Acute kidney injury (AKI) is a major complication of sepsis and is associated with increased mortality 1. Renal vascular dysfunction has emerged as a major component of AKI 234. First, on a 'macrovascular' level, renal response to vasoconstrictor or vasodilator substances at the arterial and arteriolar level is altered in the presence of sepsis toward an increased vasopressive response 5. Second, on a 'microvascular' level, several recent lines of evidence have shown that abnormalities in cortical peritubular capillary flow occur very early in sepsis 678.

In animal models of sepsis, renal blood flow (RBF) is preserved provided that cardiac output is maintained with vascular fluid loading 2. Nonetheless, the abovementioned alterations in vasoreactivity may render the kidneys unable to protect themselves against additional ischemic insults during changes in systemic perfusion pressure: for instance, mean arterial pressure (MAP) variations may result in a rapid drop in RBF and/or glomerular filtration rate (GFR). Indeed, suboptimal perfusion pressure is common in patients with septic shock and has been shown to be associated with renal dysfunction 9. Most animal studies on RBF in sepsis have used either no fluid loading leading to profound hemodynamic abnormalities or massive fluid resuscitation leading to 'overcorrection' of cardiac output and arterial pressure, such that the impact of variation of blood pressure over a range relevant to the human setting remains unknown 2.

Aside from RBF, the relationship between systemic hemodynamics and microcirculation remains a matter of debate. In experiments performed in various organs, the link between systemic hemodynamics and microcirculation has been relatively vague although the latter was found to be affected by cardiac output and arterial pressure when these were critically altered 10. Most observations in kidneys, showing a microcirculatory defect in sepsis conditions, have been conducted in non-resuscitated animals such that cardiac output, and hence RBF, may have been substantially diminished 78.

As a result, the manner in which frequent MAP variations in sepsis impact both RBF and microcirculation is unknown. Moreover, septic shock patients require vasoconstrictor infusion, mostly norepinephrine (NE), in addition to fluid loading, to maintain blood pressure which may add another level of complexity to the vascular abnormalities observed in sepsis 11.

Thus, in order to explore the relationships between variations in MAP over a clinically relevant range and renal macro- and cortical microcirculations, we conducted a study on a rat model of sepsis with and without NE infusion. Renal cortical peritubular capillary blood flow was monitored using Sidestream Dark Field (SDF) imaging 12. Our main objectives were: (1) to assess the existence of an autoregulation phenomenon between MAP and RBF in sepsis and whether it is impacted by NE; and (2) to evaluate if changes in MAP impact cortical microcirculation in a manner similar to RBF in septic condition, in presence/absence of NE.

The present animal study was aimed at delineating the 'macro' and 'micro' hemodynamic changes over a broad range of mean arterial pressures in septic condition with or without NE infusion. The original design of our study with controlled variations in MAP allowed entering MAP, NE and sepsis as independent variables and assess RBF or CBFand cortical peritubular capillary blood velocity as dependent variables. Mixed models are particularly suitable for repeated measures and for modeling correlations between measurements and were hence used in the present setting. As sepsis is characterized by altered variability of clinical and biological signals, assessing variability of measurements was considered 1617. Consequently, we paid special attention to display all collected data in figures.

As expected, positive correlations were found between MAP and RBF as well as between MAP and CBF: namely, the higher the MAP, the higher the blood flow in these vessels. Sepsis did not appear to have an independent impact on RBF; conversely, NE decreased RBF. Indeed, in the present model, for any given MAP, RBF remained similar in the presence or absence of sepsis, but was lower in the presence of NE, thus providing evidence that NE displays a renal vasoconstrictive effect in this rat model. Of further importance, the relationship between MAP and RBF was weaker at MAP levels >100 mmHg in comparison to <100 mmHg. This is in accordance with an autoregulation phenomenon, although we did not observe a true RBF 'plateau'. More specifically, the variation in RBF when MAP varied from 100 to 130 mmHg did not differ in the presence or absence of sepsis or of NE, which may argue for the existence of this autoregulation phenomenon irrespective of the presence of sepsis or NE. Some of the observations pertaining to the kidney were organ specific since sepsis, NE and the MAP 100 mmHg threshold impacted CBF differently: both sepsis and NE increased CBF while the relationship between MAP and CBF was increased above 100 mmHg. Finally, a positive relationship was observed between MAP and renal cortical peritubular blood velocity (Vm) in that the higher the MAP, the higher the Vm. For a given MAP, sepsis increased Vm while NE decreased this velocity, such that NE tended to 'antagonize' the effect of sepsis on Vm. Finally, sepsis was associated with an increase in serum creatinine, which was prevented by NE.

RBF in sepsis has long been a subject of debate given that some models displayed decreased RBF while others showed either normal or elevated RBF. Langenberg et al. reconciled these conflicting data by demonstrating that the main parameter that predicted RBF in the various published animal models was the maintenance of cardiac output through fluid loading 2. Herein, we confirm these findings, as previously observed, after restoring normal MAP through fluid loading, RBF did not differ in the presence/absence of sepsis. Furthermore, we extend these data by showing that the relationship between MAP and RBF is not altered and that the relative RBF 'plateau' observed above a MAP of 100 mmHg, usually referred to as 'autoregulation' phenomenon, appeared to be preserved in sepsis or in the case of NE infusion 15. This last observation deserves to be emphasized since 'autoregulation' alteration is frequently cited as a contributing factor to renal failure 18. The ability of the kidney to maintain RBF when MAP varies over a certain range is enabled both by the myogenic reflex and the tubuloglomerular feedback and is impacted by neuroendocrine kidney vasomotor regulation 19. These effectors are dependent upon the integrity of several renal structures (that is, vascular, tubular) working in cooperation. It should be acknowledged that we only assessed the ensuing few hours after induction of sepsis, likely before renal damage may have sufficiently evolved to alter these structures. Indeed, several experimental data have shown that renal hemodynamics may be totally different at the initiation, maintenance, and recovery phases of AKI 1620.

Cardiac output was not monitored in the present study. Inserting a Doppler probe around the thoracic aorta is very invasive in rats and we felt it would have added an unwanted form of insult to our rat model. Indeed, our model is very complex and sensitive to any supplemental form of aggression, especially in septic animals, and thus we had to carefully put into balance the relevance of any additional procedure and the risk that such procedure may destabilize the entire model. It should also be emphasized that the present study was essentially 'kidney centered' with the main objective to assess the existence or not of an autoregulation phenomenon, which is evaluated by the relationship between MAP and RBF. Thus, we chose to monitor RBF at 'controlled' levels of MAP, and not cardiac output, in order to provide a direct evaluation of this relationship. We believe that not measuring cardiac output could have been a problem in interpreting RBF had we observed a reduced RBF. Conversely, the absence of impact of sepsis on RBF indirectly indicates that cardiac output was not compromised, at least significantly, in our resuscitated model. In fact, absence of cardiac output measurements prevented us from determining whether renal and/or systemic resistances varied in a similar manner or not, that is, whether renal behavior is different from other vascular beds. A partial response to this answer was to monitor carotid blood flow, which is easy to access. Indeed, we observed that RBF and CBF did not vary identically.

Deciphering the direct renal effect of NE infusion has been especially difficult as this drug also impacts systemic hemodynamics, which in turn modifies renal circulation 18. Indeed, most models showed that NE infusion was associated with an increased RBF together with an increased MAP and cardiac output 21. The particular design of our study allowed to 'offset' the systemic effect of NE and to observe a direct vasoconstrictive effect of NE on renal vasculature. Influence of sepsis on the effect of NE on RBF adds another level of complexity to the already discrepant results reported in the literature. Some authors evidenced that the increase in RBF after NE infusion in septic animals was not completely explained by an increase in perfusion pressure indicating that, in this condition, NE may actually decrease renal vascular resistance 22. Other studies showed a preservation of the pressor response to norepinephrine in the kidney, albeit with a diminished response in other vascular beds 5. In the present study, we did not observe any modification of the renal response to NE incurred by sepsis. However, we should acknowledge the limitation of our model due to sample size, such that a small impact of sepsis on the effect of NE on renal vasculature cannot be ruled out.

Few data have been published to date on renal microcirculation changes in sepsis. To our knowledge, no report has used SDF imaging to assess kidney microcirculation until now. We observed that this technique allows an easy and direct observation of blood cell velocities in peritubular capillaries. In capillaries (blood vessel diameter <10 μm), blood cell velocity can be considered as blood velocity, such that we believe the parameter measured herein is a relevant index of renal cortical microperfusion 13. It was observed that sepsis induces cortical peritubular blood cell hypervelocity, which was corrected by NE. Also noteworthy is that sepsis seems to impact cortical peritubular microcirculation and RBF differently in sepsis. Studies using intravital microscopy have previously been conducted on non-reanimated mice and rats by the group of Mayeux et al. 7823. These experiments showed a rapid decrease in the proportion of peritubular capillaries with continuous flow after induction of sepsis, which we did not observe in this study (data not shown). However, absence of fluid loading following sepsis induction in these particular experiments is likely to have resulted in a drastic drop in MAP, cardiac output, and RBF. Several publications have described the variation in cortical and sub-cortical oxygenation in septic condition in reanimated small animals using dual-wavelength oxygen phosphorimetry 62425. In these models, cortical oxygenation parameters were altered in septic animals and were only partially corrected by volume expansion allowing the normalization of RBF 24. However, using oxygenation as a proxy for microcirculation may be debatable. Indeed, a recent study reported normal kidney oxygenation measured by blood oxygen level-dependent MRI in LPS-treated mice despite diminished RBF 26. In this latter study, the authors evidenced profound mitochondrial alterations in the septic kidney, which could result in impaired oxygen utilization, such that oxygen tension may not be a good indicator of microvascular perfusion in sepsis.

Unfortunately, our findings do not allow to pinpoint the precise implications regarding acute kidney injury of peritubular capillary hypervelocity induced by sepsis and its correction by NE. Endothelial microvascular lesions have emerged as a component of kidney injury in several models, including sepsis or ischemia-reperfusion 42728. Peritubular capillary alterations have been the parameter most closely associated with renal failure among all vascular abnormalities described 28. Peritubular capillaries immediately follow glomerular circulation, which determines the filtration capacity of the kidney. One hypothesis could be that sepsis elicits a diminished vascular tone of the glomerulus efferent artery resulting in both a hypervelocity in the ensuing capillaries and a drop in glomerular filtration pressure. NE may correct this vascular tone thus restoring normal capillary velocity and filtration pressure. To date, however, exploring glomerular circulation to confirm this hypothesis in vivo has been beyond the reach of current technical capabilities. Of note, whether velocity of blood cells is an accurate index of blood flow is difficult to establish. In addition to velocity, the diameter of blood vessels and the number of blood vessels perfused should be taken into account. Although we did not observe variations in capillary sizes (that is, no vascular ectasia) or the appearance of non-perfused vessels (data not shown), the precision of these latter observations do not allow us to extrapolate accurate blood flow measurements from our dataset. Only cortical peritubular capillaries were evaluated, while deep cortical and medullary circulations were not assessed. Given the complexity of the renal architecture corresponding to different vascular organizations, it should be acknowledged that our experiments only target a very limited spectrum of the entire span of renal capillaries. Finally, variabilities in RBF and Vm measurements were lowered by sepsis, such as the model better predicted individual values for a given MAP. This is in keeping with previous observations of a loss of spontaneous heart rate variability in patients with multiple organ dysfunction or sepsis 17. A blunted autonomic function has been hypothesized in sepsis to explain these data, however the relevance of this altered variability for organ dysfunction remains unknown.

Several limitations of our study have already been discussed above. Notably, in this 'kidney-centered' study, parameters (including cardiac output, renal venous pressure, renal interstitial pressure, inflammation parameters, and so on) with potential interest in deciphering the relationship observed between MAP and RBF could not be measured. However, this study is nonetheless relevant with regard to one of the more prevailing questions in clinical practice for patients in shock conditions, namely the MAP target that should be reached through fluid loading and NE. We were not able to measure renal filtration during our MAP steps due to limited time spent at each step. Our evaluation of renal function with serum creatinine at the end of the experiments represents the sum of all insults that occurred to the kidney throughout each experiment. It is thus impossible to discriminate precisely at which time point changes in glomerular filtration occurred and eventually became permanent, which limits what can be inferred from creatinine measurements. At least, the observation that sepsis led to increased serum creatinine confirms the relevance of our model.

Our resuscitated rat model adds to the knowledge on renal macro-hemodynamics in sepsis in that the ability of the kidney to protect itself by maintaining RBF when MAP varies above a certain MAP threshold does not appear to be altered by the presence of sepsis or NE infusion. Notwithstanding, a direct vasoconstrictive effect of NE was evidenced. In contrast with the absence of any observable effect of sepsis on the relationship between MAP and RBF in resuscitated animals, from a microcirculatory standpoint, sepsis is very rapidly associated with 'hypervelocity' of blood flow in cortical peritubular capillaries, which is reversed by NE infusion. The mechanisms and consequences of this discrepant effect of sepsis on renal macro and microcirculations may be of great importance in understanding the failure of glomerular function in sepsis.

• In a rat model, early sepsis does not modify the relationship between renal blood flow and mean arterial pressure.

• Conversely, a vasoconstrictive effect of norepinephrine was observed.

• Sidestream Dark Field imaging enabled to observe blood hypervelocity in cortical peritubular capillaries in sepsis, which was improved by norepinephrine

• Thus, sepsis appears to differently impact kidney macro- and microcirculations

AKI: acute kidney injury; CBF: carotid blood flow; CT group: control group; CT+NE group: control + norepinephrine group; GFR: glomerular filtration rate; HR: heart rate; MAP: mean arterial pressure; NE: norepinephrine; RBF: renal blood flow; Sepsis+NE group: sepsis + norepinephrine group; SDF: Sidestream Dark Field; Vm : cortical microcirculatory velocity.

The authors declare that they have no competing interests.

MB participated in the design of the study, acquired the data, and drafted the manuscript; JFH participated in the design of the study and performed the statistical analysis; MT participated in the design of the study, acquired the data, and drafted the manuscript; MdB participated in the design of the study, acquired the data, and drafted the manuscript; AD participated in the design of the study, acquired the data, and drafted the manuscript; AM interpreted the data and drafted the manuscript; PC and AM interpreted the data and drafted the manuscript; PA and NL conceived and designed the study, interpreted the data and drafted the manuscript. All authors read and approved the final manuscript.