Single-cell label-free photoacoustic flowoxigraphy in vivo

Quantitative Measurement of Oxygen Release from Single RBCs in Vivo.

To record real-time oxygen delivery, a set of B-scan images was acquired at varied time points (Fig. 2A), as single RBCs flowed from the left to the right side of the field of view. A real-time video is shown in Movie S1. The oxygen release from single RBCs was clearly imaged cell by cell. A representative RBC shown in Fig. 2A desaturated by 0.03 over a 32-µm travel distance. The oxygenation imaging speed currently is limited by the 2-kHz pulse repetition rate of the 560-nm pulsed laser, which may be improved in the future to enable 3D single-RBC FOG. Taking advantage of the ultra-short wavelength switching time, fast scanning speed, and high spatial resolution, we can simultaneously quantify C_Hb, sO₂,, v_f, and rO₂ from images of single RBCs, as shown in Fig. 2B. By operating a 532-nm single-wavelength laser at a 20-kHz pulse repetition rate, we achieved 3D imaging of flowing single RBCs with a 20-Hz rate, as shown in Movie S2.

Relationship Among Single-RBC Parameters.

FOG has the capability to measure relationships among multiple parameters of a single RBC. Here, we imaged single flowing RBCs in designated vessels in the mouse brain cortex at a 20-Hz B-scan rate while the mice were breathing air mixed with isoflurane. The imaged flowing RBCs were 60–150 µm deep from the brain cortex surface and had travel distances of 30–60 µm within the B-scan window. The vessel inclination was less than 20° with respect to the x-y plane. More than 6,000 B-scan images were acquired at each wavelength. We simultaneously calculated multiple functional parameters from the single-RBC images and averaged every 20 B-scans. Even without applying any challenge or stimulation, oxygen delivery was found to fluctuate. Although RBC oxygen saturation should be relatively uniform in arterioles, we observed dramatic sO₂ variations from cell to cell in some vessels. A plausible reason for the variations is that the imaged vessel segment was close to venules and the RBCs released different amounts of oxygen before they were imaged. We further imaged sO₂ variations in different segments of capillaries. The sO₂ of capillaries close to the arterial end had an average SD of 0.09 ± 0.03. The sO₂ of capillaries close to the venous end had an average SD of 0.12 ± 0.04. Each SD first was calculated from 500 measurements in a capillary, then was averaged over 30 capillaries. A t test showed that the arterial end had a smaller SD than the venous end with P < 0.01.

Fig. 2 C–E shows the relationships among <sO₂>,, v_f, and rO₂. As expected from Eq. 1, Fig. 2C shows that rO₂ increases with both and v_f. Whereas the former reflects the amount of oxygen released by each RBC, the latter determines the dwell time of flowing RBCs in the field of view. Fig. 2D shows that increasing v_f increases rO₂ if <sO₂> is maintained constant, but increasing <sO₂> decreases rO₂ if v_f is held constant. From Fig. 2 C and D, we also observed that for constant rO₂, increasing v_f decreases and increases <sO₂>. Fig. 2E shows that a decrease of either v_f or <sO₂> is correlated with increasing. Taken from another set of data, Fig. S1 and Movies S3 and S4 further corroborate the relationships among <sO₂>,, v_f, and rO₂.

These results show two possible approaches to altering oxygen release from single RBCs to local tissue. The first approach is via sO₂ while v_f and C_Hb are held constant. When the local tissue consumes more oxygen, increases. The second approach is via v_f while sO₂ and C_Hb are held constant. When the tissue demands more oxygen, v_f can increase so that more RBCs flow through the tissue region within a given time period.

Oxygen Release Under a Transition of Inhalation Gas from Low to High Oxygen.

Using the single-RBC FOG, we imaged the dynamic oxygen release in a mouse ear when altering inhalation gas from low to high oxygen concentration. Initially, the mouse was breathing 5% oxygen gas for over 10 min. Then, we altered the inhalation gas to pure oxygen and immediately started to image single flowing RBCs at 20 Hz. As shown in Fig. 3A, we observed a dramatic increase in single-RBC sO₂ within 60 s. A real-time movie recorded the transition of single-RBC oxygenation, as shown in Movie S5. Single-RBC functional parameters, including <C_Hb>, <sO₂>,, v_f, and rO₂, are plotted in Fig. 3 B–F. Each parameter was computed from the images of single RBCs and averaged over every second. Every 10 data points (10 s) were grouped for comparison. Statistical tests from three mice show that <C_Hb>, <sO₂>, , and rO₂ increased by 49 ± 3%, 71 ± 2%, 96 ± 7%, and 270 ± 22%, respectively, but v_f did not change significantly.

Oxygen Release in the Brain Under Acute Systemic Hypoglycemia.

In cellular aerobic respiration, the metabolic rates of glucose and oxygen are strictly proportional to each other. We demonstrated the use of single-RBC FOG to study glucose-associated oxygen release at the single-RBC level.

As shown in Fig. 4A, an sO₂ image was acquired over a large field of view in a mouse brain. Then, we selected a small region to image. We injected insulin s.c. (30 IU/kg) to induce a systemic decrease in blood glucose. Immediately after the injection, single-RBC functional images were acquired at 100 Hz for over 60 min. The blood glucose concentration was measured with a glucose meter (FreeStyle Lite; Abbott Diabetes Care Inc.) by drawing blood once every 10 min from the mouse’s tail. Fig. 4B shows the decrease of the blood glucose level with time. Fig. 4 C–G shows the multiple functional parameters computed from the single-RBC images and averaged over every second. We observed that the rO₂ changed with blood glucose level. The results, especially the rO₂, show clear biphasic responses.

We fitted the rO₂ data in Fig. 4C with a bisegmental linear regression model as follows:

where and is the separation time point between the two linear segments. At the fitted equal to 19.6 min, the blood glucose concentration is 58.5 mg/dL, which agrees well with the independently determined dividing point between normoglycemia and hypoglycemia (19). Therefore, we divide the quantitative measurements of glucose concentration, rO₂, <C_Hb>, <sO₂>,, and v_f into two phases, as marked by the two colors in Fig. 4 B–G. We define the first phase as normoglycemia (glucose concentration ≥58.5 mg/dL) and the second phase as hypoglycemia (glucose concentration <58.5 mg/dL). The data of each phase were averaged and fitted to a linear regression model to compare the means and slopes, as shown in Fig. 4 H and I. In the normoglycemia phase, rO₂ decreased with decreasing glucose concentration (Fig. 4C). The other functional parameters, including <C_Hb>, <sO₂>, , and v_f, decreased as well.

Imaging of Neuron–Single-RBC Coupling in the Brain.

The study of neurovascular coupling has gained broad interest because hemodynamics may be used as an important surrogate in exploring neuroscience and studying brain disorders (20–23). Here, we demonstrate that single-RBC FOG can image coupling between visual neural activity and single-RBC oxygen release in the brain.

As shown in Fig. 5A, optical stimulation from a bright white light-emitting diode (LED) was applied to the left eye of a white Swiss Webster mouse. Target RBCs were imaged in the right visual cortex region of the brain through a craniotomy. Fig. 5B shows the transient responses of single RBCs to single visual stimulations (a 0.5-s flash). Each parameter first was computed from images of single RBCs, then averaged over 1 s for 104 trials. Statistical analyses based on four mice showed that <sO₂>, , v_f, and rO₂ changed significantly after visual stimulation, but <C_Hb> did not show obvious changes. We observed an increase in flow speed after stimulation, reaching its peak 4–10 s after initial stimulation. The single-RBC <sO₂> exhibited a biphasic response: it remained unchanged within 3–5 s after stimulation, then proceeded to rise, and peaked after 8–10 s. The response of and rO₂ also showed behaviors similar to that of <sO₂>.

The coupling between neurons and single RBCs also was imaged under continuous visual stimulation by flashing the LED light at 1 Hz. Single RBCs in the visual cortex region were imaged at 20–100 Hz, depending on the blood flow speed. Fig. 5 C and D show representative sO₂ images acquired without and with continuous visual stimulation, respectively. The <sO₂> decreased under continuous visual stimulation. We also measured the rO₂ while alternating the continuous visual stimulus on and off, as shown in Movie S6. All single-RBC parameters, including <C_Hb>, <sO₂>, , v_f, and rO₂, were compared without and with continuous visual stimulation. Each of the parameters was normalized to its mean value computed from the data without stimulation and plotted in Fig. 5E. Experimental results from three mice show that <C_Hb> did not change significantly, the <sO₂> decreased by 4 ± 0.8%, and the , v_f, and rO₂ increased by 53 ± 6%, 8 ± 1%, and 56 ± 9%, respectively.

With the single-RBC resolution, we further quantified the probability distributions of the single-RBC parameters. Fig. 5 F–I shows the cumulative distribution functions (CDFs) of the significantly changed parameters. With continuous stimulation, the single-RBC functions exhibited distributions different from those in the control experiment, i.e., lower <sO₂> and higher, v_f, and rO₂, consistent with the data shown in Fig. 5E.

Experimental Setup.

Building on the basis of high-resolution functional PA microscopy (12, 15–18), several innovations enable real-time spectral imaging of single moving RBCs. Two lasers are used to periodically generate two laser pulses 20 µs apart at 532 nm (an isosbestic wavelength of hemoglobin) and 560 nm (a nonisosbestic wavelength of hemoglobin), whose pulse widths are <10 ns. Both lasers operate at a 2-kHz pulse repetition rate. The two laser beams are merged into a single-mode optical fiber, then delivered to a PA probe. For pulse-to-pulse calibration, the energy of each laser pulse is detected by a biased photodiode. The laser beam from the fiber is focused onto targets through a pair of optical lenses (N.A. in water: 0.1), an acoustic–optical beam combiner, and an ultrasound lens. The optical lenses may be adjusted to coalign the acoustic and optical foci accurately. The acoustic–optical beam combiner, which is composed of two prisms and a coated aluminum layer in the middle, reflects light but transmits sound (18). The tight optical focus provides a 3.4-µm lateral spatial resolution (18), which was quantified by imaging a sharp edge in water and deriving the line spread function. Laser-excited PA waves are collected by the ultrasound lens, transmitted through the acoustic–optical beam combiner, and detected by a high-frequency ultrasound transducer (V214; Olympus NDT; 50 MHz central frequency). The PA signals are amplified, filtered, and digitized at 500 MHz. The ultrasound transducer provides an axial resolution of 15 µm, as determined from a shift-and-sum method (25). A maximum imaging depth of 1.2 mm was demonstrated in a living mouse (18). For single-RBC functional imaging, FOG reached a depth of 308 µm, as shown in Fig. S2. Because of light scattering and absorption in tissue, the imaging sensitivity changes with depth. Accurate measurement of hemoglobin concentration and oxygen release rate in absolute units remains a challenge for in vivo imaging. Here, we studied only the relative change of hemoglobin concentration and the relative oxygen release rate.

At each position, two laser pulses sequentially excite nearly the same region of the target to acquire two depth-resolved PA signals (A-lines). Taking a flow speed of 10 mm⋅s⁻¹ as an example, the target moves 0.2 µm during the wavelength switching, which is much smaller than the spatial resolution and thus is negligible. Because HbO₂ and HbR have molar extinction coefficients with different spectral characteristics, and PA signals are linearly related to the concentrations of HbO₂ and HbR at low-excitation laser energy [<100 nJ per pulse (12)], the relative concentrations of HbO₂ and HbR may be computed from the PA signals (1, 16) of the same RBC excited at 532 nm and 560 nm. From the HbO₂ and HbR concentrations, we can calculate the relative C_Hb and sO₂ of single RBCs.

The PA probe is mounted onto a fast voice-coil linear scanner to enable acquisition of at least 100 cross-sectional (B-scan) images per second. Mechanically scanning the entire PA probe maintains the acoustic–optical confocal alignment and therefore achieves higher SNR than pure optical scanning in a fixed acoustic focus (18, 24). A field-programmable gate array card (PCI-7830R; National Instrument) was programmed to synchronize the trigger signals and motion control commands. By fast scanning the PA probe along RBC flowing direction, we can image single RBCs flowing through the field of view, as well as directly measure the amount of oxygen bound to each RBC.

Experimental Animals and Approved Protocols.

The ears of nude mice (Hsd: Athymic Nude-Foxl^NU; Harlan Co.; 4–6 wk old, 20–25 g) and the brains of white mice (Hsd:ND4 Swiss Webster; Harlan Co., 4–6 wk old; 20–25 g) were imaged for all in vivo studies. During imaging, mice were placed on top of a 37 °C heating pad, secured with a head holder, and anesthetized with isoflurane [Isothesia (Isoflurane USP); Butler Schein Animal Health Supply]. Ultrasound gel was applied between the imaging areas and the water tank. After the ear imaging experiments, the mice recovered naturally. Brain imaging experiments were studied through a small craniotomy. After the brain imaging experiments, the mice were killed via cervical dislocation under deep anesthesia. The laser pulse energy used for the brain imaging usually was 40–50 nJ, but up to 100 nJ when deep RBCs were imaged. To offset skin attenuation, 80 nJ was used when the mouse ear was imaged. Note that the current experimental setup can image microvessel segments within a small angle (<10°) from the voice-coil scanning axis (x-axis). For capillaries outside the angular range, either the scanner or the animal was rotated to reduce the angle. All experimental animal protocols were approved by the Animal Studies Committee at Washington University in St. Louis.