2.2.1.

Atomic force and transmission electron microscopy

The CNT length analysis and their individual dispersion were studied by atomic force microscopy (Veeco Multimode Scanning Probe Microscope with Nanoscope IIIa Controller; Veeco Instruments, Woodbury, New York) and transmission electron microscopy (TEM) using a JEOL 2100 FE electron microscope (JEOL-USA, Peabody, Massachusetts). For TEM analysis, the CNTs were deposited onto carbon-coated copper grids without staining to reduce the possibility of artifacts.

2.2.2.

Epifluorescence and light microscopy

Phase-contrast and transmittance microscopy was performed using a light microscope (Axioskop 2 Plus, Carl Zeiss, Inc., Germany) equipped with a $12\text{-}\text{bit}$ Color MicroImager II Cooled digital camera (QImaging, Burnaby, Canada) with a resolution of 1.3 million pixels. The $100\times$ or $63\times$ oil immersion objectives (Carl Zeiss) were used to visualize and acquire the images and their sequences with a digital image and video recording software, StreamPix (Norpix, Inc., Montreal, Canada). The light microscopy system was additionally equipped with a filter set (FITC, Carl Zeiss) consisting of a bandpass filter covering $450\text{to}490\mathrm{nm}$ for an exciter and an absorbance filter covering a wavelength of $515\mathrm{nm}$ to acquire epifluorescence images for the live (green fluorescent) and dead (red fluorescent) cells (see later). Fluorescent dye—FITC-dextran 150,000 (Sigma Chemicals)—was used for the visualization of lymphatics and verification of Raman data.

3.2.1.

In vitro identification of HeLa cancer cells labeled with CNTs among red blood cells

The increase of the incubation time (12, 24, to $48\mathrm{h}$ ) of human cervical HeLa cells with the DNA-wrapped single-walled CNTs led to a gradual enhancement of the Raman signal intensity collected from areas inside the cells [Fig. 3a ]. These data indicate an increased CNT concentration and number of CNT clusters inside the cells, which were present mostly around the nucleus membrane [Fig. 3a]. These results indicate a high accumulation of CNTs both in the cell cytoplasm and inside the nucleus, especially, at increased incubation times. The CNT translocation in vitro into cells most probably happened due to endocytosis and other complex processes such as diffusion through transmembrane channels, or adhesive interactions. Some of the most important factors that are responsible for the uptake of CNTs by the cancer cells include their size, surface charge, concentration, temperature, and the biochemical conditions of the delivery.¹ CNT structural characteristics such as the diameter and length are the parameters responsible for their penetration into the cells since they present a snaking behavior when they cross the intracellular membranes. In our study, surprisingly, the CNTs were found to aggregate around the membrane of the nucleus and then penetrate the nuclear membrane into the nucleus after being dispersed individually in the medium solution that was used to feed the HeLa cells. The significant enhancement of the Raman signals from the CNT aggregates present inside the cells provides a direct path for highly sensitive Raman detection of these cells in different biological environments. This is analogous to the synergistic amplification of the PA and photothermal (PT) effects due to the self-assembled nanoclusters inside the living systems.² Figure 3b at the bottom shows the capability of Raman spectroscopy to simultaneously detect and identify (using the specific CNT G band spectral position) a single HeLa cell with CNTs ( $48\text{-}\mathrm{h}$ incubation) in a monolayer of red blood cells (RBCs) from a rat. The results were obtained in vitro by spatially scanning the laser beam across the sample with a velocity of $0.01\mathrm{mm}∕\mathrm{s}$ and “spectral” scanning (i.e., collection of a $1200\text{-}\text{to}1800\text{-}{\mathrm{cm}}^{-1}$ Raman spectral window) at a $50\text{-}\mathrm{ms}$ acquisition time. Significant high SNRs and low background signal levels of the RBCs also enabled us to distinguish one cancer cell among hundreds of RBCs in a concentrated blood sample $120\mu \mathrm{m}$ thick (not shown here).

Fig. 3

(a)In vitro Raman signals (bottom) ( $633\text{-}\mathrm{nm}$ laser excitation, $20\text{-}\mathrm{mW}$ power, and an acquisition time of $50\mathrm{ms}$ ) from single He-La cells in culture petri dishes incubated for 12, 24 and $48\mathrm{h}$ with CNTs (top). (b) In vitro Raman spectra (bottom) of single He-La cells after $48\mathrm{h}$ incubation with CNTs that was introduced among a monolayer of red blood cells. (RBCs) (top); these data were obtained simultaneously by a fast spectral analysis with an acquisition time of $50\mathrm{ms}$ and relatively slow laser spatial scanning at a rate of $0.01\mathrm{mm}∕\mathrm{s}$ .

3.2.2.

In vivo detection of the carbon nanotubes distribution in rat ear tissue

To further estimate the in vivo maximal achievable sensitivity of the technique in almost ideal static conditions, a small amount of CNT solution ( $10\mu \mathrm{l}$ , concentration of $0.2\mathrm{mg}∕\mathrm{ml}$ ) was locally injected under the skin of the rat ear in an area free of blood microvessels. The high-resolution $(40\times )$ transmission imaging of the ear tissue was used to visualize the location of the blood vessels as well as the lymphatic vessels networks highlighted by a local administration of conventional Evans Blue dye²⁸ [Fig. 4a ]. The optical monitoring of this area revealed that the initially highly localized CNTs [Fig. 4b, arrow] evolved into an elongated zone about $20\min$ after delivery [Fig. 4c, arrow].

Fig. 4

Optical transmission images of the local area of small animal’s ear tissue after the administration of Evans blue dye (a) and in another set experiment at 10 (b) and 20 (c) min after CNT injections only. The arrows in (b) and (c) show the distribution pattern of the CNTs as a function of time. (d) Raman spectra of rat ear tissue areas without CNTs as a function of the Raman spectrum acquisition time, (e) Raman spectra of the rat ear tissue area with CNTs as a function of the Raman acquisition time, and (f) Raman spectra of CNT distribution in the rat ear tissue at area below the injected site as a function time at the Raman acquisition time of $2\mathrm{s}$ ; the time $T=0\mathrm{s}$ is the starting time for spectra collection (approximately $10\min$ after CNT injection). (g) Real-time Raman detection of circulating CNTs in the blood ear microvessels as a function of time after the CNT injection in the tail vein, (h) Raman spectra collected from single He-La cell injected in the ear tissue compared to the Raman background before the cell administration, and (i) 2-D Raman mapping imaging of a single HeLa cancer cells with CNTs in rat ear tissue, based on the CNTs G band intensity spatial variation. All the Raman experiments were performed with the $633\text{-}\mathrm{nm}$ laser excitation ( $20\text{-}\mathrm{mw}$ power).

The collection of specific Raman spectra of the CNT-free ear tissue (e.g., before the CNT injection) required relatively long acquisition times because of the relatively low intensity of Raman signals. As an example, Fig. 4d shows representative Raman spectra (out of over 50) of the tissue collected in vivo at 30-, 40-, 50-, and $60\text{-}\mathrm{s}$ acquisition times. Figure 4d indicates that even for relatively long acquisition times (over $60\mathrm{s}$ ) the Raman spectra of the biological tissues did not show any peak at $1568{\mathrm{cm}}^{-1}$ [therefore specific only to CNTs—Fig. 1b] among the distinctive spectral features of the tissues.

One minute after the CNTs administration, the Raman signal started being collected from the tissue area with the maximum concentration of CNTs [Fig. 4b] at different acquisition time intervals and different wavelengths. The $1568\text{-}{\mathrm{cm}}^{-1}$ , CNT-specific peak was visible among the background Raman signals from the ear tissue even for acquisition times of as low as a few dozens milliseconds [Fig. 4e]. When the acquisition time was increased, the $1568\text{-}{\mathrm{cm}}^{-1}$ peak became better defined and increased in intensity. At a $2\text{-}\mathrm{s}$ acquisition time, besides the G band, the nanotubes’ D band $\left(1312{\mathrm{cm}}^{-1}\right)$ also became evident in the Raman spectra [Fig. 4e]. These data clearly indicate the difference in the Raman scattering properties of CNTs and biological tissues, especially for short acquisition time intervals $(<5\mathrm{s})$ when there is a smooth background from tissue with no distinguished Raman spectral features [Fig. 4e, bottom]. Note that the increase of the $1568\text{-}{\mathrm{cm}}^{-1}$ peak intensity was not accompanied by a similar enhancement of the background signal collected from the surrounding tissue, thus leading to the enhancement of the SNR. Time-related Raman spectral monitoring of the ear tissue (with the acquisition time of $2\mathrm{s}$ ) in an area close to the initial spot of high CNT concentration [see arrow in Figs. 4b and 4c] and with the starting analysis time of $10\min$ after CNT injection, demonstrated a gradual enhancement of the $1568\text{-}{\mathrm{cm}}^{-1}$ peak intensity, indicating a continuous concentration increase of CNTs at the point of analysis due to their bioredistribution [Fig. 4f].

Taking into account a previous lymphatic study,²⁸ the obtained kinetic data strongly suggested that CNTs, after their migration into the tissue, were absorbed by the neighboring lymph vessels, followed by their further propagation by the lymphatics. Indeed, regular CNT tissue diffusion usually results in a relatively homogeneous CNT distribution around the injected area, while both the optical images and Raman analysis indicated a significant increase of CNT concentration in one dominant direction, which is likely associated with the lymph microvessels located in the area of CNT injection [Fig. 4a].

3.2.3.

In vivo Raman detection of circulating carbon nanotubes in rat ear blood vessels

To determine the clearance rate of CNTs, the $100\text{-}\mu \mathrm{l}$ PBS suspension of CNTs $(0.2\mathrm{mg}∕\mathrm{ml})$ was injected into the rat blood circulation through the tail vein followed by real-time monitoring of circulating CNTs in a $50\text{to}70\text{-}\mu \mathrm{m}$ blood ear vessels [Fig. 4g] at fixed wavelength near the CNT G band $\left(1568{\mathrm{cm}}^{-1}\right)$ . The Raman signals quickly appeared within the first minutes, providing slightly fluctuated continuous background with the average Raman signal amplitudes exceeding 5 to 10 times the Raman background signals from blood vessels. Simultaneously, stronger but less frequent Raman signals were observed above relatively stable Raman backgrounds from CNTs. As discussed before,¹⁷ these phenomena can be associated with the random fluctuations of the CNTs number in the detected volume (i.e., within the focal volume of the excitation laser beam) or appearance of small CNT aggregates. This study revealed the average clearance time of the CNTs from the rat ear microcirculation to be in the range of $20\text{to}30\min$ , depending on the quantity of solution and presence of CNT aggregates. No Raman signals of the $1568\text{-}{\mathrm{cm}}^{-1}$ peak were observed $30\text{to}40\min$ after injection. The observed CNT clearance rate was found to be in a reasonable correlation with the values obtained when using fluorescent and radioactive labels ex vivo.^{13, 14}

3.2.4.

In vivo detection of cancer cells tagged with carbon nanotubes in rat ear tissue

Next we investigated the potential of RFC to detect cancer cells in vivo. HeLa cells were incubated with CNTs for $48\mathrm{h}$ and injected in a similar manner as the CNTs in the rat ear tissue followed by Raman spectral and spatial scanning ( $633\text{-}\mathrm{nm}$ laser excitation, $20\text{-}\mathrm{mW}$ power, $0.5\text{-}\mathrm{s}$ acquisition time) of the tissue near the injection site. The characteristic peaks of the CNTs are evident in the spectrum (with an intensity of around 150 counts) above background signals obtained before the administration of cells [Fig. 4h]. This high SNR enabled the collection of an $X\text{-}Y$ Raman map based on the $X\text{-}Y$ variation of the $1568\text{-}{\mathrm{cm}}^{-1}$ peak intensity. The localized Raman signal 2-D distribution presented in Fig. 4i corresponds to the presence of only a single HeLa cell in the ear tissue as verified by (1) low concentration of the injected cells, (2) signal counting from individual cells before injections [Fig. 3a and 3], comparability of signal dimension localization with single cell size $(\sim 30\mu \mathrm{m})$ .

3.2.5.

In vivo Raman detection of carbon nanotubes and cancer cells in rat mesentery lymphatics

To further verify the previously presented finding of CNT uptaking by lymphatics, the CNT solution ( $0.05\mathrm{mg}∕\mathrm{ml}$ , $100\mu \mathrm{l}$ ) was topically administered onto the rat mesentery—a thin and relatively transparent structure with clearly distinguished individual $100\text{to}200\text{-}\mu \mathrm{m}$ -diam lymph vessels [Fig. 5a ], placed on a heated microscope stage and bathed in warm Ringer’s solution $37°\mathrm{C}$ , pH 7.4). The laser beam was focused onto the area behind the lymph valve [Fig. 1a], which acts as a natural nozzle, concentrating the flowing CNTs near the vessel axis.¹⁸ Similar to the CNT blood flow experiments [Fig. 4g], time-resolved Raman monitoring of lymphatics was performed by analyzing the $1568\text{-}{\mathrm{cm}}^{-1}$ peak intensity, which is in direct relationship with the number of CNTs present in the focal volume of the laser beam. Taking into consideration the laser scattering phenomena in tissue, this volume was estimated to be approximately $20\text{to}30\mu {\mathrm{m}}^3$ .

Fig. 5

Optical images of the rat mesentery with the $633\text{-}\mathrm{nm}$ laser beam in lymph vessels (diffused due to light scattering phenomena); (b) the intensity of the $1568\text{-}{\mathrm{cm}}^{-1}$ peak (G band) from a lymph vessel as a function of time after CNTs administration; the time $T=0\mathrm{s}$ is considered the time of CNTs delivery in mesentery, the insets represent the actual experimental data collected at the respective time intervals and which were provided for comparison of the G band intensity; (c) optical transmission image of one large and two small cancer (HeLa) cells in mesentery (the left side of the image), small black spots inside cells are CNT clusters; (d) Raman spectra from single HeLa cancer cells with CNTs (red) and the Raman spectrum of the background mesenteric tissue (black); and (e) 2-D temporal-spectral imaging of single cancer cell into the mesentery. (Color online only.)

As shown in Fig. 5b, the time required for the peak at $1568{\mathrm{cm}}^{-1}$ to become present in the Raman spectra of the lymph vessel (i.e., time scale of most CNTs uptaking by lymphatics) at an acquisition time of $50\mathrm{ms}$ was approximately $3–5\min$ after the local administration of the CNTs. To avoid background Raman signals due to the CNTs possibly present on the mesentery surface, the surface of the area exposed to the laser radiation was continuously washed with warm Ringer’s solution. This procedure did not influence the Raman signal levels, suggesting the presence of CNTs only inside the lymph vessels. Raman data indicated that the intensity of the G band gradually increased with the measurement time, which showed a continuous enhancement of the CNT concentration in the lymph vessels due to the CNTs uptaking from the surrounding interstitial tissue. After $15\min$ , the CNT concentration inside the lymph vessel was found to decrease, as indicated by Fig. 5b. CNT-specific Raman signals were detected even beyond $35\min$ after the initial administration, but were of a significantly lower intensity.

Identical experiments were repeated by the administration of HeLa cells tagged with CNTs ( $48\text{-}\mathrm{h}$ incubation time) onto the mesentery. The presence of HeLa cells inside the mesentery was verified by optical transmission imaging [Fig. 5c], while the presence of the CNTs in these cells was manifested by a significantly higher intensity of the $1568\text{-}{\mathrm{cm}}^{-1}$ G peak (106 counts; acquisition time of $0.5\mathrm{s}$ ) above the mesentery tissue background [Fig. 5d]. Finally, the consecutive Raman measurements at a $1\text{-}\mathrm{s}$ time interval provided simultaneous detection and identification of a single HeLa cell through spatial, temporal, and spectral changes of the Raman signals associated with short periods of time appearance of these cells containing CNTs in the detected volume [Fig. 5e]. The clusters of CNTs inside the cells provided significant enhancement of the Raman signals, which increased the RFC capabilities to detect individual HeLa cells in the mesenteric structures.

To further understand the interaction and the developing kinetics of the CNTs after their administration in the lymphatics, Raman mapping analysis was performed by scanning the excitation laser beam across the mesentery tissue (Fig. 6, left) in two directions ( $X\text{-}Y$ mapping) at spatial step sizes of $1\mu \mathrm{m}$ of an area situated about $1\mathrm{cm}$ away from the injection site. The mapping images were built by using Horiba Jobin Yvon LabSpec 5.8 software, based on the $XY$ variation of the $1568\text{-}{\mathrm{cm}}^{-1}$ G band intensity. The lighter regions of the Raman images indicated higher concentrations of CNTs while the darker areas represent the lack of or lower concentration of CNTs (Fig. 6, right). The comparison of these 2-D Raman mapping images at 2, 12, 22, and $32\min$ after the CNT injection and the conventional transmission optical image (Fig. 6, left) indicated a good correlation of the size and shapes of Raman images with the lymphatic vessel size and margins (dashed lines in Fig. 6, right). These results suggest that the CNTs were initially distributed in the broad mesenteric area and then over time became more concentrated in a localized area of lymphatics (bright areas), followed by their further aspiration, collection, and traverse along lymph vessels leading to a decreased CNT concentration in the initial zone (darker spots). In particular, Raman mapping data taken $12\min$ after the initial injection indicated a very low concentration of CNTs in the interstitial tissue neighboring the lymphatics, while most of the CNT-specific Raman signal was derived preferentially from the lymph vessel. As time progressed further (22 and $32\min$ postdelivery), the CNTs were transported along the lymph vessels by the lymph flow indicated by an arrow in the optical image, leading to a sequential lower CNT concentration inside the lymphatic area that initially had a high CNT concentration. The Raman mapping data collected for CNTs were compared with the conventional fluorescent dye (FITC) kinetics (Fig. 7 ), injected intravenously in a rat’s tail vein at a concentration that was previously used for fluorescent microscopy²⁸ and appeared in mesentery lymphatics approximately $1\mathrm{h}$ after injections. The fluorescent dye was introduced to monitor its kinetics ( $1\mathrm{h}$ after injection) into the lymphatic system and to compare its clearance rate with that obtained by Raman mapping for CNTs. The results demonstrate that fast Raman signal acquisition algorithms and their integration into 2-D Raman imaging of tissues by fast $XY$ scanning provided a good correlation of the CNTs clearance time with the fluorescent dyes (FITC) when used for the same animal model. Therefore, despite the differences in FITC, CNT properties, and injection procedures, the CNTs were found to present a high mobility and ability to travel through the interstitial tissues, to enter into the lymph system, and to propagate through the lymphatics in a rough analogy to conventional dyes.

Fig. 6

Optical (left) and Raman $X\text{-}Y$ mapping images (right) of local area of rat mesentery at 2, 12, 22, and $32\min$ after CNT injection. The Raman scattering data were collected at regular time intervals of $50\mathrm{ms}$ /spectra and an excitation wavelength of $633\mathrm{nm}$ . The laser power was kept constant at $20\mathrm{mW}$ .

Fig. 7

Fluorescent imaging of dynamic FITC distributions in the mesentery lymphatic.