Perfusion and T2 Relaxation Time as Predictors of Severity and Outcome in Sepsis‐Associated Acute Kidney Injury: A Preclinical MRI Study

Preventing sepsis‐associated acute kidney injury (S‐AKI) can be challenging because it develops rapidly and is often asymptomatic. Probability assessment of disease progression for therapeutic follow‐up and outcome are important to intervene and prevent further damage.

S epsis is a life-threatening condition caused by a deleteri- ous host response to an infection. 1The necessary steps for sepsis management are rapid identification, antibiotic therapy, fluid administration, and often interventions with vasoactive drugs. 2 Despite improved clinical outcomes over the past decades, 3 sepsis still accounts for a large proportion of critically ill patients and remains a high mortality healthcare issue. 3Acute kidney injury (AKI) can be caused by various diseases and shows a two-way relationship to sepsis 4 ; roughly one in three septic patients will develop AKI and $40% of intensive care unit (ICU) patients 5 with AKI will develop sepsis.An estimate extrapolated from rates in the United States suggests that sepsis-associated acute kidney injury (S-AKI) causes up to 19 million incidents per year worldwide. 6he pathophysiology of S-AKI is considered to be distinct as compared to other etiologies of AKI. 7,8It is characterized biochemically by decreased glomerular filtration rate and/or reduced urine output despite fluid resuscitation, and can occur without overt hemodynamic instabilities. 9Currently, laboratory diagnosis relies on elevated serum creatinine (SCr) levels and/or oliguria. 4However, oliguria is nonspecific for S-AKI, and the continuous monitoring of urine output is also difficult outside the ICU. 4 In addition, SCr is an insensitive indicator of kidney injury, and is calculated ratiometrically to a baseline, which is a frequently missing value, upon patient admission.To date, there is no consensus about the unavailable information, 10 and therefore, new diagnostic biomarkers for the assessment of disease severity, and outcomes of S-AKI are needed.
Thus, the aim of this study was to find a sensitive noninvasive diagnostic imaging marker and outcome predictor for S-AKI in a rat-model of peritoneal contamination and infection (PCI) that has higher diagnostic power than those currently used in standard clinical practice.

Animal Experiments
All experiments were conducted in accordance with the National Institute of Health Guidelines for the Care and Use of Laboratory Animals (8th edition).Animals were housed under light-dark cycle 12/12 hours, temperature 22 C AE 2 C, humidity 50% AE 10%, pellet diet, water ad libitum.A total of 140 adult female Sprague Dawley rats (6 weeks old, average body weight 204 AE 30 g, RjHan:SD, Janvier Labs, France) were included in the study and divided into two different experimental groups: (A) healthy control group (n control = 65, with i.p. injection of $400 μL NaCl 0.9%; 1.8 μL/g body weight); (B) animals with polymicrobial sepsis (n sepsis = 75, PCI model, applied volume $400 μL NaCl 0.9%; 1.8 μL/g body weight).In addition, all animals were randomly assigned to two different examination strategies: (A) SCr group: determination of SCr levels 24 hours postsepsis/control induction (n sepsis = 35; n control = 31), and (B-1) MRI histology group: noninvasive MRI 18 hours postsepsis/control induction for histology assessment (16 animals: n sepsis = 9, n control = 7) and (B-2) MRI follow-up group: noninvasive MRI 18 hours postsepsis/control induction for disease follow-up with behavioral scoring during the first 96 hours (every 6 hours until 72 hours, then 96 hours, n sepsis = 33; n control = 25).The experimental paradigm is shown in Fig. 1.The current study is part of a greater project including MRI, mRNA sequencing, enzyme activities, mitochondrial capacity in both the acute and chronic status of S-AKI.Each readout requires a precalculated and defined number of animals approved by the local ethical committee.The authors seek every power to replace, reduce and refine the use of animals in research.

Peritoneal Contamination and Infection Model (PCI)
Systemic infection was induced by the PCI model. 11Briefly, a feces batch from healthy nonvegetarian human donors was processed to optimize bacterial content and inactivate reactive oxygen species.Feces samples were stored in aliquots at À80 C until further use.To induce polymicrobial sepsis, animals were anesthetized with isoflurane (2.5%); an aliquot of feces was diluted with saline (1:4 v/v), and 1.8 μL/g body weight was intraperitoneally administered into the right lower quadrant of the abdomen.In a pilot study, the amount of feces was adjusted to the current dose, resulting in a sepsis mortality rate of $50% 96 hours postsepsis induction.The renal functional impairment in the animal model was validated by assessing SCr levels 24 hours after septic insults (Fig. 3b).

Behavioral Scoring
After sepsis induction, animals underwent behavioral scoring by WTZ every 6 hours for up to 72 hours and daily thereafter.In addition to body weight loss (0-3), the score sheet 11 included behavioral criteria based on gait (0-3), locomotion (0-3), and responsiveness to external stimuli (0-3).The sum of all scores per time point described the degree of severity during the course of the disease.Depending on the scores, supportive animal care or an ethical endpoint was applied (Table S1).Scores 18 hours postinfection were correlated with the measured MRI parameters; scores 24 hours postinfection were correlated with SCr levels.To identify the differential severity and its relation to MRI parameters, we categorized the animals into mild (score 0-4) or severe disease symptoms (score >5) at 18 hours.To assess the discriminative performance of MRI, we retrospectively divided the outcome of the animals into groups of survivors and nonsurvivors at 96 hours and associated them to the MRI data collected at 18 hours.

Serum Examination
SCr was determined in healthy controls and septic animals (n total = 66; n sepsis = 35; n control = 31).Blood samples were collected 24 hours after sepsis/control induction.To separate the serum, the blood was first stored at room temperature for 30 minutes to induce clotting.Serum was then collected and stored at 4 C for a further 12 hours until precipitation.The supernatant was collected, transferred to a clean Eppendorf tube and stored at À80 C until further use.SCr levels were determined by SYNLAB Vet Laboratory.

Histology
Kidneys were snap-frozen in liquid nitrogen and stored at À80 C until use.After thawing, kidneys were fixed in 4% formaldehyde for 24 hours and dehydrated by immersion in a graded series of alcohol solutions before being embedded into paraffin.Three micrometer thick paraffin sections were cut for hematoxylin-eosin staining.Histologies (5-16 fields of view) were examined by two nephropathologists (BH and VvM) and one nephrologist (SR) for signs of acute renal injury such as focal tubular epithelial swelling, condensation or loss of tubular nuclei, and vacuolar degeneration.

Immunohistochemistry
First, formalin-fixed, paraffin-embedded sections (3 μm thick) were deparaffinized with xylene, rehydrated in a series of graded alcohols, and washed in distilled water.Then, CD68 immunohistochemistry was performed using a primary rabbit monoclonal anti-CD68 antibody (ab283654; Abcam, Cambridge, UK, dilution 1:500) using an automated staining system (Ventana BenchMark Ultra, Roche).Slides were pretreated with cell conditioning solution (Ventana Medical Systems) for 16 minutes at 95 C. Immunohistochemistries (5-10 fields of view) were examined by two nephropathologists (BH and VvM) and one nephrologist (SR) for CD68 positive cells.CD68 positivity was confirmed visually by comparison with the positive control (splenic staining) and the negative control (no secondary antibody).

MRI Protocols
All in vivo studies (n total = 74; n sepsis = 40; n control = 34) were performed at 9.4 T on a BioSpec 94/20 small animal MRI system (Bruker BioSpin, Ettlingen, Germany) equipped with a 0.7 T/m gradient system (BGS-12, Bruker) and ParaVision software version 6.0.1.A rat body quadrature volume coil with an inner diameter of 72 mm (Bruker BioSpin, Ettlingen, Germany) was used for image acquisition.Each animal was placed in a rat cradle in prone position.Body temperature was measured with a rectal probe and maintained at 37 C AE 1 C by a feedback-controlled warm-water heating system (SA Instruments Inc., Stony Brook, NY, USA).Respiration was monitored with a pneumatic pillow sensor attached to the animal's lower chest and a pressure transducer (SA Instruments, Inc.).During MRI acquisition, respiratory rate was maintained at 40 AE 5 bpm (beats per minute) by carefully adjusting the isoflurane concentration throughout the measurement (induction 4%, maintenance 2.5%-3.5%,and 0.7/0.3air/O 2 mixture).Besides a nonselective first-order shim, an additional B 0 map was acquired for a localized, voxel-based, higher-order map-shim that covered both kidneys and yielded a water resonance line with full width at half maximum of 60-110 Hz.
A localizer sequence (fast imaging with steady state precession, repetition time [TR] = 4 msec, echo time [TE] = 1.5 msec) and a 3D variable flip angle RARE sequence (Rapid Acquisition with Relaxation Enhancement, TR = 775 msec, TE = 3.06 msec, RARE factor = 80) 12 were acquired for geometric planning.Animals underwent perfusion imaging, as well as T 1 and T 2 mapping at 18 hours postinfection/control induction.
PERFUSION IMAGING AND T 1 MAPPING.To quantify renal perfusion and extract T 1 relaxation time maps, a fat-saturated flowsensitive alternating inversion-recovery echo-planar imaging sequence (FAIR-EPI), an arterial spin labeling technique, was applied in a single axial plane through both kidneys with the following parameters: TE = 12.8 msec, matrix size = 96 Â 96, FoV = (60 Â 60) mm 2 , slice thickness = 2 mm, inversion time (TI) = 100-4500 msec, number of TIs = 16, average = 1, acquisition time (TA) = 6 minutes.A constant recovery time of 10 sec was applied to ensure full recovery of magnetization.The FAIR-EPI sequence acquired data for two T 1 maps, and used either global or slice selective inversion in an interleaved fashion.A small window ($5 msec) respiratory trigger was setup on the beginning of the respiratory plateau.A schematic illustration of the multi-TI, multidelay FAIR sequence is shown in Fig. S2.

Postprocessing
PERFUSION IMAGING AND T 1 MAPPING.MRI data were processed using an in-house developed MATLAB framework (MathWorks, Natick, MA, USA).To calculate T 1 relaxation time and perfusion maps, both global and selective FAIR-EPI image series were fitted with the following equation with TR representing the repetition time, S the measured signal, S 0 the signal intensity in the fully relaxed state, and α the inversion efficiency (α = 1 for a perfect inversion).The T 1 relaxation time map was calculated from the corresponding global image series using the nonlinear least square fit model for higher precision. 13The initial values of the fitting were S 0 = 50, α = 1, and T 1 = 1000 msec.To assess perfusion, a blood pool T 1 value of 2430 msec 14 and a bloodtissue partition coefficient (λ) of 0.8 15 were used in the following equation 16 : The unit of renal blood flow (RBF) was mL/100 g tissue/min.An example image dataset and fitting of data from a control and septic animal is presented in Fig. S3.
T 2 MAPPING.To map the T 2 relaxation time constant, the multiecho RARE data were fitted using the equation previously described, 17 where S represents the measured signal intensity, S 0 the signal at TE = 0 msec, and c defines an offset constant.The initial values of fitting were S 0 = 65, T 2 = 30 msec, and c = 9.
All regions of interest (ROI) were drawn by WTZ based on the first echo of the T 2 image series in the cortex and medulla, avoiding artifactual chemical shift areas induced by perirenal fat (see Fig. S4 for details).The T 2 image series was coregistered with the first echo as a template using the Flirt function in FSL. 18Exemplary images and T 2 fitting in control and septic animals are shown in Fig. S5.
Examples of representative MRI images from a control and septic animal are shown in Fig. 2.

Statistical Analysis
T 1 and T 2 relaxation times and perfusion were averaged within the defined cortex and medulla ROIs, yielding six values per animal.These ROI-based MRI results, along with each animal's physiological parameters (behavioral scores, pre-MRI blood glucose, and corresponding sepsis outcome) were then used for a multi-parametric correlation analysis performed in Python (using Pandas ver.1.3.4 19nd Numpy ver.1.21.3 20 ).
MRI parameters and physiological data were first tested with the K2 D'Agostino-Pearson normality test (using scipy ver.1.7.1 21 ) and either Spearman or Pearson correlation coefficients (r) were computed based on the data distribution (Table S2).Specifically, Pearson correlation was used to compare normally distributed numerical parameters, while Spearman correlation was used otherwise (i.e. to compare numerical and ordinal variables such as behavioral scores and survival or non-normal numerical parameters).Parameters in cortex and medulla were considered separately.Group analysis between control and sepsis groups as well as between survivors and nonsurvivors was performed using Mann-Whitney U test.
In addition, multivariable logistic regression was used to estimate the probability of survival at 96 hours from cortical perfusion and T 2 relaxation times measured 18 hours postsepsis induction (using scikit-learn ver.1.1.1. 22).A fitted model was used to compute a distribution of survival probability as Before fitting the logistic model, perfusion and T 2 values were standardized by subtracting the mean value and scaling the values to unit variance 22  Receiver operating characteristic (ROC) curve 22 was used to determine the best cutoff values for the logistic regression to discriminate the survival at 96 hours.To evaluate the accuracy, area under curve (AUC) was calculated.The Youden index, 23 which indicates the maximum potential effectiveness of a biomarker, was determined as J max = max t (sensitivity t (%) + specificity t (%) À 100), where t ranges over all measured values of the animals.A Youden index = 0 indicates that the test gives the same results with and without the disease. 24P < 0.05 is considered statistically significant.

Animal Model and Severity Score
The PCI model resulted in peritoneal sepsis with various behavioral characteristics in all animals, including decreased body weight, diarrhea, ocular discharge, piloerection, reduced locomotion, and reduced response to external stimuli (see Table S1 for score criteria).In PCI animals, nearly half (46%) of the septic animals survived 96 hours after initiation of disease (Fig. S1a).Similar to Gonnert et al., 11 all septic rats exhibited peritoneal inflammation with multiple adhesions and bowel edema at dissection (data not shown).Overall, the septic severity scores increased after 6-12 hours, peaked between 24 and 36 hours, decreased after 36 hours (Fig. S1b), and the phenotype lasted for nine days (Fig. S1b).

Standard Diagnostic Methods and Severity
Twenty-four hours after infection, SCr increased in the mildly ill animals (P = 0.051) and was significantly elevated in severely ill animals compared with healthy control animals (Fig. 3a,b: control animals: 34 AE 9 μmol/L, n = 31, mildly ill group: 41 AE 13 μmol/L, n = 18, and severely ill group: 70 AE 30 μmol/L, n = 16).The kidneys of infected animals showed histologic signs of damage, such as focal tubular epithelial swelling, condensation or loss of tubular nuclei, and vacuolar degeneration, corresponding to the morphological correlates of S-AKI.In addition, more CD68-positive cells (monocytes/ macrophages) were observed in kidneys from septic animals compared to kidneys from control animals (Fig. 3d).

Changes in T 1 , T 2 , and Perfusion with Respect to Severity and Outcome
In the MRI data, septic animals had significantly decreased cortical perfusion (330 AE 140 vs. 480 AE 80 mL/100 g tissue/ min), and significantly reduced cortical (37 AE 5 vs. 41 AE 4 msec) and medullary (45 AE 6 vs. 52 AE 7 msec) T 2 relaxation times compared with control animals (Fig. 4).In septic animals, there were no significant differences in the T 1 relaxation times in the cortex (P = 0.23, sepsis: 1770 AE 140 vs. control: 1820 AE 90 msec) or in the medulla (P = 0.43, sepsis: 2100 AE 200 vs. control: 2200 AE 120 msec) compared with healthy control animals.No significant correlation was found between T 1 relaxation time and disease severity (r = À0.39 and P corr = 0.32), blood glucose level (r = 0.243 and P corr = 1), or weight change (r = 0.12 and P corr = 1) among all animals (Fig. 5, Bonferroni correction applied).
Despite the fact that individual MRI parameters could be associated with the severity scores (Fig. 7), animals with the same renal perfusion and relaxation values may have a different outcome in S-AKI survival (Fig. 3c).
Overall, 33% of the animals from the mildly ill group (n = 9) and 60% from the severely ill group (n = 20) later became nonsurvivors.The mildly ill nonsurvivors revealed cortical perfusion values ranging from 260 to 520 mL/100 g tissue/min and a T 2 relaxation time ranging from 31 to 41 msec.

Correlation, Regression, and ROC Analysis
Septic animals with poor outcome showed significantly reduced cortical perfusion and reduced T 2 relaxation time in cortex and medulla (Fig. 6).In septic animals, cortical perfusion (Spearman r = À0.44) and T 2 relaxation time (Spearman r = À0.55)showed a higher correlation with severity than SCr (Spearman r = 0.33; Fig. 7a).Finally, multivariable logistic regression between cortical perfusion and T 2 relaxation time in a probability distribution demonstrated that animals with decreased renal perfusion and T 2 relaxation time values were less likely to survive sepsis (Fig. 7b).
ROC analysis showed high accuracy for discriminating between survivors and nonsurvivors of S-AKI animals at 18 hours with an AUC of 0.80.At the Youden threshold (J max ), sensitivity and specificity were 80% and 73%, respectively (J max = 52%, Fig. 7c).
As another pathophysiological measure, we examined ATII(Agtr1) mRNA expression in renal tissue 18 hours after the septic insult.ATII(Agtr1) plays the most critical role in the regulation of sodium and fluid in the kidney. 25Our preliminary results showed a significant reduction in Agtr1b in septic animals compared with healthy controls (Fig. S6b) and a markedly reduced trend in Agtr1a (P = 0.059, Fig. S6a).

Discussion
Although S-AKI is known for its complexity and high mortality rate, clinical diagnosis still relies on SCr levels and/or urine output, the use of which are limited by delayed sensitivity, as SCr can detect renal injury at earliest 24-36 hours after onset. 26Therefore, a meaningful tool reflecting sepsis severity and outcome without delay is important for early therapeutic planning in rapidly evolving S-AKI.Correlation coefficient (R) is indicated between positive (red, 1.0) and negative (blue, À1.0) values.Significant correlations were observed between cortical and medullary T 2 relaxation times (R = 0.72, P < 0.001, Spearman), blood glucose level (glc pre) measured prior to MRI examination and the severity score (R = À0.77,P < 0.001, Spearman) as well as between cortical and medullary T 1 relaxation times (R = 0.47, P < 0.001).Further correlations were identified for the severity score: versus T 2 relaxation time in cortex (R = À0.5, P < 0.001, Spearman), perfusion in the cortex (R = À0.57,P < 0.001, Spearman), and T 2 relaxation time in the medulla (R =À0.51,P < 0.001, Spearman).Pearson's correlation was used unless otherwise indicated.*: P < 0.05, **: P < 0.001, Bonferroni correction was applied.).In the cortex, both perfusion (a, P < 0.005) and T 2 relaxation time (c, P < 0.005) showed statistically significantly reduced values in the nonsurvivor group compared with healthy controls.T 2 relaxation times in survivors were also significantly decreased (e, P < 0.05).In the medulla, significantly reduced T 2 relaxation time were observed in both survivors (f, P < 0.05) and nonsurvivors (f, P < 0.005) compared to healthy control animals.
As S-AKI can develop without tissue hypoperfusion, measurement of perfusion alone is not sufficient to assess disease severity.Furthermore, endothelial injury resulting in peritubular interstitial edema may have a pivotal impact on perfusion.Therefore, we combined T 1 -, T 2 -, and perfusion-MRI in a multiparametric imaging approach and evaluated impaired vascular and interstitial macrocirculation in addition to structural changes in sepsis.
On behavioral assessment, the majority of PCI animals had a severity score of 0 in the first 12 hours (data not shown) and developed initial symptoms of disease after 18 hours.At this time point, our multiparametric MRI investigations revealed that animals with S-AKI had a significantly hypoperfused cortex and reduced T 2 relaxation times in cortex and medulla despite fluid resuscitation.The observed decrease in T 2 relaxation time was negatively correlated with the sepsis severity score but not with the change in renal perfusion.In fact, an increase in T 2 was expected with increased renal injury due to pronounced inflammation and local tissue edema formation. 27However, decreased T 2 relaxation time was observed in septic animals, reflecting decreased water content in renal tissue, despite increased infiltration of CD68-positive monocytes/macrophages, along with histological signs of renal damage.Indeed, the manifested complications of sepsis or peritonitis, such as vomiting and diarrhea, would result in water shift, salt loss, and exposure to acute tubular necrosis risk.
An mRNA expression profile of renal tissue assessed 18 hours after septic insult demonstrated a significant reduction in Agtr1b and a markedly reduced trend in Agtr1a.Given the role of angiotensin II (Agtr1) in the regulation of sodium and fluid in the kidney, 25 the observed reduction in T 2 relaxation time could potentially be connected to an impaired ATII system in sepsis. 28t has been suggested that medullary perfusion and hypoxia play a "triggering" role in the various metabolic cascades in S-AKI. 29However, no significant perfusion differences between nonsurvivors and survivors were observed in the current study.The mechanistic relationship between medullary hypoxia due to a redistribution of intrarenal perfusion and sepsis survival requires further investigation. 29ndividual MRI parameters like perfusion and T 2 relaxation time showed a stronger correlation with disease severity than SCr.ROC curve analysis showed that the combination of perfusion and T 2 relaxation time obtained at 18 hours was highly likely to correctly predict the outcome of S-AKI animals at 96 hours.Instead of the expected severely ill animals, it was the animals with mild symptoms at early stage (18 hours) that eventually evolved into nonsurvivors at 96 hours and attracted our attention.The combination of partially perfused kidney and relatively low T 2 relaxation time constants could be a warning sign in early stage S-AKI nonsurvival.These results also suggest that T 2 relaxation time and perfusion MRI are a potential early screening tool for preselection of supportive care in patients on admission and/or with unknown history.
In recent years, several translational animal models of S-AKI/sepsis have been criticized because they do not parallel human sepsis. 3For example, the lipopolysaccharide model is widely used to study sepsis pathology. 30However, it fails to represent the complexity of S-AKI and provides only a simplified and more intense immune response compared to human sepsis. 30In the cecal ligation and puncture model, septic mice have shown marked and rapid metabolic suppression that is not comparable to human sepsis. 31The current PCI model is a polymicrobial sepsis model with fluid resuscitation, 11 which has been reported to mimic cytokine profiles, multiorgan dysfunction, metabolic acidosis, and impaired microcirculation in a dose-dependent manner.A comparison of commonly used S-AKI animal models with their advantages and disadvantages is summarized in Table S3.
It is worth noting that despite a significant increase in SCr levels in PCI animals, there is a substantial overlap between the groups.Usually, the SCr levels for healthy rats are between 13 and 31 μmol/L. 32Our PCI animals showed 41 AE 13 μmol/L in the mildly ill group and 70 AE 30 μmol/L in the severely ill group.According to the criteria for diagnosing AKI, 26 an increase of 26.5 μmol/L of SCr is required to enter the AKI stage.Creatinine production relies predominantly on the amount of skeletal muscle, but sepsis causes catabolic muscle wasting. 3As a result, SCr as a biomarker for renal injury may be weakened by sepsis and show lower SCr accumulation.This circumstance may explain why a few animals in the severe group had relatively low SCr levels, leading to the observed overlap of the data.Accordingly, the reliability of SCr reflecting renal injury in S-AKI may be limited.
Our current study focused on rapid MRI techniques based on established protocols that can be easily translated to the clinic.The perfusion map has an intrinsic "free" T 1 map, whereas the T 2 map collects free induction decay signal and typically runs within minutes-both are fast and can be easily adapted as a future screening method.

Limitation
There are a few limitations in our study.First, not all animals developed S-AKI.It is possible that the animals died from multiorgan failure, heart arrest and endotoxemia before S-AKI presented. 30This is because in polymicrobial sepsis, a mild dose of feces suspension might not cause AKI, but a slightly higher dose might result in death without AKI. 30eing aware of this condition, we carefully stratified the linear dose-dependent concentration when validating the animal model in a pilot study.At the current dose, the majority of animals gradually show increasing severity scores at 18 hours postsepsis induction.When increasing the dose of feces suspension, the animals demonstrated a high mortality rate and started to show a bipolar severity distribution-either by sudden death or by appearing completely healthy throughout the whole observation time.This bipolar behavioral phenotype indicates septic shock conditions which should be avoided.Still, a current ground truth for renal injury would be an insensitive and commonly performed SCr analysis or an invasive tissue biopsy for histology.Second, PCI allows classification into multiple severities and survivors/nonsurvivors that can be correlated with independent parameters but sepsis severity scores are susceptible to interobserver variation.MRI parameters (eg perfusion, T 1 and T 2 ), in contrast, do not exhibit inter-observer variability and are suitable to describe the high heterogeneity of S-AKI.Other parameters such as blood oxygenation level-dependent MRI can provide information about oxygenation levels in renal tissue, 33,34 but their suitability to represent hypoxia and to stage kidney function is still controversial. 35Apparent diffusion coefficients derived from diffusion-weighted MRI, intravoxel incoherent motion, or diffusion tensor imaging 27 could also provide further information about intracellular/extracellular fluid and more accurately and clearly assess renal interstitial hydrostatic pressure and renal microstructure without contrast agent. 36

Conclusion
We studied an animal model of polymicrobial sepsis with various severities using multiparametric MRI, which reflects the characteristic of S-AKI.The systemically identified changes in renal tissue perfusion and water content (reflected in T 2 ) provide a matrix for probability representation of the course and outcome of S-AKI.

FIGURE 1 :
FIGURE 1: Experimental paradigm.Animals received intraperitoneally either a slurry suspension of human feces (PCIgroup) or normal saline (Ctrl-group) (1.8 μL/g body weight).(a) Animals underwent behavioral scoring every 6 hours and were sacrificed 24 hours postinjection for SCr collection.(b) MRI measurements were conducted with prior blood glucose examination 18 hours possepsis/control induction.Animals were scored every 6 hours until 72 hours and daily thereafter.Body weight and temperature were determined daily.PCI, peritoneal contamination and infection model.

FIGURE 2 :
FIGURE 2: Representative figures of perfusion MRI, T 1 , and T 2 relaxation time maps in healthy animals and S-AKI animals.ROIs of cortex (pink) and medulla (blue) were manually drawn.Perfusion MRI and T 1 relaxation map were acquired by FAIR-EPI.T 2 relaxation time map was obtained by using a multiecho RARE sequence.

FIGURE 3 :
FIGURE 3: Sepsis-associated renal injuries: renal perfusion and SCr compared to healthy controls (n control = 31; n sepsis_mild = 19, n sepsis_severe = 16).(a,b) Significantly increased SCr levels were detected in septic animals (P < 0.001), especially in severely ill animals.(c) Septic animals exhibited reduced cortical perfusion compared to healthy controls (P < 0.001).Septic animals with similar perfusion values may exhibit different outcomes in surviving sepsis.(d) Exemplary hematoxylin-eosin staining from a control and sepsis animal (d1-2).Eighteen hours after sepsis induction, kidneys with S-AKI showed focal tubular epithelial swelling, condensation, or loss of tubular nuclei compared to controls.Moreover, infiltration with CD68-positive cells (arrows) was more prevalent in septic kidneys (d-4) than in control kidneys (d-3).Scale bar: 100 μm.

FIGURE 4 :
FIGURE 4: Perfusion and T 1 and T 2 relaxation times measured in cortex (a,c,e) and medulla (b,d,f) 18 hours postsepsis induction (n control = 25, n sepsis = 33).Septic animals showed significantly reduced perfusion in (a) the cortex but not in (b) the medulla compared to healthy controls.(c,d) No significantly different T 1 relaxation times were observed in either cortex or medulla in septic and healthy animals.(e,f) Both cortex and medulla showed a significant reduction in T 2 relaxation times in septic animals compared to healthy control animals.*: P < 0.05, ***: P < 0.005.N.S., not significant.

FIGURE 7 :
FIGURE 7: MRI parameters depict sepsis severity and survival.(a) Individual MRI parameters such as cortical perfusion or T 2 relaxation time in cortex and medulla showed a higher correlation (r) with the sepsis severity score than SCr levels.(b) Distribution of survival probability over perfusion and T 2 relaxation time p SurvivaljT 2 ,Perfusion ð Þ estimated from multivariable logistic regression.Yellow dots: survivors, dark purple dots: nonsurvivors.(c) ROC analysis showed high selectivity (80%) and sensitivity (73%) (AUC = 0.803, Youden's index J max = 0.52, red dot) in predicting survival.

FIGURE 6 :
FIGURE 6: Perfusion and T 1 and T 2 relaxation times measured in cortex (a,c,e) and medulla (b,d,f) 18 hours postsepsis induction in the nonsurvivor and survivor groups (n control = 25, n nonsurvivor = 18, and n survivor = 15).In the cortex, both perfusion (a, P < 0.005) and T 2 relaxation time (c, P < 0.005) showed statistically significantly reduced values in the nonsurvivor group compared with healthy controls.T 2 relaxation times in survivors were also significantly decreased (e, P < 0.05).In the medulla, significantly reduced T 2 relaxation time were observed in both survivors (f, P < 0.05) and nonsurvivors (f, P < 0.005) compared to healthy control animals.