Cardiac Magnetic Resonance Studies in a Large Animal Model that Simulates the Cardiac Abnormalities of Human Septic Shock

Background: Septic shock, in humans and in our well-established animal model, is associated with increases in biventricular end diastolic volume (EDV) and decreases in ejection fraction (EF). These abnormalities occur over 2 days and reverse within 10 days. Septic non-survivors do not develop an increase in EDV. The mechanism for this cardiac dysfunction and EDV differences is unknown. Methods: Purpose-bred beagles randomized to receive intrabronchial Staphylococcus aureus (n=27) or saline (n=6) were provided standard ICU care including sedation, mechanical ventilation, and fluid resuscitation to a pulmonary arterial occlusion pressure of over 10mmHg. No catecholamines were administered. Over 96h, cardiac magnetic resonance imaging, echocardiograms, and invasive hemodynamics were serially performed, and laboratory data was collected. Tissue was obtained at 66h from six septic animals. Results: From 0–96h after bacterial challenge, septic animals vs. controls had significantly increased left ventricular wall edema (6%) and wall thinning with loss of mass (15%) which was more pronounced at 48h in non-survivors than survivors. On histology, edema was located predominantly in myocytes, the interstitium, and endothelial cells. Edema was associated with significantly worse biventricular function (lower EFs), ventricular-arterial coupling, and circumferential strain. In septic animals, from 0–24h, the EDV decreased from baseline and, despite cardiac filling pressures being similar, decreased significantly more in non-survivors. From 24–48h, all septic animals had increases in biventricular chamber sizes. Survivors biventricular EDVs were significantly greater than baseline and in non-survivors, where biventricular EDVs were not different from baseline. Preload, afterload, or HR differences did not explain these differential serial changes in chamber size. Conclusion: Systolic and diastolic cardiac dysfunction during sepsis is associated with ventricular wall edema. Rather than differences in preload, afterload, or heart rate, structural alterations to the ventricular wall best account for the volume changes associated with outcome during sepsis. In non-survivors, from 0–24h, sepsis induces a more severe diastolic dysfunction, further decreasing chamber size. The loss of left ventricular mass with wall thinning in septic survivors may, in part explain, the EDV increases from 24–48h. However, these changes continued and even accelerated into the recovery phase consistent with a reparative process rather than ongoing injury.

were calculated.All measurements were calculated by an experienced technician and reviewed by a senior echocardiologist to avoid inter-observer bias.

Cardiac MR
Each animal was transported to the scanner sedated, mechanically ven�lated, and con�nuously monitored by a technician or clinician.A 3 Tesla MRI scanner (Philips Healthcare) acquired cardiac MRIs for animals at baseline (T0), 42 hours a�er challenge, and at the end of the study (96h).Electrocardiogram-gated steady state free-precision cine and T2 images were acquired in mid-ventricular short axis and assessed for average plane values.Epicardial and endocardial contours were drawn on the short-axis slices at end-diastole and end-systole.For contrast enhancement, animals were given gadobutrol 0.1 mmol/kg (Bayer Healthcare).Throughout the scan, animals were sedated, ven�lated, and con�nuously monitored by technicians and or clinicians.
All measurements for the CMRs were conducted using dedicated analysis so�ware (NEOSOFT suiteHEART) by appropriately trained clinicians and further confirmed by an experienced independent clinician.Papillary muscles were included in the volumetric quan�fica�on of the le� ventricle.

Electron Microscopy and Histology
Histology and electron microscopic analysis were performed as previously described. 13Briefly, the le� ventricle anterior wall �ssues were fixed and embedded in paraffin.Sec�ons were then stained with HE and Masson trichrome.The slides were scanned with NDP digital slide scanner (Hamamatsu) for histology evalua�ons.For Electron Microscopy (EM), the cardiac �ssues were dissected from the le� ventricle anterior por�on and brief fixed with 4% formaldehyde and 1% glutaraldehyde then sliced to 200 m slices with vibratome, further fixed overnight, followed by post fixa�on with 1% OsO 4 and embedded in Epoxy.Ultrathin sec�ons were stained with lead and uranyl and images were taken from JEOL-1400 electron microscope.

Animal Inclusion Criteria
Animals were pooled from two original experiments, the first assessed the effects of catecholamines on myocardial func�on during sepsis (n = 14) and non sep�c controls (n = 12), and the second assessed a bacterial dose response on the myocardial func�on in sepsis (n = 14).
Of these total 40 animals, 27 animals did not receive catecholamines and were included in a secondary analysis in this manuscript.Six addi�onal sep�c animals were employed for cardiac histology and EM.

Sta�s�cal Methods
Using CMR measurements, we calculated approximate "water content" as mass*edema (in %), and "dry weight" as mass -water content.Unless noted otherwise, data was analyzed using linear mixed models to account for repeated measures and ploted as model es�mate (+/-SE).We first tested the group-�me interac�on.If the interac�on term is significant, groups are compared at each �me point; otherwise, group comparisons are based on the main effects.
Standard residual diagnos�cs were used to check model assump�ons.All p-values are two-sided and considered significant if p ≤ 0.05.For some variables, logarithm transforma�on was used when necessary.Random-effect models were used to es�mate the overall mean difference (MD) or standardized mean difference (SMD) between groups.Heterogeneity among studies was assessed using the Q sta�s�c and I 2 value.Meta-analyses were conducted using R (version 4.2.2) and packages meta (version 6.2-0) 38 and metaphor (version 3.8-1) 39 .All other sta�s�cal analysis was performed using SAS version 9.4 (Cary, NC) and figure crea�on using GraphPad Prism 9.

Sepsis vs. controls
In sep�c animals at 4, 6, 8, 12, 16, 20 and 24h  Non-sep�c controls during this �me had no significant changes from baseline in mean IL-6 plasma levels.Compared to controls, the mean IL-6 levels in sep�c animals were significantly elevated at 8, 12, 16, 20 and 24h a�er bacterial challenge (Panel A).The mean IL-8 levels at 6 and 8h a�er bacterial challenge (Panel B) and the mean IL-12 levels at 72 and 96h a�er bacterial challenge (Panel C) were significantly elevated compared to baseline.During this �me, there were some significant decreases in mean IL-8 at later �me points and no significant changes in mean IL-12 in controls throughout compared to baseline.Furthermore, there were no significant differences in the mean values of these two cytokines in controls compared to sep�c animals.In sep�c animal, mean IL-10 levels at 16 and 20 h (Panel D) were elevated and mean Interferon-γ levels (Panel E) at 6, 8, 12, 20 and 48h were depressed compared to non-sep�c controls.Addi�onally, sep�c animals did have significant depressions compared to baseline in the mean levels of Von Willebrand factor (VWF) (8 and 96h, Panel F), TNF-α (8,12,16,20, 48,72    and 96h, Panel G), and there were not significant differences during these �mes in mean values of the following cytokines (VWF, TNF-α and p-selec�n) compared to controls.In sep�c animals at 4,6, and 8h a�er bacterial challenge, there were significant eleva�ons of mean levels of Monocyte Chemoatractant Protein-1 compared to baseline (Panel I).In controls, there was no significant change in the mean value of this cytokine during this �me compared to baseline.Thus, the major cytokine elevated most profoundly in this sepsis model was IL-6; however, mul�ple other cytokines showed changes from baseline with either eleva�ons or depressions that was not observed in controls.

Septic Survivors vs non-Survivors
There were no significant differences in the mean changes from baseline at 48h for any of the 9 cytokines measured (e-supplementary Figure 5 Panels A-I) between sep�c survivors and nonsurvivors.Of note, during this �me 5 cytokines were nominally higher in survivors (IL-6, -10, Interferon-γ, von Willebrand factor, TNF-α).Two cytokines (IL-and -12) were nominally only minimally elevated over �me in survivors and two of the nine Cytokines were remarkably similar over this �me (p-selec�n and Monocyte Chemoatractant Protein-1) in survivors and non-survivors.Overall, there were no remarkable increases in cytokine levels in non-survivors vs. survivors in this experimental sepsis model and if anything, survivors overall had higher cytokine levels not lower.

Sepsis vs. controls
In sep�c animals, from 4 to 96 h a�er bacterial challenge, mean pH was significantly decreased from baseline at all �me points measured.In non-sep�c controls, mean pH is only significantly decreased from baseline at 24 to 72h a�er bacterial challenge.However, there is overall from baseline to 96h a significant decrease in mean pH comparing sep�c and control animals (esupplementary Figure 6, panel A all, p=ns).There were no significant changes from baseline in mean PCO2 in both sep�c and control animals throughout the study, as this parameter was maintained within a set range by protocolized mechanical ven�la�on (Panel B).Arterial mean PaO2 in sep�c animals was significantly lower vs. controls at 8 and 12h a�er bacterial challenge and not significantly different from controls at all other �me points measured.This parameter was also ar�ficially maintained by protocolized �tra�on of inspired FiO2 on the mechanical ven�lator (Panel C).Mean lactate levels were not significantly different in sep�c animals compared to controls throughout the dura�on of the study (Panel D).
There were no significant differences in change from baseline for mean Crea�nine, BUN, Total Protein, Alanine Transaminase, and Albumin levels in sep�c vs. control animals throughout the study (e-supplementary Figure 7).In sep�c animas there were, at random �me points, isolated significant differences compared to baseline (at levels of no clinical significance) in mean BUN (Panel B), BUN/Cr ra�o (Panel C), Total protein (Panel D) and Albumin (Panel F).
Sep�c animals had significant increases in mean glucose levels at 12, 16, 20, 24, 48, 72, and 96h    compared to baseline.Mean glucose levels were normal in controls were significantly higher in sep�c animals throughout the study (e-supplementary Figure 8, Panel A).There were significant increases compared to baseline in mean potassium levels at 24, 48, 72, and 96h only but no significant difference from controls throughout.There were no other significant differences between sep�c and control animals in mean sodium (Panel C), and Chloride (Panel D) levels.
Sep�c animals had significantly lower mean WBC counts at 8, 12, 16, and 20h compared to controls (e-supplementary Figure 9, Panel A).There were no significant differences in mean Hemoglobin, Hematocrit, Lymphocyte, Eosinophil, and Platelet levels between sep�c animals and non-sep�c controls throughout the study (Panel B-F).

Survivors vs. Non-survivors
In survivors vs. non-survivors, there were no significant differences in mean pH (Panel A), PCO2 (Panel B), PaO2 (Panel C) and Lactate (Panel D) throughout the study (e-supplementary Figure 10).Non-survivors had significant increases compared to baseline in mean crea�nine levels at 16, 20, and 24h.Mean crea�nine levels were normal in survivors throughout the study but significantly lower compared to sep�c non-survivors (e-supplementary Figure 11, Panel A) at 16, 20, and 24h a�er bacterial challenge.In non-survivors, BUN levels were significantly elevated both compared to survivors (Panel B) and from baseline at 16, 20, 24, and 48h.There were no significant differences in mean BUN/Cr ra�o, Total Protein, Alanine Transferase, and Albumin between survivors and non-survivors throughout the study (Panel C-F).
Mean glucose levels were not significantly different in survivors vs. non-survivors throughout the study (e-supplementary Figure 12, Panel A).Potassium levels were significantly increased from baseline in non-survivors at 24 and 48h (Panel B).Sodium (0 to 48h) and Potassium levels (0 to 48h) were significantly higher in non-survivors compared to survivors (Panel B and C, p = 0.03 and 0.02 respec�vely).There were no significant differences in chloride levels between survivors or non-survivors and no significant changes from baseline throughout the study (Panel D).
There were no significant differences in WBC, Hemoglobin, Hematocrit, Lymphocyte, Eosinophil, and Platelet levels between survivors and non-survivors of sepsis throughout this study (e-supplementary Figure 13, Panel A-F).

e supplementary Literature Search for Meta-analysis
A search of electronic databases PUBMED, EMBASE, and BIOSIS was performed to iden�fy studies from 1980 un�l 02/22/2023.Any study inves�ga�ng LVEF and EDV in sep�c survivors and non-survivors through ECHO, MUGA or CMR was included.Relevant �tles were also iden�fied by hand searching previous reviews on the topic.Duplicates were filtered through EndNote once the databases were combined.Studies that were excluded were pediatric popula�ons, conference abstracts, reviews or editorials, non-sepsis related cardiomyopathies, or did not use the imaging modali�es listed above (e-supplementary figure 14).Search terms that were used included cardiomyopathy OR "contrac�le dysfunc�on" OR "systolic failure" OR "diastolic failure" OR "impaired cardiac relaxa�on.These terms were combined with "ejec�on frac�on" OR "diastolic volume" OR "diastolic volume index" OR 'echo' OR 'echocardiography', as well as the results from the combined set, 'sepsis' OR 'sep�c shock'.Results were screen further for imaging using, 'mri' OR 'magne�c resonance imaging'; 'rnca' OR 'radionuclide cineangiogra''; 't2 weight' OR 't2wi' AND edema./exp OR 'sep�c shock'.
a�er bacterial challenge, compared to baseline, there is a significant increase in mean IL-6 plasma levels (e-supplementary Figure4, Panel A).
e-supplementary Figure 3: For Panel A and B, the format is similar to Figures 5 Panel A. For Panel C, the format is similar to Figure 1.For Panel D, the format is similar to Figure 2. e-supplementary figure 4: The format is similar to Figure 1.