Multiscale X-ray phase-contrast CT uncovers adaptive changes and compensatory mechanisms of circulatory pathways during acute liver injury

Abstract

Intrahepatic circulation is essential for the repair of acute liver injury (ALI); however, very limited information is available concerning changes in the circulatory pathways during ALI. Therefore, multi-scale X-ray phase-contrast CT combined with three-dimensional (3D) visualization is used to quantitatively analyze the intrahepatic circulation pathway (including the hepatic vein, portal vein and hepatic sinusoid) in the mouse model via the intraperitoneal injection of carbon tetrachloride (CCl₄) from acute injury to recovery. The results demonstrate that the liver still preserves some vessel-like channels accessed to the central vein when the injury causes the severe collapse of the hepatic sinusoids that cannot be observed in two-dimensional pathologic slices. Moreover, angiogenesis is observed in the terminal branches of the hepatic vein and portal vein. Additionally, we extend the two-dimensional primary lobule to a 3D model and find that the sinusoids in zone III have the most severe injury. The sinusoids in different zones also show changes in parameters such as density and mean diameter during the ALI. In conclusion, phase-contrast CT can reveal the intact vascular system within the liver lobes, thus providing critical information for studying the mechanisms involved in the evolution of circulatory structures from damage to repair.

Multiscale phase-contrast CT combined with 3D visualization provides insights into the morphological mechanisms of damage, compensation and repair of blood circulation pathways within the intact liver lobe during acute liver injury in mice.

Introduction

Acute liver injury (ALI) is a serious clinical syndrome that is characterized by sudden and massive death of hepatocytes¹. The etiology of ALI involves exposure to toxins, drug or alcohol abuse, and hepatitis virus infection. It manifests cellular dysfunction, immune cell infiltration, massive hepatocyte apoptosis, and necrosis of the liver tissue^2–5. Mild ALI can repair itself through hepatocyte proliferation, whereas severe or persistent liver injury often leads to liver fibrosis and cirrhosis; eventually, this may develop into liver cancer and lead to death⁶. The rapid disease course and complex etiology of ALI make its treatment very difficult; therefore, monitoring the entire process of ALI development and repair may help to propose therapeutic strategies for blocking disease progression to a more severe stage in a timely manner.

Hepatotoxic acute liver injury (especially that induced by pharmacologic toxins) is a common clinical event that usually manifests as a loss of hepatocyte-specific function and cellular necrosis, which occurs around the central vein^7,8. In addition, direct toxic stress can be delivered to other targets that are considered important for the initiation and progression of significant tissue damage, such as liver sinusoidal endothelial cells (LSECs), which leads to hepatic sinusoidal collapse, occlusion, congestion, and other adverse phenomena^9–11. The process of liver repair is dominated by cell proliferation, which fills in the original necrotic areas and gradually restores liver function. The physical sensation that is generated by changes in the morphological parameters of sinusoids is indispensable in the process of liver repair and termination¹². The hepatic sinusoid is an important part of the microcirculatory structure, and damage to it directly induces microcirculation disorders. Most of the current research on ALI has focused on substances that can ameliorate ALI; however, basic research on the vasculature is still scarce, and detailed reports on hepatic sinusoids during ALI are lacking. In addition, the studies involving vasculature have focused on two-dimensional pathological analysis, which makes it difficult to determine the complex three-dimensional (3D) structural changes in hepatic sinusoids, as well as to demonstrate how the liver maintains circulation when blood outflow is obstructed by sinusoidal collapse as part of blood transport^13,14. To answer these questions, a suitable imaging system is needed that can perform multiscale 3D imaging of the liver from the macroscopic vascular tree to the microscopic hepatic sinusoids to analyze the alterations in the different blood vessels.

Phase-contrast computed tomography (PCCT), which is a cutting-edge imaging modality, enables full 3D visualization with isotropic resolution and without destructive slicing of the specimen in contrast to 2D histology¹⁵. Conventional CT relies on the attenuation of X-rays when they pass through an object to generate contrast. As a result, dense structures containing elements with high atomic numbers produce high contrast, whereas soft tissues that are mainly composed of low atomic number elements (such as carbon, hydrogen, and oxygen) exhibit rather low contrast¹⁶. Unlike absorption-based imaging, PCCT utilizes the phase shift of X-rays by the investigated object as the source of contrast, which is theoretically approximately 1000 times more sensitive than the absorption-based approach for soft tissue imaging¹⁷. Currently, PCCT has been widely used for imaging studies of various organs, such as the liver, kidney, brain, and spinal cord, with particular advantages in the micromorphological examination of vascular lesions in these organs^18–21. Especially in the field of hepatic imaging, neoangiogenesis and abnormal vascular architecture in fibrosis/cirrhosis, as well as the entire vasculature system, have been clearly demonstrated on PCCT images^21–23. In recent years, submicro- or nano-PCCT technology has advanced, and multiscale imaging systems based on this technology can simultaneously visualize the vascular tree in intact liver lobes, as well as the hepatic sinusoids in hepatic lobules, thus providing a technological basis for probing the dynamic evolution of integral circulatory pathways in liver diseases.

Herein, the carbon tetrachloride (CCl₄)-induced liver injury model was used to simulate ALI. This model is similar to human acute drug-induced liver injury in terms of mechanism and symptoms²⁴. Subsequently, multiscale PCCT was used to observe the adaptive changes in the macroscopic vascular system and microscopic hepatic sinusoidal structure, as well as the junctions between them, during the onset, progression and repair of ALI. Finally, these changes were quantitatively analyzed to fully elucidate the morphological mechanisms of circulatory injury, compensation and repair in ALI.

Results

PCCT correlated well with the corresponding histologic sections

In the acute liver injury model, PCCT correlated well with histologic sections of different vascular structures in the liver lobes. The entire vascular tree from the normal and injured liver was imaged at 3.25 μm resolution (Fig. 1a, b), and the terminal tissue of interest was imaged at 0.65 μm resolution; at this resolution, the PCCT demonstrated the morphology of the hepatic sinusoids, as well as the ability to image necrotic areas (Fig. 1c, d). PCCT can be considered a 3D virtual pathology technique that provides coherent spatial information compared to histologic sections; moreover, when combined with 3D visualization techniques, PCCT enables multidirectional observation and further quantitative analysis of the liver vasculature. As shown in Fig. 1e, the 3D structures of the liver during blood circulation were demonstrated, including the portal vein (PV) tree and the hepatic vein (HV) tree for blood distribution and collection, the basic unit of the liver - the hepatic lobule - responsible for blood transport, and the sinusoidal capillaries within the hepatic lobule.

Fig. 1. X-ray phase-contrast CT correlated well with histologic sections.

a, b HE-stained sections and PCCT of intact liver lobes imaged at 3.25 μm resolution in the normal and injury groups. c, d HE-stained sections and PCCT of the terminal tissue of interest imaged at 0.65 μm resolution in the normal and injury groups. The green color represents the necrotic area. e 3D reconstruction of the hepatic circulation; the purple color indicates hepatic vein (HV), the pink color indicates portal vein (PV), and the red color represents the hepatic sinusoids. Red and yellow arrows represent the direction of blood delivery.

The hepatic veins and portal veins tended to change synergistically

Adaptive changes from liver damage to repair were observed in both the PV and HV systems (Fig. 2a). 3D vascular reconstruction demonstrated that both veins tended to change synergistically (Fig. 2b, c). The magnified images of vascular branching showed that the variation in the number of small blood vessels from the PV and HV was prominent. The number of small blood vessels exhibited a pattern of initial increases and subsequent decreases, peaking on the fifth day (mean, PV: normal group: 21.87 ± 1.46; 1-day: 24.47 ± 2.47; 3-day: 30.20 ± 2.78; 5-day: 36.53 ± 1.92; 7-day: 30.87 ± 1.46; 9-day: 25.13 ± 1.51; F = 115.18, P < 0.001; HV: normal group: 20.07 ± 1.98; 1-day: 21.73 ± 1.75; 3-day: 28.40 ± 1.50; 5-day: 34.80 ± 2.96; 7-day: 28.67 ± 2.06; 9-day: 24.40 ± 1.59; F = 105.78, P < 0.001) (Fig. 2d). The interaction effect of the location (terminal or internal) of vascular branches and the time of disease progression on the Murray’s deviation was obtained through the two-way repeated measures ANOVA, which was conducted on both vascular systems (PV: F = 9.70, P < 0.001, HV: F = 11.64, P < 0.001). This demonstrated that there was a distinct Murray’s deviation pattern between terminal and internal vascular branches with the progression of the disease. As illustrated in Fig. 2e, f, the variation in Murray’s deviation was more pronounced at the terminal vessel branches and higher than in the internal branches of the vessel in both vascular systems. Post hoc multiple comparisons by the Bonferroni test demonstrated that there were no significant differences between groups in the Murray’s deviation of the internal branches as the disease progressed in both vascular systems (PV: F = 0.64, P = 0.67, HV: F = 0.59, P = 0.70). Conversely, the experimental group was significantly different from the normal group in both vessels in the terminal branches on Murray’s deviation (PV: F = 25.93, P < 0.001; HV: F = 25.87, P < 0.001). Detailed statistical parameters were described in Supplementary Appendix E1. This result suggested that the disturbance in the macrovasculature mainly occurred in the terminal branches.

Fig. 2. Adaptive changes in the 3D morphology of macrovasculature in acute liver injury.

a 3D reconstruction of the intact liver lobule. b, c 3D reconstruction of the vascular tree of the hepatic vein (HV) and portal vein (PV) with magnified vessel segments in the purple dashed box and the green dashed box. d The number of small blood vessels in the PV and HV. n = 15 vessel segments in each group were included in the calculation; mean ± SD was displayed in bar graphs. Significant difference between two groups with different letters, not significant between two groups when one letter is the same. e, f Murray’s deviation of the vessels of the PV and HV at the internal bifurcation (n = 30) and the terminal bifurcation (n = 30), the horizontal lines of the box plots represent the maximum, upper quartile, median, lower quartile, and minimum values. The purple color indicates HV, and the pink color indicates PV.

Acute liver injury disrupted the structure of the microcirculation—reduced connections were evident between the hepatic sinusoids and the central vein

High-resolution imaging at 0.65 µm was employed to image the terminal blood vessels. From the 3D virtual pathology sections, it was clear that the necrotic areas were effectively captured and exhibited the same trend as the necrosis contained in the pathologic sections (Fig. 3a, b). Through the combination of high-resolution imaging and 3D visualization techniques, the 3D hepatic lobular structure (previously shown as a pattern diagram) was effectively segmented and reconstructed in this study (Fig. 3c). The volume of necrotic tissue gradually increased from the normal group to the 3-day group, peaking on the third day with a large cavity surrounded the central vein. From the 5-day group to the 9-day group, the number of sinusoids in the necrotic area gradually increased. In the 9-day group, the vicinity of the central vein had almost returned to a normal state (Fig. 3c, e). In addition, the number of connections between hepatic sinusoids and the central vein changed with time from the normal group to the 9-day group (mean, normal group: 83.50 ± 5.08; 1-day: 34.40 ± 4.55; 3-day: 13.90 ± 2.88; 5-day: 25.50 ± 3.44; 7-day: 46.50 ± 5.64; and 9-day group: 61.50 ± 7.52; F = 248.69, P < 0.001 with significant differences between all groups) (Fig. 3d, f). There were six times more connections in the normal group than in the 3-day group, thus leading to obstruction of blood flow to the central vein on the third day, which severely impaired microcirculation. Based on the 3D reconstruction of the hepatic lobules in the 3-day group, it could be concluded that the connecting channels of the hepatic sinusoids to the central vein remained present even during severe necrosis (Fig. 4a, b). The mechanism through which the remaining channels manage to deliver blood to the central vein was shown in Fig. 4. The 3D structure of one of the preserved channels was reconstructed, and both transverse and longitudinal hepatic sinusoids were connected to the channel (Fig. 4c). In addition, blood in collapsed hepatic sinusoids could achieve blood conduction through alternative pathways (Fig. 4d, e). Thus, through the interconnections of the hepatic sinusoids, the channel allowed rapid drainage of blood from this region into the central vein, and we call this channel the ‘fast channel’.

Fig. 3. CCl₄ caused hepatocellular necrosis, hepatic sinusoidal absence, and reduced blood effluent channels within the liver lobules.

a HE-stained sections of the normal and experimental groups. b 3D virtual pathologic sections of the normal and experimental groups. c 3D structure of the hepatic lobules and necrotic areas within them. d The vascular centerline of the hepatic sinusoids and the central vein were included in the orange box in (c). e Relative necrosis volume fraction in the ROI (n = 10). f Number of hepatic sinusoidal channels connecting to the central vein in the ROI (n = 10), all error bars represent the standard deviation. Significant difference between two groups with different letters, not significant between two groups when one letter is the same. The purple color represents the central vein, the pink color represents the interlobular vein and the red color represents the hepatic sinusoids.

Fig. 4. 3D visualization of residual connection channels in severe necrosis.

a 3D structure of liver lobules with severe necrosis. b Channels to complete circulation were preserved within the necrotic area (arrows point to the connection pathways). c The connection channel indicated by the red arrow in b unfolded, and both the transverse and longitudinal hepatic sinusoidal networks were connected to this channel. d Longitudinal viewing channel. e Transverse viewing channel. The blocked blood could be transported through other pathways. The dotted line represents the pathway in which blood efferent was blocked, the arrow represents the pathway in which blood was able to be efferent and the asterisk represents the breakpoint of the hepatic sinusoid at the edge of the necrotic region. The purple color indicates hepatic vein, and the pink color indicates portal vein.

Adaptive changes in the hepatic sinusoids in the primary lobule

As blood flows from the interlobular veins to the central vein, there is functional compartmentalization of hepatocytes between afferent and efferent vessels. Based on the three-zone theory of the primary lobule, the hepatic sinusoids were divided into three zones of equipotential (Fig. 5a, b). The hepatic sinusoids in all three zones underwent corresponding adaptive changes with time (Fig. 5c–g). For changes in hepatic sinusoidal diameter, all five experimental groups showed an initial increase and a subsequent decrease in the mean hepatic sinusoidal diameter in all three zones, with temporal inflection occurring on Day 5 for zone I and on Day 3 for zones II and III (Fig. 5h). Moreover, the differences in diameter gradually changed in different zones within each group. In the normal group, the average diameter of hepatic sinusoids in zone II was smaller than that in zones I and III, and the average diameters in zones I and III were closer in value to each other. As the injury progressed, in the 1-day group, the hepatic sinusoids in all the zones swelled, and the diameters were close in value to each other. On the third day, the average diameter in zone III became the largest, whereas that in zone I became the smallest. Afterwards, in the 5-day group, the diameter in zone I reversed to become the largest at this time, and zones II and III were close together in value. Finally, on Days 7 and 9, the mean diameter of the hepatic sinusoids and the differences in the three zones were closer in value to those of the normal group. All three zones in the experimental group showed a decrease followed by an increase in the relative surface area of the hepatic sinusoids. The temporal inflection time was the same as that for the diameter, and the remaining changes were opposite to the diameter (Fig. 5i). When the spatial density of the hepatic sinusoids was measured, it was similar in the three zones of the normal group. In the experimental groups, the hepatic sinusoids located in zones I and II had similar densities in each group and between the different groups, both of which were lower than those in the normal group. The main change in density occurred in zone III, where it tended to decrease and then increase, reaching its lowest value on the third day (Fig. 5j). The results of the Kruskal-Wallis test indicated that there was no statistically significant difference in the mean diameter and relative surface area of the hepatic sinusoids in the three regions between the normal group and the 9-day group (except for the relative surface area in zone II between the normal group and the 9-day group). This suggested that the morphology and structure of the hepatic sinusoids tended to normalize (Fig. 5h, i). Detailed statistical parameters are shown in Supplementary Appendix E2.

Fig. 5. Adaptive changes in the hepatic sinusoids within the three zones of the primary lobule.

a 3D structure of the primary lobule in the normal group. b Division of the three zones. c–g 3D structures of the primary lobules, zones I, II, and III at different time points. h Mean diameter of the hepatic sinusoids. i Relative surface area of the hepatic sinusoids. j Density of hepatic sinusoids. N = 17 primary lobules per group were included for statistical analysis. The horizontal lines of the box plots represent the maximum, upper quartile, median, lower quartile, and minimum values. Significant difference between two groups with different letters, not significant between two groups when one letter is the same.

Discussion

The present study employed multiscale PCCT in a mice model of CCl₄-induced ALI to provide comprehensive 3D visualization of the complex vasculature in the liver (from the macrocirculation of the portal and hepatic venous system to the microcirculation of the hepatic sinusoidal structure). The accuracy of the technique was confirmed by comparison with corresponding histologic sections. By observing and quantitatively analyzing the 3D vasculature architecture, we were able to gain new insights. In the liver injury stage (days 1–3), severe damage results in the collapse of the hepatic sinusoids in zone III, thus causing microcirculatory disturbances. The liver then adjusts to circulatory disturbances by preserving part of the connections in zone III to form vascular “fast channels” and by terminal vessel hyperplasia of the PV and HV to generate new hepatic sinusoidal connection channels. During the repair phase (days 5–9), the hepatic sinusoids located at zone III gradually proliferate to form new circulation channels, followed by gradual regression of vascular terminal hyperplasia.

In recent years, several imaging techniques have been used to effectively depict the anatomical structure and 3D spatial characteristics of macroscopic liver circulation. Nevertheless, the structure of the microcirculation within the hepatic lobules remains ambiguous. Clinical imaging techniques, including CT and MRI, are limited in their ability to capture fine structures due to insufficient resolution. Micro-CT offers high resolution, but poor soft tissue contrast necessitates the use of contrast agents. To date, optical imaging techniques such as scanning electron microscopy (SEM) and confocal microscopy have been employed to visualize the 3D microstructure within the liver^12,25,26, but these techniques are confined to imaging superficial sinusoidal network due to limited tissue penetration depth. PCCT is an innovative and nondestructive imaging technique that uses phase shift as the imaging principle and is promising for observing intricate spatial structures in soft tissues. In this study, we used this technique to achieve 3D visualization from the complete vascular tree to the hepatic lobules to the hepatic sinusoids by changing the resolution, confirming that this technique can be a powerful tool for the complete study of the hepatic vasculature and providing a foundation for the systematic analysis of ALI.

In this study, we observed critical morphological structural characteristics that enabled continued circulation during the severe period of ALI. A portion of the channels connected to the central vein were still preserved, which could not be observed in the 2D pathologic sections. In addition to the connectivity properties of the channels, we designated the remaining channels as “fast channels”. Due to the collapse of the hepatic sinusoids and the redirection of blood, more blood flowed into the remaining channel; additionally, with the expansion of the hepatic sinusoids, the blood could flow more rapidly through the channel, which enables the blood blocked in the liver to drain into the central vein as quickly as possible.

We found that vascular proliferation plays an important role in the compensation for and recovery from ALI. Vascular proliferation occurred in both the PV and HV. The neovascular and original vasculature exhibited the same microenvironment through virtual pathology in multiple directions. Moreover, new connecting channels were formed between the new PV branches, HV branches and the original or newborn HV branches, PV branches to compensate for the loss of connecting channels between the hepatic sinusoids and the central vein due to the injury. In the early stage of repair, the adaptive changes in macrovasculature lagged the changes in microcirculatory structure. As the number of hepatic sinusoids gradually recovered, the necrotic area gradually shrank, but the macroscopic vascular proliferation continued. At that time, the proliferating vessels occupied the original space of the hepatic sinusoids, thus causing the distance from the inlet vessels to the outlet vessels to shorten, which reduced the oxygen and nutrient consumption from the blood in the hepatic sinusoids; consequently, the hepatocytes were provided with more nutrients to boost their reproduction. Subsequently, the hyperplastic vasculature began to vanish, which was accompanied by continued hepatocyte proliferation and increased access to the central vein. Finally, the macrocirculation and microcirculation were similar to those of normal structures. The evolution of the circulatory structure at different times during ALI can be summarized as follows: from a less vascular and more hepatic sinusoidal connection in the normal state (Fig. 6a) to a more vascular and less hepatic sinusoidal connection during the period of injury (Fig. 6b, c), followed by a more vascular and more hepatic sinusoidal connection at the early stage of repair (Fig. 6d) and a subsequently less vascular and more connected connection at the later stage of the repair (Fig. 6e). Thus, it is speculated that vascular proliferation has more of a structural role during the injury period, compensating for the loss of hepatic sinusoidal pathways. During the recovery period, vascular proliferation plays more a functional role by providing more oxygen and nutrients.

Fig. 6. The evolution of the circulatory structure at different times during liver injury.

a Vascular structure of the normal group. The black dashed box represents the microenvironment of the hepatic vein (HV) branches and portal vein (PV) branches. b Hepatic sinusoidal collapse due to CCl₄. c Vascular proliferation generated new channels between the PV branches and HV branches to compensate for collapsed hepatic sinusoids. d In the early period of recovery, the vasculature continued to proliferate, and the connection of the hepatic sinusoids to the central vein increased. e In the later stage of recovery, the vessels regressed, and the connection of the hepatic sinusoids to the central vein normalized. The blue arrows represent blocked blood to achieve blood efflux through other pathways, and the brown, orange, and blue dashed boxes represent combinations between the newborn and original vessels, which behave consistently in the microenvironment. The dotted lines on the vessels represent regressed hyperplastic vessels. The purple color indicates HV, and the pink color indicates PV.

In previous research on hepatic sinusoids, regions of interest were frequently chosen at random, thus ignoring the spatial heterogeneity of the hepatic sinusoids in the liver lobules. Therefore, this study utilized the primary lobule as the fundamental structural unit for assessing the difference in hepatic sinusoids among three different regions. In the normal group, the mean diameter of the hepatic sinusoids in zone II was smaller than that in zones I and III, and the relative surface area was the largest. It is possible that because the hepatic sinusoids of zones I and III play a major role in blood transfer, with a limited number of connections to the vasculature, a larger average diameter allows for better blood transportation. The hepatic sinusoids in zone II were found in the middle of the hepatic lobule. A smaller diameter likely slows the blood flow, whereas a larger relative surface area facilitates material exchange between the blood and hepatocytes. As the disease developed, the differences among the three zones also changed. The density of the liver sinusoids in zone III adjacent to the central vein gradually decreased, and the connections between the sinusoids and the central vein were reduced, thus leading to obstruction of the outflow of blood and causing the overall average diameter of the liver sinusoids in the three zones to increase. When the hepatic sinusoids in zone III were severely damaged, their density decreased, and the efferent channels were dramatically reduced, thus causing blood congestion in the hepatic sinusoids in zone II and the remaining channels and resulting in significantly greater increases in the diameters of the hepatic sinusoids in zones II and III than in zone I. In the early stage of repair, the continued proliferation of the macrovessels allowed for more blood flow into the liver, therefore causing expansion of the hepatic sinusoids in zone I. At this time, hepatocytes multiplied, the number of hepatic sinusoids increased, and the volume of necrotic areas gradually decreased. The dilated hepatic sinusoids were compressed by hepatocytes, and the more connections between the sinusoids and the central vein allowed more outlets for the blood to flow out, resulting in little change compared to that on the third day in the hepatic sinusoids of zones II and III. With liver repair, the liver sinusoidal parameters within the primary lobule and differences among its three zones gradually normalized on Days 7 and 9. From this study, it was concluded that, in addition to the adaptive changes produced by the hepatic sinusoids near the central vein, which have been predominantly mentioned in previous studies^27–29, the hepatic sinusoids in zones II and I still had morphological alterations, thus affecting the structure and function of the microcirculation.

To validate the findings in the CCl₄ model, an imaging experiment was complemented with another classic mouse model of acute liver injury, i.e., the bile duct ligation (BDL)-induced cholestatic liver injury model^30,31 (detailed results in Supplementary Appendix E3). Hepatocellular necrosis in the BDL model was primarily due to cholestatic toxicity and its location was not fixed in the liver, unlike the CCl₄ model where it was concentrated around the central vein. Although the two models differed in terms of causative factors and the location of necrosis, there were similarities in the adaptive changes in circulatory structures in response to circulatory disturbances caused by hepatic necrosis. Regarding the macrocirculation, both models showed vascular proliferation at the end of the portal and hepatic veins, which shortened the distance from the entrance to the exit of the blood circulation and accelerated the rapid outflow of blood. These changes reduced blood exchange and increased the efficiency of blood transport to compensate for the circulatory impairment caused by the necrosis. In addition, we retrospectively investigated the alterations in circulating structures in human liver injury based on pathological sections and found that the hepatic sinusoids surrounding the necrosis also expand (detailed results in Supplementary Appendix E4). This finding was consistent with the morphological changes, such as the increased diameter of the hepatic sinusoids, observed in the mouse models of liver injury. Overall, vascular proliferation of the portal and hepatic veins, along with adaptive changes in the hepatic sinusoids surrounding the necrosis, may be the main way to compensate for circulatory disturbances caused by the necrosis in liver-damaging diseases. Although PCCT has demonstrated excellent imaging capabilities for multiscale imaging of liver injury, the technique has one limitation. This was an ex vivo study of liver tissue, and the realization of liver imaging using PCCT in vivo remains challenging due to inherent technical limitations. The technology that was used in our study is not widely available.

In conclusion, we comprehensively observed the adaptive changes in the different vascular structures through which blood circulation during the process of ALI from injury to recovery and demonstrated evidence via 3D morphology that circulation can continue even with severe injury. We also demonstrated the evolution of blood circulation patterns at different time points. The detailed adaptive changes in the 3D morphology of blood vessels and the process of autonomous recovery by the liver provide a concrete foundation and potential targets for subsequent studies into treating acute liver injury.

Methods

Sample preparation

All the animal experiments were conducted in accordance with the guidelines for the care and use of laboratory animals approved by the Ethics Committee of Experimental Animal Welfare, Tianjin Medical University General Hospital (IRB2022-DW-12). Intraperitoneal injection of CCl₄ solution (purchased from Aladdin Industrial Corp, Shanghai, China) was used to induce ALI in mice. Sixty male wild-type C57BL/6 mice (eight to ten weeks old, weight, 20–22 g, purchased from Beijing Vital River Laboratory Animal Technology, Beijing, China) were randomly divided into six groups: the normal group and Groups 1, 3, 5, 7 and 9 days after CCl₄ treatment (ten mice were included in each group). The mice were maintained with free access to food and water in a temperature-controlled room with a 12-h light/dark cycle. A single intraperitoneal injection of 20% CCl₄ dissolved in olive oil (purchased from the National Pharmaceutical Group Chemical Reagent Co., Ltd, Shanghai, China) was administered to the mice at a dose of 100 µl/10 g body weight (the day of CCl₄ injection was designated as day 0), after which the mice were sacrificed at 1, 3, 5, 7 and 9 days after CCl₄ treatment. The ten mice in the normal group were not subjected to any intervention and were euthanized simultaneously on Day 0 to illustrate the normal hepatic structure. All of the liver samples were removed after the mice were sacrificed and preserved in 10% neutral buffered formalin solution for the subsequent experiments. Before image acquisition, selected liver samples were subjected to different concentrations of ethanol (50%, 70%, 80%, and 95% for two hours each and 100% ethanol solution for 24 h to achieve gradient dehydration).

Image acquisition

Phase contrast imaging experiments of the liver samples were performed at the BL13HB Beamline station of the Shanghai Synchrotron Radiation Facility (SSRF). The X-ray beam was produced by a 16-pole wiggler source and monochromated by liquid-nitrogen cooled Si (111) double crystal monochromator tuning into an energy range from 8 keV to 72.5 keV. A high-resolution X-ray detector consisting of an sCMOS detector (Hamamatsu, Japan) and an X-ray conversion microscope system (OptiquePeter, France) were used in this experiment, and various resolutions were acquired by changing the microscope objective. We employed 3.25 μm resolution to image the macrovasculature in the intact liver lobes (the two smallest lobes of the liver which could be imaged completely in the present imaging field of view) and 0.65 μm high resolution for detailed scanning of vessel termination and capillary architecture. Dark field (a dark signal with no photons) and flat field (no sample in the beam) images were taken for dark current corrections during detector imaging. Detailed information on the experimental setup, imaging procedure, and parameters used during image acquisition at two different resolutions is described in Supplementary Appendix E5.

CT reconstruction and 3D visualization

After the dark-field and flat-field images were used to rectify the raw projection, the phase-attenuation duality Born-type approximation (PAD-BA) algorithm was used to extract the phase information that was generated during the scanning of the subject from the light intensity information. A series of CT slices were subsequently reconstructed by using filtered back-projection algorithms, with images retaining only phase information. Reconstructed CT stacks were imported into Amira software (version 2019.1; Thermo Fisher Scientific, USA) for 3D visualization as well as for further image processing. The hepatic vein, portal vein, and hepatic sinusoid were semiautomatically segmented, as their gray value ranges differed from those of the parenchyma on PCCT images. The distinction between PV and HV is that PV is always accompanied by a hepatic artery or bile duct.

Image analysis

In combination with PCCT and 3D reconstruction, 3D virtual histopathological information on the liver vasculature can be obtained in any direction. To analyze the changes within the vasculature in detail, we divided the vasculature into three parts based on the pathway of blood circulation: the macrovascular tree, the macro-microvascular junctions in the hepatic lobules and the hepatic sinusoids in the primary lobules.

Macrovascular tree- The number of small blood vessels and Murray’s deviation at the intravascular branches and terminal branches of vessels were used to assess adaptive changes in the macrovasculature. We applied the generation numbers to the branches based on the method used by Jiang et al.³². The terminal vasculature was ranked as the first generation (generation 1), and the ranking number of the branches increased with increasing branch level. The total number of terminal vessels in the segment of the vessels with a diameter less than 100 μm located at generation five was used to define the number of small vessels. The terminal branches of a vessel indicated that neither of the bifurcating vessels generated a new branch. Murray’s deviation is a crucial indicator for evaluating the circulatory state of vascular branching. It is defined as the ratio of the cube of the radius of the parent vessel to the sum of the cubes of the radius of the daughter vessels. Fifteen vessel segments, thirty intravascular branches and thirty terminal branches of vessels in the PV and HV systems that met the requirements were chosen at random for the assessment of the number of small blood vessels and Murray’s deviation at the intravascular branches and terminal branches in each group.

Macro-microvascular junction- The parameters of the number of sinusoid-central vein connections and relative necrosis volume fraction were used to evaluate the adaptive changes occurring at the junction of macro-microvascular connections. Ten 3D regions of interest (ROIs) with a volume of 240 × 240 × 160 pixels around the central vein within the hepatic lobules were randomly selected for quantitative evaluation in each group. The number of sinusoid-central vein connections refers to the number of sinusoids in the ROI that can effectively transport blood into the central vein. Due to the fact that the gray value of the necrotic region was close to the liver parenchyma in PCCT, the volume of the necrotic region was mixed with the liver parenchyma, so that the necrotic volume fraction calculation in the ROI was performed by aggregating the volumes of both the necrotic tissue and liver parenchyma. Hence, in this analysis, the necrosis volume fraction is relative. For the normal group, the result was the liver parenchyma’s volume fraction within the ROI. The relative necrotic volume fraction was defined as:

$$V_n=\frac{V-V_a-V_s}{V-V_a},$$

the ROI volume was defined as $V$, the central vein volume was defined as $V_a$, the hepatic sinusoid volume was defined as $V_s$ and the relative volume fraction of the necrotic region was defined as $V_n$.

Hepatic sinusoids- 3D spatial density, relative surface area and mean diameter were used to assess the adaptive changes in hepatic sinusoids quantitatively. Based on the three-zone theory proposed by Matsumoto³³, the primary lobule was divided into three equipotential zones, and adaptive changes in the morphology of the liver sinusoids were observed in different zones. Seventeen primary hepatic lobules were chosen in each group, and ROIs of 60*60*60 pixels were randomly selected under different zones in the primary hepatic lobules for subsequent analysis. 3D density was the ratio of the voxels of the hepatic sinusoids to the total voxels in the ROI. The relative surface area was defined as the ratio of the liver sinusoid surface area to its volume in the ROI. The average of the mean diameter of liver sinusoids in each segment was used to represent the mean diameter of the ROI.

Histological analysis

After the scanning experiments, the imaged liver samples were embedded in paraffin, and 4-µm-thick slices were cut by using predetermined cross-sectional orientations. Hematoxylin-eosin (HE) staining was performed on all of the slides. The alignment of the pathology slides to the reconstructed CT was manually achieved to confirm the accuracy of the PCCT images.

Statistical and reproducibility

The statistical analyses were performed using SPSS (version 20; IBM, Chicago, USA). The normal distribution of the data was confirmed using the Shapiro-Wilk test. Numerical results were presented as mean ± standard deviation. For group comparisons, the one-way ANOVA followed by a post hoc Bonferroni test was used for continuous variables when the data satisfied the normal distribution and homogeneity of variance, such as the number of vessels, necrosis volume fraction and the connections between hepatic sinusoids and hepatic veins. The Kruskal–Wallis test followed by a post hoc Bonferroni correction was used when the data satisfied the normal distribution but not homogeneity of variance, such as the density, mean diameter and relative surface area of hepatic sinusoids. The two-way repeated measures ANOVA followed by a post hoc Bonferroni test was used to explore the effect of the time and the location (terminal or internal) of the vessel branches on Murray’s deviation. A P value < 0.05 was considered statistically significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

All authors contributed to the study conception and design. Yuanyuan Zhao, Wenjuan Lv, Chunhong Hu and Jianbo Jian contributed to the conception and design of the trial. Yuanyuan Zhao, Wenjuan Lv, Yi He, Beining Qi, Xianqin Du, Yuqing Zhao, Shan Shan, and Xinyan Zhao contributed to the acquisition, analysis and interpretation of data. Yuanyuan Zhao, Wenjuan Lv, Chunhong Hu and Jianbo Jian were responsible for writing, reviewing and editing. Chunhong Hu and Jianbo Jian were responsible for supervision and project administration. All authors discussed the results and approved the manuscript. All authors read and approved the final manuscript.

Peer review

Peer review information

Communications Biology thanks Samuel Huber, and the other anonymous reviewer for their contribution to the peer review of this work. Primary Handling Editor: Dario Ummarino.

Data availability

The numerical source data for figures and plots can be found in Supplementary data 1. All other data are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-024-07044-1.