Quantifying Injury Expansion in the Cervical Spinal Cord with Intravital Ultrafast Contrast-Enhanced Ultrasound Imaging

Abstract

Spinal cord injury is characterized by hemodynamic disruption at the injury epicenter and hypoperfusion in the penumbra, resulting in progressive ischemia and cell death. This degenerative secondary injury process has been well-described, though mostly using ex vivo or depth-limited optical imaging techniques. Intravital contrast-enhanced ultrasound enables longitudinal, quantitative evaluation of anatomical and hemodynamic changes in vivo through the entire spinal parenchyma. Here, we used ultrasound imaging to visualize and quantify subacute injury expansion (through 72 hours post-injury) in a rodent cervical contusion model. Significant intraparenchymal hematoma expansion was observed through 72 hours post-injury (1.86±0.17-fold change from acute, p < 0.05), while the volume of the ischemic deficit largely increased within 24 hours post-injury (2.24±0.27-fold, p < 0.05). Histology corroborated these findings; increased apoptosis, tissue and vessel loss, and sustained tissue hypoxia were observed at 72 hours post-injury. Vascular resistance was significantly elevated in the remaining perfused tissue, likely due in part to deformation of the central sulcal artery nearest to the lesion site. In conjunction, substantial hyperemia was observed in all perilesional areas examined except the ipsilesional gray matter. This study demonstrates the utility of longitudinal ultrasound imaging as a quantitative tool for tracking injury progression in vivo.

Introduction

Spinal cord injury (SCI) results in loss of sensorimotor function and substantially reduced quality of life for most patients (Adriaansen et al., 2016; Westgren and Levi, 1998). SCI is comprised of a primary injury phase, wherein the initial physical insult causes irreparable damage within the lesion center, and a secondary phase, during which the lesion progressively expands (Zhou et al., 2014). Vascular disruption plays a critical role in the secondary injury process. Expansion of the ischemic deficit generated by the initial insult contributes to persistent hypoxia and cell death (Tator and Fehlings, 1991). Furthermore, disruption of the blood-spinal cord barrier results in increased permeability and enhanced inflammatory response in the lesion center and the penumbra (i.e., hypoperfused perilesional tissue that is not fully ischemic) (Fitch and Silver, 1997; Tator and Fehlings, 1991; Whetstone et al., 2003). Elevated intraspinal pressure, driven by edema and tissue swelling, can also reduce the perfusion pressure in the spinal cord; this further exacerbates persistent hemodynamic disruption and contributes to worse patient outcomes (Khaing et al., 2017; Saadoun et al., 2017; Varsos et al., 2015). Vascular contributions to secondary injury progression are clearly significant. However, there is a dearth of in vivo longitudinal data tracking hemodynamic changes during the critical first few days post-SCI, especially in cervical contusion injuries, which are most representative of typical clinical presentation (Wang et al., 2022). Previous studies have relied either solely on histological and molecular methods, which offer only a single timepoint per animal, or light-based vascular imaging techniques such as laser speckle or two photon imaging, which are incapable of imaging the entire dorsal-ventral extent of the spinal cord (Cheng et al., 2019; Jacques, 2013; Wu et al., 2022).

Ultrasound imaging achieves sufficient penetration depth to assess blood flow throughout the full spinal cord, including both gray and white matter. Recent advancements in ultrafast plane wave imaging (~ 20 kHz) applied in conjunction with an intravenously administered microbubble contrast agent have enabled visualization and quantification of spinal hemodynamics with excellent spatiotemporal resolution (~ 100 um, < 1 sec) (Khaing et al., 2018; Tanter and Fink, 2014; Tremblay-Darveau et al., 2014). Previous studies have shown that metrics derived from acute contrast-enhanced ultrasound (CEUS) scans (area of ischemic deficit, time of contrast arrival) strongly correlate with injury severity and chronic behavioral deficits following thoracic contusion injury (Bruce et al., 2022). Furthermore, standard B-mode imaging can be used to provide complementary anatomical metrics by quantifying changes in tissue morphology and the extent of the hematoma. Given the ability to conduct repeated scans within the same animals at multiple timepoints, ultrasound is uniquely well suited for longitudinal assessment of injury expansion and evolution. Ultrasound imaging, and CEUS in particular, has yet to be comprehensively applied in the cervical spinal cord and is more commonly utilized at chronic timepoints rather than during the critical acute to subacute phase post-SCI when neuroprotective interventions may still be effective (Hilton et al., 2017; Khaing et al., 2023; Liu and Xu, 2012).

Here, we have applied CEUS imaging to quantify progressive hemodynamic changes within the first 72 hours following a unilateral contusion injury in the cervical spinal cord. We observed significant progressive hemorrhage through 72 hpi in conjunction with a regression in total cord parenchyma volume. CEUS further elucidated the temporal window for both expansion of the ischemic deficit and anatomical disruption of the vascular network. In terms of hemodynamic changes, we observed a sustained increase in vascular resistance, and significant hyperemia in all remaining perfused tissue (i.e., outside of the lesion center) by 72 hpi. Histological analysis supported these findings and elucidated the time course of tissue sparing, apoptosis, vascular disruption, and sustained tissue hypoxia following SCI. The current study represents a more thorough characterization of injury expansion than has previously been possible with existing technologies, which may play a role in future development and optimization of neuroprotective therapies.

Materials and Methods

Cervical spinal cord injury model

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Washington (Protocol #4512-01). Experiments were conducted in accordance with NIH Office of Laboratory Animal Welfare (OLAW) ARRIVE guidelines.

Female Long Evans rats (N=25; 8–12 weeks old; Harlan Labs, Indianapolis, IN) were anesthetized with isoflurane (5% induction, 1–3% maintenance in 1 L/min O₂). A tail vein catheter (24 Ga) was placed, flushed with heparinized saline (1%, 0.5 mL), and fitted with a three-way stopcock for administration of ultrasound contrast during image acquisitions (0.1 mL Definity, 0.2 mL saline flush per bolus; Lantheus, Billerica, MA). The area above the C2-T2 segments was shaved and sterilized prior to subcutaneous administration of topical analgesic (Lidocaine, 1.5 mg/kg, and Bupivacaine, 1 mg/kg). A 2.5 cm longitudinal incision centered over C4-C6 was made using a #10 scalpel blade. Following subperiosteal dissection of paraspinal muscles, a three-level laminectomy was performed to expose levels C4-C6 for epidural ultrasound imaging. Unilateral contusion injuries were generated at C5 using an Infinite Horizon device (Precision Systems and Instrumentation, VA; 150 kDyn with zero dwell time). Animals were stabilized at the C4 and C6 spinous processes during injury, and the C2 and T2 processes during ultrasound imaging using a custom frame. After the imaging session, the surgical site was closed, and animals were allowed to recover from anesthesia between acute and terminal imaging timepoints. Pain associated with the surgery was managed with buprenorphine (0.03 mg/kg SC) immediately before surgery and every 8 to 12 hours for 48 hours.

Experimental design and ultrasound imaging

Animals were randomly assigned to one of three separate cohorts. Epidural ultrasound imaging was conducted at baseline, acutely (~ 30 min) post-injury, and at either 4 hours post-injury (hpi; N=9), 24 hpi (N=8) or 72 hpi (N=8). All ultrasound imaging was performed in two anesthesia sessions per animal: baseline and acute post injury imaging sequences were performed during the first anesthesia session, and the final imaging session (at 4, 24, or 72 hpi) was performed under a second anesthesia session. Custom imaging sequences were implemented on a Verasonics Vantage research ultrasound system (Verasonics, Kirkland, WA) using a 15 MHz linear array transducer (L22–14vX, Vermon, Tours, France). Warmed saline and sterile ultrasound gel (Aquasonic) were used for acoustic coupling between the imaging transducer and the spinal cord. Each set of ultrasound scans was acquired axially, centered on C5, and consisted of the following: (1) a wide-beam 3D B-mode swept 6–7 mm along the rostrocaudal axis (50 μm step size); (2) plane-wave Doppler images acquired at the center of each of the three exposed segments; (3) plane-wave contrast-specific imaging to capture a dynamic bolus inflow curve; (4) sparse microbubble imaging 5–6 minutes after a bolus injection; and (5) a 3D plane-wave contrast-specific sequence swept 6–7 mm along the rostrocaudal axis (25 μm step size). Additional details on the imaging setup as well as the imaging sequences have been previously reported by our group (Bruce et al., 2022; Bruce et al., 2020; Harmon et al., 2022). Following terminal imaging, animals were euthanized using an overdose of pentobarbital anesthesia (Euthasol, IP, ≥50 mg/kg).

Hematoma, spinal cord parenchyma, and perfusion deficit volumes

Volumes were measured using a custom script written in Matlab (R2019b, Mathworks, Natick, MA). Researchers were blinded to the experimental condition during measurement. The hematoma and cord parenchyma volumes were measured from 3D B-mode acquisitions. Following contusion SCI, the hematoma appears as a highly echogenic region which can be clearly distinguished from tissue background. Volumes were normalized to either their corresponding baseline (parenchyma volume) or acute timepoint measurements (hematoma volume).

Perfusion deficit volume measurements were taken from 3D contrast-specific acquisitions. In brief, these consisted of short 40 frame plane-wave ensembles (667 Hz pulse repetition frequency; PRF) acquired at each stepped location. Each ensemble was filtered using the singular value decomposition (SVD) to remove corrupting signal from larger vessels and specifically analyze only microcirculatory flow (i.e., tissue perfusion) (Bruce et al., 2020). The first 5 projections were retained, and a power Doppler image was generated for each location. Subsequently, masking and volume measurements were conducted in the same manner as the B-mode measurements.

Bolus kinetics

Bolus acquisitions consisted of short 40 frame plane-wave ensembles acquired at an inter-ensemble repetition rate of 16 Hz. Imaging was initiated shortly prior to bolus injection and continued for 32 seconds total dwell time. Tissue perfusion signal was isolated using the SVD as described for the 3D contrast acquisitions. A lognormal curve fitting routine was applied on a pixelwise basis in order to extract quantitative measurements of relative tissue health while retaining spatial information. Specifically, the arrival time delay (ATD), rise time (RT), and area under the curve (AUC) were calculated. Physiologically, these measurements are analogous to vascular resistance (ATD, RT) and blood volume (AUC). A representative bolus inflow curve and a visual explanation of the parameters of interest is depicted in Fig 1. Parametric maps of each parameter were generated, centered on the injury epicenter, at each timepoint. Using a custom Matlab script, images were divided into regions of interest encompassing the ipsi- and contralateral sides of the spinal cord, and the gray and white matter. The ischemic deficit was masked out and excluded from analysis.

Fig 1. Contrast-enhanced ultrasound imaging (CEUS) and analysis.

Images depict C5 acutely following a unilateral contusion injury (30 min). (A-C) A contrast-specific pulsing scheme was used to cancel tissue background. Subsequent filtering using the singular value decomposition (SVD) separated tissue perfusion and vascular flow. (D-E) Ultrasound localization microscopy (ULM) was used to produce super-resolved maps of larger vasculature. Spatially discrete circulating microbubbles were localized and tracked across successive frames; tracks were accumulated across long acquisitions (> 20 sec) to generate ULM images. Perfusion and ULM images were used to quantify perfusion deficit volume and to assess progressive vascular deformation. (F) Bolus inflow kinetics were utilized to quantify “dynamic” parameters to complement the “static” anatomical parameters described in (A-E). Plots depict dynamic inflow curves in the contusion, penumbra, and contralateral gray matter; the regions of interest (ROIs) used are depicted in (B). Lognormal curve fits were used to derive the rise time (i.e., absolute vascular resistance) and area under the curve (i.e., blood volume). Arrival time was also quantified and normalized to the arrival time of the first bubble in the contralateral gray matter (arrival time delay, i.e., relative vascular resistance). The inset depicts the arrival time (arrows) and rise time (time to reach 95% of peak intensity; dashed black lines) for representative curves from the contralateral gray matter and penumbra. (G-I) Bolus parameters were quantified on a pixelwise basis to produce parametric maps.

Ultrasound localization microscopy

Between 15–20 longer contrast-specific acquisitions (720 frames, 400 Hz PRF) were captured for ultrasound localization microscopy (ULM) (Errico et al., 2015; Harmon et al., 2022). These were acquired at sparse microbubble concentrations such that each circulating microbubble appeared spatially discrete. Super-resolved vascular maps were generated by localizing and tracking individual circulating microbubbles between subsequent frames. Following SVD filtering to retain signal from larger vessels (projections 125–500), a cross-correlation based localization method was used prior to tracking with a partial assignment algorithm based on the Hungarian linker (Harmon et al., 2022; Song et al., 2018). Tracks shorter than 7 frames were excluded to reduce noise. Prior to track accumulation and display, each track was smoothed using a Savitsky-Golay filter (first order, 3 frame window) and linearly interpolated by a factor of 10. ULM images were binarized prior to skeletonization and extraction of a tortuosity index using established methods (Khansari et al., 2017; Telea and van Wijk, 2002).

Histology and immunohistochemistry

After obtaining terminal level anesthesia with Euthasol, animals were sacrificed via trans-cardiac perfusion with 200 mL of ice-cold PBS (pH 7.4) followed with 200 mL of 4% paraformaldehyde (PFA). Following post-fixation and tissue harvesting, spinal cord tissue was cryoprotected and immersed in optimal cutting temperature medium (Fisher Healthcare Tissue Plus O.C.T. Compound, Clear) before being frozen on dry ice. Spinal segments were sectioned at approximately −20°C using a cryostat (Leica, CM1850). 20 μm thick coronal sections were thaw mounted onto gelatin coated microscope slides and dried.

Standard Cresyl Violet and Eriochrome Cyanine (CV/EC) staining was performed to determine the extent of tissue sparing. Fluorescent immunohistochemistry was performed to analyze more specific markers (e.g., blood vessels, apoptotic cells). Briefly, tissue sections were thawed, crosslinked to the gelatin coating on the slide (4% PFA, 10 mins), and washed with PBS. Antigen retrieval was performed by submerging the slides in pH 6.0 citrate buffer (Ricca Chemical Company, Arlington, TX). Slides were washed with PBS and blocked in a solution consisting of 3% Normal Goat Serum (Vector Laboratories Inc, Burlingame, CA) and 0.3% Triton-X 100 with 1X PBS as the diluent. Slides were then incubated overnight at 4°C in a primary antibody solution made from the blocking solution. The following primary antibodies were used: rabbit anti-Caspase 8 (Novus, NB100-56116, 1:500), rabbit anti- GFAP (Dako, Z0334, 1:1000), and rabbit anti-Laminin (LN; Sigma Aldrich, L9393, 1:500). The next day, slides were washed with PBS and subsequently incubated at 37°C in a secondary antibody solution made from the blocking solution. The secondary antibodies used were: Alexafluor 568 goat anti-rabbit (Invitrogen, A11036, 1:500) and Alexafluor 488 goat anti-mouse (Invitrogen, A32723, 1:500). Slides were washed with PBS prior to applying DAPI (1:1000, 10 mins). A final wash was performed prior to cover-slipping with Fluoromount G mounting medium (Electron Microscopy Sciences, Hatfield, PA).

Tissue hypoxia was analyzed in separate cohorts of animals (i.e., no ultrasound imaging; 4 hpi N=5, 24 hpi N=6, 72 hpi N=5). IP injections of pimonidazole (Hypoxyprobe, Inc., 60 mg/kg) were administered 2 hours prior to sacrifice. Tissue preparation and immunostaining was conducted in the same manner as the other markers, except primary antibody incubation was conducted for one hour at room temperature (~22°C) rather than overnight. The following primary antibodies were used: goat anti-mouse biotinylated fluorophore (Hypoxyprobe Inc., HP1–200Kit, 1:50) and goat anti-rabbit GFAP (Dako, Z0334, 1:1000). Alexafluor goat anti-rabbit 647 (Invitrogen, AB-2535813, 1:500) and Alexafluor streptavidin 488 (Invitrogen, S11223, 1:500) were used as secondary antibodies. Slides were washed with PBS, stained with DAPI, and cover-slipped as described above.

Analysis was conducted using ImageJ (Schindelin et al., 2012). Researchers were blinded for all measurements. The area of spared tissue and number of apoptotic cells were manually analyzed. Photomicrographs of Laminin-stained sections were captured at constant microscope exposure time and zoom. Images were binarized in order to segment vessels from background, and the total LN⁺ stained area was used as a measure for total vessel coverage. The area of hypoxic tissue was measured in a similar manner; the threshold for image binarization was empirically selected to include the most stained cells while fully excluding background fluorescence. Measures were conducted either on the epicenter alone (vessel coverage), or on serial sections spaced 200 μm apart comprising a 3 mm (tissue sparing, apoptotic cells) or 5 mm (tissue hypoxia) rostrocaudal span.

Statistical analysis

Summary statistics are presented as mean ± standard error. Python (ver. 3.8.10) was used in conjunction with the Pandas (ver. 1.4.2), Scipy (ver. 1.8.1), Statsmodels (ver. 0.13.2), and Seaborn (ver. 0.11.2) packages for all statistical analyses and data visualization. One-way analysis of variance (ANOVA) tests paired with Tukey post-hoc analyses were used to determine statistical significance (α=0.05).

Results

Anatomical changes and progressive hemorrhage

Swept 3D B-mode scans were used to quantify both the volume of the spinal cord parenchyma within a 3 mm rostrocaudal window and the hematoma volume. Representative images of C5 at baseline, acutely post injury (30 min), and at either 4, 24, or 72 hpi are shown in Fig. 2 (A–I). While the spinal cord exhibited swelling through 4 hpi (1.15±0.02-fold change from baseline), the cord had regressed to baseline volume by 24 hpi (1.01±0.03-fold; p < 0.05 vs 4 hpi, N=8) and continued to regress through 72 hpi (0.93±0.03-fold; p < 0.05 vs 4 hpi, N=8; Fig. 2J). In contrast, the hematoma exhibited limited expansion between 4 hpi and 24 hpi (1.20±0.07- and 1.46±0.09-fold change from acute, respectively), and expanded significantly by 72 hpi (1.86±0.17-fold; p < 0.05 vs 4 hpi and 24 hpi, N=8; Fig. 2K).

Fig 2. Hematoma expansion and spinal cord parenchyma volume.

B-mode images are depicted for representative animals from the 4 hpi (A-C), 24 hpi (D-F), and 72 hpi (G-I) cohorts across all three imaging timepoints. (J) The spinal cord parenchyma volume was quantified within a 3 mm rostrocaudal window. Swelling was observed through 4 hpi prior to progressive volumetric reduction (p < 0.05, 24 hpi and 72 hpi vs. 4 hpi). By 72 hpi, the volume was smaller than at baseline (0.93±0.03-fold change from baseline). (K) The volume of the entire hematoma was measured at each timepoint. Significant progressive expansion of the hematoma was observed by 72 hpi (p < 0.05, 72 hpi vs. 4 hpi and 24 hpi; 1.86±0.17-fold change from acute).

Perfusion deficit expansion

Swept 3D CEUS scans were conducted to quantify the perfusion deficit volume. SVD filtering was applied to ultrafast contrast-specific image ensembles to segment and independently analyze tissue perfusion independent of flow in larger vessels. Images depicting the injury epicenter (C5) at baseline, acutely post-injury (30 min), and 72 hpi in a single animal are shown in Fig. 3 (A–C). The perfusion deficit continued to expand from 4 hpi to 24 hpi (1.53±0.11- and 2.24±0.27-fold change from acute, respectively; p < 0.05, N=8) prior to plateauing between 24 hpi and 72 hpi (2.45±0.20-fold, p > 0.05, N=8; Fig. 3G).

Fig 3. Perfusion deficit expansion and vascular deformation.

(A-C) Perfusion images of C5 are depicted at baseline, acutely post-injury, and 72 hpi for the same representative animal. The ischemic deficit appears as a dark region on the right side of the image. (D-F) Ultrasound localization microscopy (ULM) images illustrate progressive disruption of vascular anatomy; in particular, the central sulcal artery (yellow arrow) continues to bend through 72 hpi. (G) The volume of the ischemic deficit was quantified from perfusion images using a stepped 3D acquisition. The deficit expanded primarily within the first 24 hpi (2.24±0.27-fold change from acute; p < 0.05, 4 hpi vs. 24 hpi and 72 hpi; p > 0.05, 24 hpi vs. 72 hpi).

Deformation of vascular anatomy

ULM was conducted at a single imaging plane due to the necessity of acquiring upwards of 20 seconds of data per output image. Super-resolved vascular maps were generated at the same imaging plane for each timepoint to track deformation of the same vessels across time. In addition to a progressively more defined ischemic deficit, substantial deformation of the central sulcal artery was observed by 72 hpi (Fig. 3D–F). A tortuosity index was calculated for each central sulcal artery where the ULM image was of sufficient quality for robust skeletonization. Progressively larger changes in tortuosity index were observed with increasing time post-injury (4 hpi: 2.82±0.84-fold change from baseline, N=5; 24 hpi: 4.85±2.01-fold, N=4; 72 hpi: 9.04±3.22-fold, N=4), though this trend was not statistically significant.

Hemodynamic changes and hyperemic response

Parametric maps depicting bolus inflow kinetics were generated for a single imaging plane due to long acquisition times (32 sec) and the necessity of capturing a full time-intensity curve. Two sets of parameters were quantified – those relating to vascular resistance (RT, ATD) and an analog to local blood volume (AUC). RT represents an absolute vascular resistance, whereas ATD is normalized within a given image and depicts spatial differences in vascular resistance with respect to the injury site. Representative examples of RT and ATD maps are shown for an animal from the 72 hpi cohort in Fig. 4 (A–F).

Fig 4. Changes in vascular resistance.

Representative images of rise time (RT; A-C) and arrival time delay (ATD; D-F) are displayed at baseline, acutely post-injury, and 72 hpi. The perfusion deficit was masked out and excluded from analysis; low contrast inflow resulted in noisy parameter estimates. RT is analogous to the absolute vascular resistance, whereas ATD is analogous to a relative vascular resistance or gradient within an image. (G-H) RT and ATD were quantified using four ROIs – either the ipsi- or contralateral gray matter (GM) or white matter (WM). A significant increase was observed acutely post-injury in all regions except the contralateral WM (p < 0.05, injury vs. baseline). In both the ipsilateral GM and the contralateral WM, a significant progressive increase was observed through 72 hpi (p < 0.05, 72 hpi vs. injury). A significant increase in ATD was observed acutely post-injury in all regions except the contralateral GM (p < 0.05, injury vs. baseline). This was sustained through 4 hpi but returned to baseline levels by 24 hpi (p > 0.05, 24 hpi vs. baseline). While absolute vascular resistance continued to increase by 72 hpi, spatial differences within the imaging plane were resolved within 24 hpi.

Parameters were quantified within four ROIs – ipsilateral and contralateral gray matter and white matter. To focus on quantifying the relative health of spared tissue, the area of perfusion deficit was excluded from ROI analysis. RT increased significantly acutely post-injury in all regions except the contralateral white matter (Fig. 4G; p < 0.05, acute vs baseline). RT returned to baseline levels by 72 hpi in the contralateral gray matter only (p > 0.05). In the ipsilateral white matter, elevated RT was sustained, and in the ipsilateral gray matter and contralateral white matter, a further increase in RT was observed by 72 hpi (p < 0.05, 72 hpi vs injury). Arrival time delay increased significantly acutely post-injury in all regions except the contralateral gray matter and was sustained through 4 hpi (Fig. 4H; p < 0.05, acute and 4 hpi vs baseline). This returned to baseline by 24 hpi and remained at baseline levels through 72 hpi for all regions (p > 0.05).

Parametric maps of AUC are shown in Fig. 5 (A–C). Even in tissue outside of the primary ischemic deficit, a global decrease in blood volume was observed acutely post-injury (Fig. 5D; p < 0.05, all regions). However, while this reduction was sustained through 4 hpi in the gray matter (p < 0.05 vs baseline), AUC had returned to baseline levels in all regions except the ipsilateral gray matter by 24 hpi and progressed to exhibiting significant hyperemia by 72 hpi (p < 0.05, 72 hpi vs baseline). The ipsilateral gray matter recovered to baseline levels by 72 hpi.

Fig 5. Blood volume reduction and subsequent hyperemia.

(A-C) Representative images of the area under the bolus inflow curve (AUC; blood volume) are displayed at baseline, acutely post-injury, and 72 hpi. White arrows indicate example regions with visible hyperemia at 72 hpi. (D) AUC was quantified using the same ROIs drawn for RT and ATD analysis. The perfusion deficit was excluded from analysis. A consistent global decrease in blood volume was observed acutely post-injury (p < 0.05, injury vs. baseline); this was sustained through 24 hpi in the ipsilateral GM (p < 0.05, baseline vs. 4 hpi and 24 hpi). Whereas the ipsilateral GM exhibited recovery to baseline levels by 72 hpi (p > 0.05, baseline vs. 72 hpi), all other regions exhibited a significant hyperemic response (p < 0.05, baseline vs. 72 hpi).

Histological Analysis

Tissue was collected from each of the 4, 24, and 72 hpi groups to elucidate cell and tissue-level changes during injury expansion. Spared tissue analysis was conducted using Cresyl Violet (CV; nuclei) and Eriochrome Cyanine (EC; myelin) stained sections (Fig. 6A–C). In tandem, activated Caspase 8 staining was conducted to quantify the spatiotemporal extent of apoptotic cell death following injury (Fig. 6D–F). A significant reduction in spared tissue was observed between 4 and 24 hpi both at the epicenter (8.84±0.40 vs. 7.64±0.36 mm²; p < 0.05) and caudal to the lesion (10.3±0.35 vs. 8.87±0.30 mm²; p < 0.05). Progressive tissue loss was observed through 72 hpi, particularly rostral to the lesion center (10.3±0.33 vs. 8.86±0.32 mm²; p < 0.05; Fig. 6G). Alongside significant tissue loss, a sharp increase in the number of apoptotic cells was observed by 24 hpi (11.2±2.77 vs. 138±15.1 Cas8⁺ cells; p < 0.05) and sustained through 72 hpi (155±21.8 Cas8⁺ cells; p > 0.05; Fig. 6H).

Fig 6. Lesion expansion and apoptotic cell death.

(A-C) Representative images of the injury epicenter stained with Cresyl Violet (CV; nuclei – purple) and Eriochrome Cyanine (EC; myelin – blue) are shown for 4, 24, and 72-hours post injury (hpi). (D-F) Representative images of tissue stained for caspase 8, zoomed on the ipsilateral side, are displayed. (G) Spared tissue analysis conducted using CV/EC-stained tissue revealed significant tissue loss at the epicenter and caudal to the lesion within 24 hpi (8.84±0.40 vs. 7.63±0.36 mm²; 10.3±0.35 vs. 8.87±0.30 mm²; p < 0.05), and progressive tissue loss rostral to the lesion between 24 and 72 hpi (10.3±0.40 vs. 8.86±0.32 mm²; p < 0.05). (H) A significant increase in the number of apoptotic cells (Cas8⁺) was observed between 4 and 24 hpi (11.2±2.77 vs. 138±15.1 Cas8⁺ cells; p < 0.05), which was sustained through 72 hpi (155±21.8 Cas8⁺ cells; p > 0.05).

Hypoxyprobe and laminin staining were used to quantify tissue hypoxia and vessel loss following SCI. Composite images showing overlaid Hypoxyprobe, glial fibrillary acidic protein (GFAP; astroglia cells), and nuclear staining (DAPI) are shown for each timepoint in Fig. 7A–C. Representative laminin-stained tissue sections, zoomed on the ipsilesional side, are shown in Fig. 7D–F. Elevated tissue hypoxia was observed within 4 hpi and was sustained through 72 hpi (p > 0.05, 4 hpi vs 72 hpi; Fig. 7G). Significant vessel loss was observed by 24 hpi (0.22±0.02 vs. 0.10±0.02 mm² LN⁺ area; p < 0.05, 4 hpi vs. 24 hpi) and sustained through 72 hpi (0.15±0.02 mm²; p > 0.05, 72 hpi vs. 24 hpi; Fig. 7H).

Fig 7. Sustained tissue hypoxia and decrease in vascularity.

Hypoxyprobe staining was used to determine the spatial distribution of hypoxic cells at 4, 24, and 72 hpi. (A-C) Representative images of cervical spinal cord tissue at the injury epicenter (C5) are displayed for each timepoint. Sections were labeled with Hypoxyprobe (green), glial fibrillary acidic protein (GFAP; red) and a nuclear stain (DAPI; blue). (D-F) Laminin antibody staining (LN; green) was used to detect blood vessels within the spinal cord. Representative images of the ipsilateral side of C5 are displayed for each timepoint. (G) The hypoxic area (e.g., area positive for Hypoxyprobe) was measured within a 5 mm rostrocaudal segment of the spinal cord at 200 μm intervals. Tissue hypoxia was sustained through 72 hpi (p > 0.05, 4 hpi vs. 72 hpi). (H) Vessel coverage (i.e., total LN⁺ area) was significantly reduced by 24 hpi (0.22±0.02 vs. 0.10±0.02 mm²; p < 0.05, 4 hpi vs. 24 hpi), and remained low through 72 hpi (0.15±0.02 mm²; p > 0.05, 72 hpi vs. 24 hpi).

Discussion

The current study provides, for the first time, a comprehensive description of anatomical and hemodynamic disruption during the secondary injury phase post-SCI in vivo. First, we used 3D B-mode imaging to examine tissue swelling followed by subsequent regression as clearance of debris and necrotic tissue was initiated, and to quantify progressive bleeding through 72 hpi (Neumann et al., 2009; Wang and Sun, 2011). The spinal cord exhibited swelling through 4 hpi and had significantly regressed by 72 hpi, in tandem with progressive expansion of the hematoma (Fig. 2). Extravascular blood and its components can induce a mass effect onto surrounding parenchymal tissue and can also be directly neurotoxic, thereby contributing to further swelling and tissue loss (Leonard et al., 2015; Losey et al., 2014). Ferroptosis, a programmed cell death process that occurs in the presence of excessive iron, has also been suggested to be a contributing mechanism to secondary degeneration post-SCI (Chen et al., 2020; Li et al., 2022; Li and Jia, 2023). Limiting hematoma expansion within the subacute phase serves as a potential therapeutic target for future work.

Contrast-specific plane wave imaging enabled segmentation and independent analysis of flow in the microcirculation and in larger vasculature, a capability that at time of publication is unique to our group (Bruce et al., 2020; Harmon et al., 2022). Our imaging and processing methods enable analysis and quantification of flow in the capillary network, the critical level at which nutrient and oxygen exchange occur. This represents a significant advancement compared to conventional ultrasound imaging, which is only able to visualize and measure blood flow upwards of 1 cm/sec. Here, these specialized CEUS imaging modalities were used to visualize and quantify the spatiotemporal evolution of the ischemic perfusion deficit, deformation of vascular anatomy, and hemodynamic disruption in the microcirculation in remaining, non-contused tissue. These quantitative features will be used in future work to identify tissue at risk of further secondary damage in order to guide neuroprotective interventions in a targeted manner.

Significant progressive expansion of the perfusion deficit was observed through 24 hpi and was sustained at 72 hpi despite the regression in total parenchyma volume (Fig. 3). Histological analysis corroborated this finding; the number of actively apoptotic cells (Cas8⁺) increased significantly between 4 and 24 hpi and remained steady through 72 hpi (Fig. 6). Vessel loss (i.e., reduced LN⁺ area at epicenter) followed the same temporal pattern (Fig. 7). In conjunction, a significant reduction in spared tissue was observed between 24 and 72 hpi, particularly rostral to the lesion center (Fig. 6). Microvascular disruption during the secondary injury cascade has been well-documented, largely using histological methods (Mautes et al., 2000; Ng et al., 2011). In addition to disruption of the blood spinal cord barrier and extravasation of neurotoxic blood components, endothelial cell death and capillary fragmentation contribute to progressive hemorrhagic necrosis and ischemia-driven tissue loss (Figley et al., 2014; Noble et al., 2002; Popovich et al., 2012; Simard et al., 2007; Simard et al., 2012; Simard et al., 2013). The current work is novel in that it visualizes the evolution of this degenerative process in vivo by tracking the same injuries at multiple timepoints. Furthermore, rather than looking solely at vessel structure and density ex vivo, the current study quantifies actual functional perfusion in situ, rendering it well-suited for both longitudinal preclinical studies and future translational work where noninvasive methods are required.

Ultrasound localization microscopy (ULM) revealed significant progressive deformation of the central sulcal artery closest to the lesion site and surrounding microvasculature (Fig. 3D–F). While previous studies have reported significant increases in vessel tortuosity in the chronic phase post-SCI (8 weeks), the current study uniquely tracks individual vessels in the critical subacute phase (Beliard et al., 2022). Both the ATD and RT, derived from bolus inflow CEUS acquisitions, were significantly elevated immediately following injury; rise time, a measure of global vascular resistance, continued to increase through 72 hpi especially in the ipsilateral gray matter and the contralateral white matter (Fig. 4). This is likely due in part to the increase in CSA tortuosity (Han, 2012; Helmberger et al., 2014). Given that our bolus parameters represent estimates of microcirculatory hemodynamics, it is also likely that alterations in vessel structure and function at the capillary level are contributing to the observed increase in vascular resistance. Previous work from our group has also described the utility of ULM in quantifying acute pathological anatomical changes in the microvessel network in situ, which were then confirmed with histology (Harmon et al., 2022).

In addition to the well-documented multifaceted breakdown of the microvessel network described above, early-stage regenerative processes may also play a role in elevated vascular resistance. Previous studies have reported endothelial cell infiltration and increased microvessel density at the lesion center within the first 3–4 days post-injury, suggesting that 72 hpi may capture the very early stage of endogenous angiogenesis following SCI (Casella et al., 2002; Imperato-Kalmar et al., 1997; Loy et al., 2002). As further supporting evidence, we observed a trend toward increasing LN⁺ stained area between 24 and 72 hpi in the current study (Fig. 7H). Tortuous capillaries (both existing and newly forming) have been observed in a mouse ear wound-healing model within the first 3 days post-wound and have been implicated in the initial steps of angiogenesis (Chong et al., 2017). Increased capillary tortuosity may therefore play a role in the observed increase in microcirculatory vascular resistance in the current study. This will be directly investigated in future studies; in vivo, ULM applied to microcirculatory flow enables tracking of bubble trajectories and direct quantification of disruption of the capillary network (Harmon et al., 2022).

Despite the observed increase in vascular resistance, significant hyperemia was observed in white matter and contralateral gray matter at 72 hpi (Fig. 5). A number of older articles report observing an early hyperemic response which subsided within the first few hours post-injury (Kobrine and Doyle, 1976; Smith et al., 1978). More recent clinical studies have reported metabolic patterns characteristic of hyperemia in spinal cord tissue following contusive injuries, suggesting this phenomenon is preserved in humans (Chen et al., 2016; Phang et al., 2016; Saadoun and Papadopoulos, 2020). While hyperemia has been reported with some regularity, there is no conclusive consensus on its functional consequences following SCI and most observations rely on invasive, ex vivo, or depth-limited measurements (e.g., intraparenchymal catheters, histology, laser Doppler). It has been proposed that hyperemia may be a downstream effect of sustained tissue hypoxia and may be actively detrimental to recovery (e.g., by reperfusion-type injury or edema) (Dumont et al., 2001; Sjovold et al., 2013; Streijger et al., 2017). More prolonged hyperemia, such as that observed in the current study, has been reported following brain injuries and has been implicated in the secondary injury cascade if not brought into appropriate balance (though this can also vary depending on the presence or absence of comorbid intracranial hypertension) (Kelly et al., 1996; Kinoshita, 2016; Nair and Rajagopal, 2017). It is therefore possible that the hyperemic response observed in this study contributes to the secondary injury cascade, but further work is required to elucidate the role and complete temporal evolution of hyperemia in the penumbral tissues post-SCI. A lack of adequate in vivo blood flow monitoring techniques may partially explain the lack of a comprehensive and mechanistic description of post-traumatic penumbral hyperemia; CEUS is uniquely well-suited to measure this in follow-up subacute and chronic studies.

As discussed, vascular degeneration and tissue remodeling are hallmarks of the secondary injury cascade (Mautes et al., 2000; Ng et al., 2011; Tator and Fehlings, 1991). The current study addresses two longstanding limitations of work in this area. First, the use of in vivo ultrasound imaging enabled investigation of tissue through the entire depth of the spinal cord, including both gray and white matter, which represents a substantial improvement over light-based in vivo imaging methods (e.g., laser speckle, two-photon) which are depth-limited in part due to light scattering in the heavily myelinated white matter (Cheng et al., 2019; Jacques, 2013; Wu et al., 2022). Second, repeated ultrasound imaging sessions in the same animals enabled longitudinal assessment of injury expansion. By contrast, myriad studies have relied solely on histological or other cellular and molecular-specific techniques. These only provide information at a single, terminal timepoint, with tissue removed and anatomically disrupted from its in vivo configuration. Here, we used histology as a secondary technique to corroborate and expand upon our ultrasound findings.

The current study provides a description of vascular and hemodynamic alterations in situ following SCI. Future work will leverage these data and this imaging technique to acutely evaluate the efficacy of neuroprotective interventions. For example, Glibenclamide is a promising drug with an established pre-clinical history and ongoing clinical trials for spinal cord injury and traumatic brain injury (e.g., NCT05148403, NCT05426681) (Popovich et al., 2012; Simard et al., 2012; Simard et al., 2013). It has been shown to limit progressive bleeding and capillary fragmentation post-SCI. Another therapy to which our imaging technique would potentially be well-suited is surgical decompression (i.e., durotomy). Our lab has previously published on the efficacy of durotomy in reducing elevated intraspinal pressure post-SCI, which may alleviate vascular disruption due to “spinal compartment syndrome” (Khaing et al., 2021; Khaing et al., 2017; Yang et al., 2022). CEUS imaging could be used to monitor spatiotemporal hemodynamic changes following durotomy and could inform the critical window within which durotomy must be performed to be effective (e.g., within 4 hpi).

One of the primary limitations of the current study is the relatively short time window post-SCI. In future work, longer time scales will be studied to characterize injury expansion and tissue remodeling beyond the 72 hpi endpoint used in the current study. The first few hours to days post-SCI exhibit rapid injury expansion and represent a critical window within which neuroprotective interventions may be applied (Ahuja et al., 2020). However, deeper understanding of hemodynamic changes in the late subacute and chronic phases post-injury may be critical to aid in the development and implementation of regenerative therapies, which will be more broadly applicable for patients already living with SCI.

Conclusions

In summary, significant progressive hemorrhage, tissue remodeling, and hemodynamic disruption were observed in the first 72 hours post contusion type SCI in the cervical spinal cord. The current study is descriptive of secondary injury progression but did not utilize a therapeutic intervention. Future studies will utilize these techniques and the data from the current study to inform neuroprotective treatment strategies and to acutely evaluate their efficacy.

• Ultrasound enables robust quantification of spinal cord injury expansion in vivo

• Expansion of the ischemic deficit peaked within 24 hours post-injury

• Hemorrhage and hemodynamic disruption progressed through 3 days post-injury

• Histology (spared tissue, apoptosis, hypoxia, vessels) corroborated ultrasound data

• Ultrasound imaging may facilitate development of neuroprotective therapies