 HV pulses, as well as SP, resulted in lower fluorescence intensities of EGFP detected at the muscle surface (Table 2). Such findings could have been prognosticated from the noninvasive multispectral imaging data in vivo [Figs. 6(a)–6(f)]. A linear regression was found to describe the relationship between Table 2 listed parameters measured in vivo (represented by $X$-axis) and at muscle surface ($Y$-axis), revealing the regression slope of $k=0.91\pm 0.13$ for EGFP fluorescence mean; $k=1.23\pm 0.16$ for EGFP fluorescence area and $k=1.85\pm 0.26$ for EGFP fluorescence integral values.

Histological samples were prepared from each muscle examined ex vivo. We sliced the muscle perpendicularly to the longitudinal axis of the fibers, and analyzed the same parameters in the transverse muscle section: area of EGFP fluorescent region, mean intensity, and integral intensity of EGFP positive pixel values. These results are presented in Table 2. In agreement with our previous findings, the highest number of EGFP positive fibers in the transverse sections of the muscle as well as the most intense fluorescence of the fibers had been detected when 8 LV or 1 HV + 4 LV electric pulses were used for pDNA transfection [Figs. 7(c), 7(d) representative images]. Meanwhile, after plasmid injection followed by SP (20% DC) the long-term transfection was limited to the presence of single EGFP positive fibers within the muscle [Fig. 7(b)]. EGFP fluorescence was not detected in the control group [Fig. 7(a)] and SP (100% DC) group.

Fig. 7

Transverse sections of tibialis cranialis muscle (left panel): (a) control mice; (b) mice one year after sonotransfection using $2\mathrm{W}/{\mathrm{cm}}^2$, 5 min, 20% DC, 1 MHz US; (c, d) mice one year after electrotransfection using 8 LV or 1 HV + 4 LV pulses. The size of a single transverse section image is limited by the microscope field of view (image scale bar: $200\mu \mathrm{m}$). The intense green color stained fibers in EP and SP groups express EGFP. Results of image processing (right panel): (e) dependency of EGFP positive fiber integral fluorescence measured at the skin above the muscle in vivo (filled squares) or at the muscle surface (open squares) with respect to EGFP integral fluorescence measured in transverse muscle sections ex vivo; (f) dependency of EGFP positive fiber fluorescence area measured in vivo (filled squares) or at the muscle surface (open squares) with respect to EGFP fluorescence area measured in transverse muscle sections ex vivo. Number of samples for transverse muscle sections: 56 (from 28 mice).

The analysis of fluorescence intensity parameters obtained in transverse muscle sections showed the strong linear correlation of integral EGFP fluorescence in transverse muscle sections with the integral fluorescence detected in vivo ($k=27.5\pm 4.6\times {10}^{-3}$), as well as detected on the surface of muscle ex vivo ($k=12.4\pm 4\times {10}^{-3}$), Fig. 7(e). The mean fluorescence intensity registered in transverse muscle section was a less sensitive metric for correlation with EGFP multispectral imaging data in vivo or ex vivo, due to its low-dynamic range. In comparison to integral fluorescence intensity, the dynamic range of mean fluorescence intensity (expressed as the ratio between the largest and smallest values from Table 2) was lower by one order of magnitude. Finally, Fig. 7(f) shows the linear correlation between EGFP fluorescent fiber area (which represents the certain number of fluorescent muscle fibers) in transverse section and EGFP fluorescent fiber area measured in vivo ($k=1.4\pm 0.2$) or at muscle surface ($k=1.7\pm 0.4$). The following conclusion is that multispectral imaging data of EGFP fluorescence in live mice tibialis cranialis muscle can also provide information related to the number of EGFP fluorescent fibers and EGFP fluorescence intensity levels within the muscle.