1. Introduction

The hippocampus plays a role in learning, memory formation, spatial navigation, and emotional behavior. Disturbances in the structure and function of the hippocampus have been associated with age-related cognitive decline and Alzheimer’s disease [1]. In contrast to our extensive understanding of cortical vasculature, there exists a significant gap in our knowledge regarding the hemodynamic characteristics of the hippocampus and their potential relationship with cognitive impairment.

Vasculature in the cortex and hippocampus differ in orientation and vascular density, with inverted and fewer blood vessels in the hippocampus [2]. The different vasculature distribution risks hippocampal oxygen supply in disease due to its lower resting blood flow, decreased neurovascular coupling capabilities, and lower blood oxygenation [3]. These anatomical differences may put the hippocampus at major risk of accumulating neurotoxic proteins that lead to vascular dysfunction and neurodegeneration [4]. Understanding the regulation of oxygen supply in the hippocampus can advance knowledge on interventions targetting early stages of neurodegenerative disease development.

Optical imaging methodologies are well suited for imaging blood flow in small vessels and detecting microvascular abnormalities because they offer high spatial and temporal resolution, enabling the visualization of fine vascular networks and rapid changes in blood flow [5]. Optical imaging techniques can reflect the neuronal activity in the brain by measuring changes in blood oxygenation [6], single capillary hemodynamics [7], and changes in oxy- and deoxyhemoglobin [8]. Two-photon microscopy (TPM) could be considered the gold standard for correlating neuronal activity to brain hemodynamics with high spatial and moderate temporal resolution [9]. However, TPM signal depends on the delivery of exogenous fluorescent dyes, requiring, in some cases, repeated intravascular injections of such dyes that are usually conjugated with high-molecular dextrans that can affect blood viscosity [9]. Another disadvantage of TPM is the limited translation to clinical settings. In light of this constraint, it becomes imperative to employ techniques that use endogenous signals such as fluctuation patterns of red blood cell speckles, to investigate the vascular dynamics in various brain regions.

Laser speckle contrast imaging (LSCI) examines real-time brain perfusion dynamics. LSCI typically utilizes a near-infrared laser and a standard CMOS camera to record an interference (or speckle) pattern formed by back-scattered light. Moving particles create intensity fluctuations, blurring the speckle pattern when integrated over a finite exposure time. Blurring degree is directly related to the particles’ dynamics and, thus, to blood flow, and can be quantified in terms of the speckle contrast $K$ [10]. Such simplicity in hardware and data processing aspects, as well as non-invasiveness and high spatiotemporal resolution, resulted in the broad translation of LSCI to clinical application to monitor blood flow in the brain [11–13], skin [14–16], retina [17,18] and other organs. These characteristics also make LSCI an obvious choice when measuring perfusion in the exposed hippocampus. Access to the hippocampus by removing the overlying cortex [3,19] has given access to an essential area of the brain, allowing the application of standard optical imaging techniques such as TPM and LSCI and extending our knowledge of brain hemodynamics beyond the cortex. Understanding the optical properties of various biological tissues will refine research in laboratory animals and the development of blood vessel phantoms for testing intraoperative clinical diagnostic devices aimed at real time perfusion monitoring during surgical interventions [20]. Although perfusion monitoring in clinical settings is currently only applicable to superficial areas of the brain [21], the optical properties of hippocampal tissue are relevant to consider as techniques progress, especially due to the area’s increased vulnerability to small vascular changes [3].

In this study, we applied LSCI to monitor perfusion in mouse hippocampus via a chronic, optically transparent window. After inspecting contrast images, we found that the hippocampal blood vessels appear out-of-focus at all objective-sample distances, unlike the similar caliber vessels in the cortex. To study the possible causes of such observation and explore how the hippocampus differs from the cortex from the laser speckle imaging point of view, we used dynamic light scattering imaging (DLSI) in the same animals. DLSI enables estimation of contributions from different scattering regimes, including static scattering and prevalent dynamics such as multiple scattering unordered motion (MU), single scattering from unordered motion, or multiple scattering from ordered motion (SU/MO) versus single scattering ordered motion (SO) [22,23]. We hypothesized that, due to the inverse vascular topology in the hippocampus compared to the upper layers of the cortex, there might be an increase in the contribution of multiple-scattered photons and static scattering. DLSI results supported the hypothesis, showing a change in the form of the intensity correlation function g2, which corresponded to 13 times increased contribution of MU regime in large hippocampal vessels and 4 times increased amount of static scattering in the parenchyma when fitting the DLSI model.

2. Methods

2.1 Animal preparation

For in vivo imaging, C57BL/6 mice (n=4, age 12 weeks, Janvier, Denmark) were used, n=1 with a chronic cortical cranial window and n=3 with chronic hippocampal windows. Animals were housed in conditions of 12 h light/dark cycle, 22–24$^{\circ }$C, 55 $\pm$ 10% humidity, and ad libitum access to food. All experimental protocols were approved by the Danish National Animal Experiments Inspectorate (permit no.: 2022-15-0201-01188) and conducted according to the ARRIVE guidelines and guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes. For surgical procedures and imaging sessions, animals were anesthetized in an induction chamber with 3% isoflurane in 100% oxygen and reduced to 1-2% on a homeothermic heating pad to maintain body temperature at 37$^{\circ }$C. Analgesic (Buprenorphine, 0.1 mg/kg body weight), anti-inflammatory (Carprofen, 10mg/kg), antibiotic (Ampicillin, 200 mg/kg), and edema-reducing corticosteroid (Dexamethasone, 4.8 mg/kg) were administered intraperitoneally before incision according to body weight. Local anesthesia (Xylocaine, 10mg/ml) was injected subcutaneously at the surgical site. The animal preparation and surgical procedures have been previously described in detail [3,24–26]. Briefly, a small region of the skin covering the cranium was removed, and the edges were secured to the bone using adhesive. The exposed skull was scraped to remove the periosteum and scored to increase surface area. For cortical and hippocampal windows, a craniectomy was performed over the barrel cortex (1.5–2 mm anterior-posterior – AP, 3 mm medio-lateral – ML to Bregma) and the dCA1 hippocampus (2 mm AP, 2 mm ML to Bregma), respectively. The location of the craniectomy was determined based on skull size and mouse age. For the cortical window, an optically transparent glass coverslip of 4mm in diameter was placed onto the window and fixed to the skull with cyanoacrylate glue. After craniectomy, the hippocampal window requires aspiration of approximately 1.3 mm of the overlying cortex. A constant saline flow ensured the hydrophobic tissue and fibers loosened for easier removal of cortical matter. A 3 mm glass coverslip glued to a stainless-steel tube (1.5 mm x 3 mm) was inserted into the space and fixed to the skull with cyanoacrylate glue. A metal head plate was glued to the skull for both procedures to fix the head during the imaging sessions. The skull was then covered with dental cement to seal the surgical region and fix the window to the head plate. It is important to note the craniectomy and aspiration of tissue may result in bleeding from the broken vessels in and around the cortical matter, and some bleeding may also occur on the surface of the hippocampus during removal of the overlying corpus callosum fibers. Active bleeds were stopped with a hemostatic sponge or by flushing with saline before placement of the window to prevent obscuring of the glass. Windows with damaged vessels and blood in the field of view at time of recording may not reflect accurate physiological conditions of the vasculature and were not included. Mice were placed in a heated recovery chamber during anesthesia waning before they were returned to their home cage. During a recovery period of at least five days, weight and behavior were closely monitored, and mice received pain relief medication (Buprenorphine, Carprofen), antibiotic (Ampicillin), and fluids (isotonic glucose, saline) IP for four days following surgery as well as soft food at the bottom of the cage to encourage weight gain.

In vivo imaging

The animals were induced for anesthesia with isoflurane (3%) before the imaging sessions. During the LSCI imaging sessions, the mouse was placed on a homeothermic heating pad, and the isoflurane concentration was reduced to 1-2%. For all imaging, mice were head-fixed to the imaging stage using clamps attached to the implanted metal head plate. The imaging stage components are modified from [26] and can be found on our GitHub (https://github.com/CFIN-Optics/Laser-Speckle-Imaging-of-the-Hippocampus). LSCI and DLSI were performed to assess brain perfusion in each mouse using a custom-made imaging system. The light from a fiber-coupled volume holographic grating stabilized laser diode (785 nm, Thorlabs FPV785P), controlled with a laser driver (Thorlabs LDC210C) and temperature controller (Thorlabs TED200C), and mounted on the arm perpendicular to the imaging axis, is delivered on the sample co-axially using a polarising beamsplitter positioned behind the objective. The back-scattered light collected by the objective (Leica UVI 6.3x, NA=0.13) passes through the polarizing cube, which acts as a cross-polarizer, tube lens (Thorlabs TTL200-B), and a long-pass filter (cut on at 750nm). It is then directed to a 9:1 non-polarising beamsplitter and delivered to the sensors of both a high-speed (Photron Nova S6, 20x20 $\mu m^2$ pixels) and a conventional (Basler aca2040-90um NIR, 5.5x5.5 $\mu m^2$ pixels) CMOS camera for DLSI and LSCI data collection, respectively. The DLSI recordings were performed for 15 seconds at a frame rate of 30000 frames (512x384 pixels, $\approx$3x2.25 mm field of view) per second and an exposure time of 33 $\mu s$. The LSCI recordings were performed for 120 seconds at a frame rate of 50 frames (2048x2048 pixels, $\approx$ 3x3 mm field of view ) per second and exposure time of 5000 $\mu s$.

Two-photon microscopy (TPM) was used to assess the topology of the vascular network in two exemplary mice, one with a cortical window and one with a hippocampal window. Mice were imaged in a conscious state and habituated to restraint in increasing increments of 30 minutes over five days for a total habituation time of 150 minutes. A tail vein catheter for infusion of fluorophores was placed under anesthesia before the imaging session. Anesthesia was induced with 3% isoflurane in 0.6L/min oxygen and maintained at 1-2%. During anesthesia waning, mice were fixed to the imaging stage to prepare for awake-restrained imaging. Imaging was performed on an Ultima-IV two-photon system (Bruker Corporation, Billerica, MA, United States) with PrairieView software version 5.5 (Bruker Corporation). A load of 300$\mathrm {\mu }$l of 5% (5mg/ml) solution of Texas Red 70kDa was administered via a tail vein catheter for vessel visualization and z-stack acquisition of the cortex. The 25x objective (Olympus, Scaleview, NA = 0.9, WD = 8 mm) was used to acquire z-stacks of $190 \mu m$ (FOV=515x512, $461.8\mu m^2$, laser excitation 950nm, 587 GaAsP). A load of 180 ${\mathrm {\mu }}$l of 5% (5mg/ml) solution of FITC 70 kDa was administered via a tail vein catheter for vessel visualization and z-stack acquisition of the hippocampus. A 25x objective (Olympus SCALEVIEW, NA=0.9, WD=8mm) was used to acquire z-stacks of $300\mu m$ (FOV=515x512, $461.8\mu m^2$, laser excitation 810nm, exponential laser compensation 300-350 pockels power and 800 GaAsP).

Data analysis

LSCI and DLSI data processing details are available in other studies [22,27], so here, we provide a concise overview. For LSCI analysis, the contrast was calculated as $K=\frac {\sigma (I)}{<I>}$, where $\sigma (I)$ and $<I>$ are the standard deviation and mean of intensity. These parameters were calculated either for spatial contrast $K_s$ over 5x5 pixels neighborhood using a standard CMOS camera or for temporal contrast $K_t$ over all DLSI frames after averaging them into frames with surrogate exposure times matching the longest time lag (T=33ms). For DLSI analysis, the time-average intensity autocorrelation function for each pixel was calculated as:

TPM-acquired angiograms were reconstructed in ImageJ using bleaching correction, contrast enhancement with normalization (saturated pixels=0.4%), background subtraction (rolling ball=50), and 3D filtering (median 3D: x=2, y=2, z=2) and presented as Z projected maximum intensity projections (MIP) and 3D renderings (3D viewer).

3. Results

Examples of time-average contrast images of the cortex and hippocampus are shown in Fig. 1, A, B. In addition to noticeable topology differences and reduced density of visible vessels, most mid-sized and large vessels in the hippocampus appear out-of-focus and do not become sharper at different object distance values. Such appearance could be explained by the quality of the cranial window or scattering properties of the hippocampal tissue. To exclude the first option, we performed TPM angiograms of the same cranial windows. In Fig. 1, C, D, TPM angiograms appear to be of similar quality for both cortex and hippocampus, suggesting the window quality is unlikely to be the cause of distortions in the hippocampus contrast images. Aspiration does not seem to significantly affect the hippocampal vasculature, despite higher expression of inflammation near the hippocampal cranial window [3]. Second Harmonic Generation (SHG) scans of the hippocampal tissue in other internal experiments did not show collagen assembly that would indicate damage (data not included). Furthermore, angiograms (Fig. 1, E, F) show, as expected, that vascular geometry in the hippocampus is inverted compared to the cortex; larger vessels are located deeper in the tissue, while smaller vessels and capillaries sprout towards the surface. Such topology can result in differences in the light scattering properties, particularly the dynamics regime and amount of static scattering.

 

Fig. 1. A,B - Examples of laser speckle contrast images of the cortex and hippocampus, respectively. C,D - Two-photon maximum intensity projection of z-stack angiograms of cortex and hippocampus . E,F - 3D rendering of z-stacks (cortex=190 ${\mathrm {\mu }}$m, hippocampus=300${\mathrm {\mu }}$m)

Download Full Size | PDF

DLSI was used to probe the light-scattering characteristics of the hippocampus tissue and its effect on the laser speckle fluctuations. As the light scattering properties are primarily reflected in shape differences, to visualize them in Fig. 2, we further normalized the normalized intensity autocorrelation function as $g2(\tau )_{norm}=\frac {g2(\tau )-g2(0)}{g2(0)-1}$. For reference, g2(0) was $\approx 1.27$ and $1.26$ in the parenchyma and large vessels, respectively. From Fig. 2(A)-(D), which represents the data from the same animals as in Fig. 1, it is clear that pixels belonging to the hippocampus decorrelate faster than the cortex for early time lags but become slower with time. The same behavior is observed in the normalized g2 averaged over regions of interest (ROIs) in the vessels of comparable size (Fig. 2, E) and parenchyma with similar contrast values (Fig. 2, F) in the cortex (n=1) and hippocampus (n=3). Such shape differences can be expected when comparing the field correlation functions with different $n$ values, e.g. $g_1^{n=0.5}$, which corresponds to multiple scattering unordered motion dynamics regime, and $g_1^{n=1}$, which corresponds to versus single scattering from unordered motion or multiple scattering from ordered motion. As we know from previous DLSI studies, large and mid-sized cortex vessels are typically characterized by the SU/MO and SO regimes, and therefore, a stronger contribution of the MU regime could explain the observations we made in the hippocampus.

 

Fig. 2. Dynamic Light Scattering measurements. A,B - Example of decorrelation progression (based on normalised g2) at 0.066 ms time lag value for cortex and hippocampus, respectively. C,D - decorrelation progression at 3.3 ms. E,F - normalized g2 curves from vessels of comparable sizes and parenchymal regions with similar contrast values from cortex (n=1) and hippocampus (n=3). All subplots show that normalized g2 in the hippocampus decreases faster than in the cortex for early time lags but then slows, eventually resulting in more decorrelation of the cortex intensity at later time lags.

Download Full Size | PDF

Results of fitting the g2 data with the DLSI model (Eq. (2)) are shown in Fig. 3 and Table 1. The decorrelation time constant $\tau _c$ appears to be comparable, though slightly smaller in the hippocampus, particularly in parenchyma. Interestingly, this occurs despite higher contrast values ($K_{t}$) in the hippocampus than in the cortex. Such disparity between $tau_c$ and $K_t$ also reflects the effects of other light scattering parameters - dynamics regime and static scattering. Supporting our hypothesis, the MU dynamics regime dominates in the hippocampus’ larger vessels with $d\approx 0.06pm0.05$ compared to $d\approx 0.78$ in the cortex and parenchyma $d\approx 0.0126pm0.005$ compared to $d\approx 0.143$ in the cortex. The static scattering contribution ($1-\rho$) is also increased in the hippocampus, reaching upwards of 0.6 in the parenchyma and 0.1-0.3 for deep large vessels, while for the cortex, the corresponding values are 0.1-0.2 in the parenchyma and typically below 0.1 in the visible vessels. Despite the results matching our expectations, it is also important to note the reduction in $R^2$ in the hippocampus compared to the cortex, which likely reflects that the existing DLSI model does not fully account for the dynamic light scattering complexity in the hippocampus tissue.

 

Fig. 3. Example of Dynamic Light Scattering Imaging fitting results for cortex (A,C,E) and hippocampus (B,D,F). A,B - decorrelation time $\tau _c$ in microseconds. C,D - dynamics regime, represented by the parameter $d$ ranging from MU (d=0) to SU/MO regime (d=1). E,F - static scattering contribution.

Download Full Size | PDF

Table 1. Contrast values and fitted DLSI model parameters for different ROIs in the cortex (n=1) and hippocampus (n=3). Cortex$_{v}$, Cortex$_{p}$, and Cortex$_{ff}$ correspond to the values measured in the chosen vessel, parenchymal regions, and averaged over the entire field of view. Similarly, for the hippocampus, presented as mean$\pm$ standard deviation calculated over the corresponding regions in 3 animals.

View Table

4. Conclusion

We have demonstrated the application of laser speckle imaging, specifically LSCI and DLSI, to monitor perfusion in the hippocampus. Although LSCI is, perhaps, the most effective tool for 2D imaging of the perfusion dynamics, it meets several limitations in the hippocampus and requires additional considerations. The spatial features of most of the large vessels appear more blurred compared to similar caliber vessels of the cortex. This blurring makes it more challenging, if not nearly impossible, to accurately measure vessel diameters or mask them. Furthermore, it increases the difficulty of identifying parenchymal regions where the presence of blurred large vessels does not alter the signal. Other considerations are related to interpreting the hippocampus’ contrast values and dynamic light scattering properties. While these properties can be associated with neuronal activity [28] or differences in neuronal density between cortex and hippocampus [29], the distribution and dynamics of red blood cells across the sampling volume over time suggest that the primary contributions to differences in scattering effects in our study are vascular tree topology and tissue composition. Our data suggest that multiple scattering unordered motion dynamics are prevalent in the hippocampus, including the large vessels, meaning that the corresponding contrast model must be used to estimate the decorrelation time accurately. Furthermore, the increased amount of static scattering in the hippocampus’ vessels and parenchyma will make tools like Multi-Exposure Speckle Imaging or DLSI preferable compared to conventional LSCI, or would require using some strategies for estimating the static scattering contribution and correct LSCI interpretation. One should also notice that the hippocampus data’s goodness of fit $R^2_{hippocampus}=0.9947$ is decreased compared to the cortex $R^2_{cortex}=0.9984$. Although the difference appears to be small, the comparable $R^2$ change can be achieved in the cortex by using a simpler g2 model with fever parameters (e.g., without the mixed dynamics). It implies that the existing DLSI model might be missing some of the features crucial for hippocampus imaging, for example, the effect of the deeper location of large vessels, which might cause some of the photons detected in the same pixel to be scattered from particles moving with different dynamics (e.g., from capillaries and large vessels). While the aim of the study was not on performing associations between behavior, perfusion, and cell morphology, we recommend future investigations to assess the influence of the hippocampal preparation on these parameters.