ABSTRACT
Background: Correction of acute metabolic acidosis using sodium bicarbonate is effective, but has been hypothesized to exacerbate intra-cellular acidosis causing cellular dysfunction. The effects of acidemia and bicarbonate therapy on the cellular components of the glomerular filtration barrier, crucial for the integrity of the renal filter, are as yet unknown. Controversy persists regarding the most appropriate method to assess acid-base status: the “Stewart approach” or the “Siggaard-Andersen approach” using the standard base excess (SBE).
Methods: Here we performed physiological studies in anesthetized Sprague-Dawley rats during severe metabolic acidosis (HCl iv 6 mmol kg-1) and following bicarbonate (2.5 mmol kg-1) administration. We assessed glomerular permeability using sieving coefficients of polydisperse fluorescein isothiocyanate (FITC)-Ficoll 70/400. Acid-base status was evaluated using SBE, standard bicarbonate, total CO2, the Stewart-Fencl strong ion difference (ΔSID = Na – Cl – 38) and a theoretical model of plasma and erythrocyte strong ion difference.
Results: Our data show that neither acidosis nor its correction with NaHCO3 altered glomerular permeability. We identified ΔSID as a strong estimator of plasma base excess (as assessed using the Van Slyke equation). In silico modeling indicates that changes in the strong ion difference in erythrocytes would explain their buffering effect by means of a shift of anions from the extracellular fluid.
Conclusion: These data demonstrate a remarkable tolerance of the glomerular filter to severe acute acidosis and bicarbonate therapy. Our results also cast light on the buffer mechanism in erythrocytes and the ability of different acid-base parameters to evaluate the extent of an acid-base disorder.
IMPORTANCE STATEMENT Metabolic acidosis is a frequent complication of acute kidney injury in critically ill patients and is associated with a high risk of mortality. Correction of acidosis using sodium bicarbonate is simple and effective, but could possibly induce intracellular acidosis causing cellular dysfunction. The effects of acidemia and subsequent bicarbonate treatment on the cellular components of the glomerular filtration barrier, crucial for the integrity of the renal filter, are unknown. We show that neither severe acidemia nor bicarbonate therapy appear to have negative effects on glomerular permeability. Our analysis also highlights the buffering effects of erythrocytes, which appear to be mediated by a shift of strong anions into the red cells, increasing the strong ion difference in the extracellular fluid.
INTRODUCTION
Metabolic acidosis is associated with poor outcome in critically ill patients with reported mortality rates exceeding 50% when the pH stays below 7.20 1. Bicarbonate therapy is a simple and effective treatment to correct acute acidemia and chronic buffer depletion. However, the safety and feasibility of buffer treatment in acute metabolic acidosis are to date little studied and remain controversial 2. Adverse effects of bicarbonate therapy are well-described in the literature and include hypokalemia 3, hypocalcemia 4, hypercapnia 5, hypernatremia and fluid overload 6. Buffer therapy will also, by mitigating acidemia, lead to a left-shift of the oxygen dissociation curve (reduced Bohr effect) with increased binding of oxygen to hemoglobin and reduced delivery to tissues, potentially contributing to an increase in tissue hypoxia. On the other hand, it is well-recognized that acute metabolic acidosis can have direct negative effects on a number of organ systems, especially in the form of reduced cardiac output and contractility, which mainly occurs at a pH < 7.10-7.20 7.
The method by which the extent of a metabolic acid-base disorder should be determined was the topic of the historical “Great Trans-Atlantic Acid-base Debate” between the North American “Boston school” (favoring bicarbonate) and the European “Copenhagen school” (favoring base excess) which also comprise the buffer effects of erythrocyte hemoglobin 8. New controversy emerged by the introduction of a “Canadian school” with the so-called “Stewart approach” described by the Canadian physiologist Stewart in the 1980ies 9. This method, which does not include erythrocyte buffers, is based on the balance between blood plasma buffer ions (usually conjugate base anions of weak acids such as carbonic acid) and strong ions, which do not act as buffers. Due to the complexity of the original Stewart model, a simplified methodology, the so called Fencl-Stewart approach has been proposed, introducing a simplified version of the SID = Na – Cl. The normal SID between strong ions in blood plasma is ∼38 meq/L (25 meq/L bicarbonate, 12 meq/L albumin and 1 meq/L phosphate) and thus the strong ion gap ΔSID = Na – Cl – 38 is an approximation of the amount of buffer ions in plasma.
Metabolic acidosis is an invariable feature of acute and chronic kidney disease and results mainly from reduced ammoniagenesis and impaired tubular reabsorption of bicarbonate. The observation that a drop in plasma pH could cause proteinuria is not new, and there are some reports 10, 11 showing that metabolic acidosis leads to increased proteinuria. In an early study, Gardner 10 provided data that indicated alterations in protein transport at the glomerular level during acidemia. In contrast, Throssell and colleagues 11 found significant low molecular weight proteinuria in acidotic rats, implying a tubular origin. Recent data reveal that the integrity of the glomerular filter is crucially determined by the concerted actions of the cellular components of the glomerulus 12-14, especially podocytes 15, which appear to tightly regulate glomerular permeability. Moreover, several in vitro and in vivo studies have indicated that both extra-cellular acidemia and bicarbonate therapy can exacerbate intra-cellular acidosis 6, 16, conceivably leading to cellular dysfunction. Thus, we hypothesized that both acidemia and its correction using bicarbonate, could affect glomerular permeability. Also, recent experiments indicate that the buffering effect of erythrocytes is facilitated chiefly via the Anion exchanger 1 (AE1) 17, making up to 25% of the red cell membrane surface, with no involvement of other transporters. On this basis, we developed a strong ion theory that includes the buffering effect of erythrocytes and show that it can explain the difference between ΔSID and SBE in our data. We also investigated the capability of several parameters to assess buffer deficit, since a number of different parameters for acid-base status are currently in clinical use.
CONCISE METHODS
Animals
Experimental studies were performed in 8 male Sprague-Dawley rats (Möllegard, Lille Stensved, Denmark) having an average body weight (BW) of 269 g (255-289 g) with free access to food and water. The local Animal Ethics Committee at Lund University, Sweden approved all procedures.
Experimental metabolic acidosis
All experiments commenced with a resting period of 20 min following surgery (Figure 1). The left carotid artery was cannulated and utilized for arterial blood samples for analysis of electrolytes, pCO2, urea, Hb, and pH using the I-STAT EC8+ cassette (I-STAT; Abbot Point of Care Inc, Abbot Park, IL). To prevent respiratory compensation of the metabolic acidosis, neuromuscular blockade was then induced with rocuronium bromide at a dose of 0.6 mg/kg (Esmeron, Merck Sharp & Dohme B.V., Haarlem, Netherlands) followed by a dose of 0.3 mg/kg every 40 min. The animal was carefully observed during the course of the experiment to ensure that no spontaneous breathing occurred. After an initial 5 min period for control (baseline) measurements, hydrochloric acid 6 mmol/kg (Saltsyra APL 1 mmol/mL, APL, Stockholm, Sweden) was given as an infusion during a period of 60 minutes starting directly after the baseline measurement period. After the infusion had stopped, new measurements were performed during a 5 min period after which sodium bicarbonate 2.5 mmol/kg (Natriumbikarbonat Fresenius Kabi 0.6 mmol/mL, Uppsala, Sweden) was given as an intravenous infusion during a period of 10 minutes.
Glomerular transport studies
During the entire course of the experimental procedure, a continuous iv infusion (10 ml kg-1 h-1) of FITC-Ficoll (FITC-Ficoll-70, 20 µg ml-1; FITC-Ficoll-400, 480 µg ml-1; FITC-Inulin, 500 µg ml-1 and 51Cr-EDTA, 0.3 MBq ml-1) was given after an initial bolus dose (FITC-Ficoll-70, 40 µg; FITC-Ficoll-400, 960 µg; FITC-Inulin 0.5 mg and 51Cr-EDTA 0.3 MBq). This 1:24 mixture of Ficoll provides a broad range of molecular sizes. Sieving measurements were performed by a 5 min collection of urine from the left ureter with a mid-point (2.5 min) plasma sample. Glomerular filtration rate was assessed using the clearance of 51Cr-EDTA. The distributed pore model by Deen, Bridges, Brenner and Myers 18, modified by using two pore size distributions 19, was used to analyze the θ data for FITC-Ficoll (mol. radius 1.5–8.0 nm) by means of weighted nonlinear least squares regression to obtain the best curve fit as described previously 20. The solution of the model differential equations and calculation of renal plasma flow are described in 20.
Physiological acid-base parameters
The normal values assumed in the current work are shown in Supplemental Table 1. The concentration of titratable hydrogen ions (also called base deficit) in whole blood (ctH+B) and plasma (ctH+P) was calculated using the Van Slyke equation (see Supplemental material). The standard base excess (SBE) is the base excess of the extracellular fluid 21 and was calculated as follows
Statistical methods
The data are presented as median ± median absolute deviation unless otherwise specified. We used an alpha level (probability of type I error) of 0.05 and a beta level (probability of type II error) of 0.20 unless otherwise specified. A previous power analysis showed that four was the minimal number of animals needed to detect a difference in the sieving coefficient of Ficoll70Å by a factor of at least 2 22. Significant differences were assessed using an ANOVA on aligned rank transformed data (ARTool version 0.10.5), essentially performing a non-parametric test using parametric methods 23. We applied the modification of the F-statistic by Kenward and Roger 24 for small sample sizes. Tukey’s Honest Significant Difference post-hoc tests were applied. Linear regression was performed using the lm function in R. Statistical analysis was performed using R version 3.5.1 for macOS (The R Foundation for Statistical Computing).
RESULTS
Severe acidemia induced by HCl and its correction using sodium bicarbonate does not alter glomerular permeability to Ficoll
Infusion of hydrochloric acid over the course of 60 minutes, 6 mmol/kg, evoked severe acidemia in Sprague-Dawley rats (n = 8), being partly improved by a 10 min iv infusion of sodium bicarbonate (0.6 mmol/mL) 2.5 mmol/kg (Figure 1 and Figure 2). The standard base excess 21, defined as the base excess (i.e. negative value denotes base deficit) of the extracellular fluid, decreased from 3 ± 1 mmol/L to −22 ± 2 mmol/L (Figure 2b). Assuming the extra-cellular volume (ECV) to be ∼30% of the body weight, this corresponds to a H+ dose of ∼20 mmol per L ECV, in concordance with the obtained result. Bicarbonate infusion (4.2 mL kg-1) apparently resulted in hemodilution (Figure 2c), corresponding to a ∼15% increase of the plasma volume, whereas there were no significant alterations in hematocrit or hemoglobin level in acidemia compared to baseline (Figure 2c). Mean arterial pressure decreased from 125 ± 6 mmHg to 115 ± 9 mmHg (P = 0.02) in acidemia (Figure 2d). The partial CO2 pressure (pCO2) increased in acidemia, presumably due to buffer consumption (Figure 2e). Conceivably, since ventilator settings were set constant over the course of the experiment, a similar increment in pCO2 should occur after administration of bicarbonate, but the difference was non-significant (P = 0.06). The impact of acidemia on glomerular filtration rate was heterogeneous with some animals developing severe hypofiltration (Figure 2f). Plotted in Figure 3 are the glomerular sieving curves (θ vs. Stokes-Einstein radius) at baseline, before, and after bicarbonate infusion. Statistical comparison (Figure 3: bottom plot) of the sieving coefficients revealed no significant alterations in glomerular permeability, neither before nor after bicarbonate treatment. Using the distributed two-pore model, we noted no significant alterations except a small increment in the large pore radius post-HCO3- (Supplemental Table 2). Bicarbonate therapy lead to sodium loading (Figure 4a) 6 and a potassium shift into the intra-cellular compartment (Figure 4b) 5, decreasing K+ from 4.0 ± 0.3 mmol/L in acidosis to 3.5 ± 0.3 mmol/L after sodium bicarbonate, corresponding to a decrease in K+ of 0.4 mmol/L per 0.1 unit increase in pH. Following HCl infusion, plasma chloride increased from 101 ± 2 mmol/L to 121 ± 1 mmol/L, in good agreement with the administered dose of HCl (vide infra), and decreased to 116 ± 2 mmol/L after buffer infusion (Figure 4c). Blood urea nitrogen levels remained stable throughout the course of the experiment (Figure 4d).
Several established methods to assess buffer deficit are highly correlated but result in different estimates of the buffer deficit
Many different methods have been proposed to assess metabolic acid-base disorders and we find that the capability of several methods (Standard bicarbonate, total CO2 et cetera) to detect the existence of a significant acid-base disturbance is very similar (Supplemental Fig. 1). However, to guide buffer therapy it is also of clinical importance that a method can correctly identify the actual base deficit. In this regard, the various methods differed (Table 1) when compared to a calculated buffer deficit from the HCl dose assuming an ECV of 288 mL/kg 25 and no metabolic compensation.
The Simplified ΔSID Systematically Underestimates Buffer Deficit Assessed by SBE
Shown in Figure 5a is the simplified ΔSID = Na – Cl – 38, which decreased from 2 ± 1 at baseline to −17 ± 2 following HCl infusion, and increased to −9 ± 1 after bicarbonate therapy. Very similar alterations occurred for the plasma base excess (ctH+P; Figure 5b) and linear least-squares regression revealed a very high degree of correlation (R2 = 0.974), with nearly a 1:1 agreement, between ΔSID and ctH+P (Figure 5c). Indeed, such a similitude is theoretically expected on the basis of the equality between true SID and plasma buffer base 26 and thus, it would appear from our data, that the simplified ΔSID is a strong predictor of the true change in SID. In contrast, inasmuch as the standard base excess (the amount of titrable acid in the extracellular fluid) and simplified ΔSID were highly correlated (R2 = 0.968), ΔSID systematically underestimated the buffer deficit assessed by standard base excess (SBE = 1.3543 × ΔSID + 0.1086, P < 0.001) (Figure 5d).
Changes in the Strong Ion Difference in Erythrocytes Mediates a Buffering Effect on the Extracellular Fluid Which Explains the Difference Between Simplified ΔSID and SBE
It is well-recognized that erythrocytes, mainly via their hemoglobin content, have a buffering effect on the extracellular fluid. The main difference between the above simplified ‘Fencl-Stewart’ ΔSID and the ‘Siggaard-Andersen’ SBE is that the former parameter does not take this buffer into account. Accordingly, in this study buffer deficit assessed with ΔSID was found to be consistently lower than SBE (Figure 5d). Thus, due to their buffer effect, erythrocytes would act to increase plasma strong ion difference (lowering ΔSID), presumably by lowering plasma chloride via the abundant and highly efficient AE1 (Band 3) HCO3-/Cl- exchanger, present on the cell surface of erythrocytes 17. To quantitate this effect, we developed a theoretical model for the strong ion difference, or more precisely, the amount of titratable acid in the extracellular fluid and in erythrocytes, based on the observation that the change in erythrocyte internal pH = 0.77 × external pH change (see Supplemental material) over a wide range of pH 17. The pH in erythrocytes is 7.25 at a normal blood plasma pH of 7.40 17, which, if taking the most important intracellular buffers (bicarbonate and hemoglobin tetramer 17) into account, corresponds to an erythrocyte SID of 38 mEq/L (per liter of cells; see Supplements). Our theoretical model is in a near 1:1 linear relationship with the simplified ΔSID (Figure 6a) if the effect of erythrocytes is excluded (i.e. taking only bicarbonate, albumin and phosphate buffers into account). In contrast, when effects of erythrocytes also are included, the model showed a near perfect agreement with SBE (Figure 6b). Thus, our in silico simulations predict that the effects of metabolic acidosis would strongly affect the SID of erythrocytes (Figure 7), giving rise to a buffering effect due to a shift of strong anions from plasma to erythrocytes, and increasing the SID of the extracellular fluid. This ‘anion shift’ should be much larger in vitro and, in a separate simple experiment, titration of whole blood with hydrochloric acid confirmed that a substantial amount of chloride indeed appears to be shifted into the erythrocytes (Supplemental Figure 2).
DISCUSSION
Emerging evidence unveils the crucial role of the cellular elements in the glomerular filtration barrier for the maintenance and regulation of glomerular permeability and function 13, 22, 27. Previous data from our group indicate that many disease mechanisms, including inflammatory stimuli 14, 28, angiotensin II 12, decreased NO bio-availability 13, ureteral obstruction 22 and hypoxia 29 lead to a glomerular hyperpermeability response. In light of these previous findings, it is surprising that no alterations in glomerular permeability were observed in the current study, with sieving coefficients for large Ficolls actually being numerically lower during severe acidemia.
An intracellular pH of 7.10-7.30 is essential for the normal function of cells 30 and it is commonly assumed that intracellular acidosis will occur whenever there is an extracellular acidosis and that it is the former that leads to the negative effects 16. There is however uncertainty and some controversy over whether the intra-cellular pH decrease occurs uniformly in all tissues and whether these alterations are sustained or transient and/or depend on the mechanism causing the acidosis 16. Indeed, other in vivo studies of acidosis due to HCl (or lactic acid) infusion have shown little or no effect on intra-cellular pH 31, while ischemic lactic acidosis caused marked effects on intra-cellular pH 32. Also, Shapiro et al 33 found that NaHCO3 treatment caused intracellular brain acidification in rats subjected to ammonium chloride acidosis. However, Bollaert et al 34 found no intra-cellular acidosis in the muscle tissue in septic rats with lactic acidosis after correction with NaHCO3 33. Similarly, in an elegant study, Levraut and colleagues 35 demonstrated that intra-cellular acidosis only occurred in vitro when using a non-bicarbonate buffering system. Albeit, from a mechanistic point of view, increasing the pH of both compartments should be beneficial in comparison with just increasing the extracellular pH. Thus, it may at first seem obvious that sodium bicarbonate therapy would be beneficial in the treatment of acidosis. However, the potential benefit of buffer therapy has been a matter of much debate. For example, nephrologists and critical care physicians had a clear disparity of opinion when polled about the use of buffer therapy in severe organic acidosis 36. While the beneficial effects of sodium bicarbonate on hemodynamics are robustly supported by experimental studies, its ability to improve hemodynamics and morbidity in critically ill patients is as yet unproven. In a very recent report 37, Jaber and colleagues randomized patients with pH ≤ 7.20 to either NaHCO3 to maintain a pH ≥ 7.30 or no treatment and found no difference in the primary outcome of mortality at day 28 or organ failure at day 7. However, in a predefined stratum of patients with acute kidney injury network (AKIN) scores of 2-3, mortality was less frequent in the bicarbonate group compared to controls. Furthermore, the need for renal replacement therapy was overall lower in the bicarbonate group 37.
A common indication for buffer therapy is hemodynamic instability in severe acidemia. In addition, the condition of critically ill patients may be further complicated by acidemia-induced hyporesponsiveness to vasopressors. It is also well known that acidemia affects the potency and risk of side effects of numerous drugs used in anesthesia and intensive care. For example, volatile anesthetics become more arrhythmogenic while opiates and neuromuscular blockers are potentiated. In the current study we found a decreased mean arterial pressure and heart rate during acidosis, but our data failed to show improvements following bicarbonate therapy. However, in mild to moderate acidemia, pH > 7.20, cardiac output may actually increase due to enhanced catecholamine release 6, but even mild acidemia has been associated with hemodynamic instability 6, 38. For example, Kellum and colleagues induced different degrees of hyperchloremic acidosis in septic rats and demonstrated a significantly lower arterial pressure in the animals with mild acidemia (SBE −5 to −10) compared with non-acidotic animals 38. However, neither controlled clinical studies 39, 40 nor experimental studies 41 in lactic acidosis showed any benefit on cardiovascular function following NaHCO3 administration.
Acute and chronic metabolic acidosis are common acid–base disorders and their diagnosis relies on the correct estimation of the buffer deficit, identification of respiratory compensation, and an assessment of the plasma anion gap, preferably corrected according to the plasma albumin concentration 21 so as to be able to identify the etiology of the disorder. Here we provide evidence of significant differences between the standard base excess compared to other methods. A priori, this result can be expected since none of the other methods include the buffering effect of erythrocytes. Furthermore, the current results highlight the difference between the quantification of buffer deficit and the assessment of ion mass balance, e.g. in the form of the anion gap. Mathematically, the albumin corrected anion gap is identical to using the simplified ΔSID and adding 1 mEq/L per every 4 g decrease in plasma albumin from a normal concentration of 42 g/L 21, 42. The clinical rationale for such computations is to detect organic, anion-gap acidosis 21. While the use of only sodium and chloride and bicarbonate/base excess to assess ion mass balance may seem simplistic compared to summing up all plasma cations and anions, one should bear in mind that including more measurements will inflate the coefficient of variation of the composite parameter and, thus, decrease precision. Indeed, our current data show that the simplified ΔSID = Na – Cl – 38 can predict the actual ΔSID with surprising accuracy.
We developed a strong ion theory similar to that by Stewart 9, which also included the buffering effect of erythrocytes. We show that this modification leads to the near equality of the buffer deficit approximated by SBE and that approximated with the novel model (Supplemental Equation 21). Our modelization relies heavily on the assumption that most of the buffering effect is mediated via the AE1 isoform of the anion-exchange transporter (also called Band 3), being highly expressed on erythrocytes with more than 106 copies per red cell 17. It cannot however be excluded that other transport proteins contribute. For example, the sodium concentration also increased in our in vitro experiment (see Supplemental Figure 2), which may imply a role for the sodium-hydrogen antiporter (NHE-1) in erythrocytes. Many other simplifying assumptions were made in our model. For example, we assumed that the only non-carbonate buffers are albumin and phosphate in plasma; and bicarbonate and hemoglobin in erythrocytes. We also used a simplistic equation for hemoglobin titration. Indeed, far more sophisticated models have been developed for this purpose 43. Nevertheless, despite its simplicity, we find that our model provides an excellent fit to our data. It also serves to convey a fundamental understanding, and simple quantitation, of the buffering effects of erythrocytes.
We conclude that neither acidosis nor its correction using sodium bicarbonate appear to have any negative effects on glomerular permeability. Furthermore, the striking agreement between the plasma base excess and the simplified ΔSID supports the use of the simplified Fencl-Stewart approach 42. Inasmuch as the difference between SBE and ΔSID was significant, and most likely caused by the fact that SBE also takes erythrocyte buffering into account, the differences were rather small from a clinical perspective, even in a very severe acid-base disturbance such as in the present study. Thus, both methods should provide reasonable results in clinical practice. We conclude that our strong ion modelization may provide part of the missing link between strong ion theory and the Danish base excess concept. It however deserves mentioning that, in the present study, these similarities were shown in Sprague-Dawley rats and not in patients, and clinical studies could be performed to confirm or refute the present results. Regardless of with what method one prefers to assess acid-base status, it appears that the standard base excess, robustly supported by several clinical studies 21, is a more suitable method to determine the extent of a metabolic acid–base disorder while the simplified Fencl-Stewart approach 42, or the albumin corrected anion gap 21, aids the clinician in identifying the mechanism.
AUTHOR CONTRIBUTIONS
J.S. and C.M.O. designed the study; A.R. and C.M.O. carried out experiments; C.M.O and J.S. analyzed the data; C.M.O. made the figures; C.M.O, P.B., D.P., N.C. and J.S. drafted and revised the paper; all authors approved the final version of the manuscript.
ACKNOWLEDGEMENTS
This study was funded by the medical faculty in Lund (ALF grant). The authors declare no competing interests pertaining to the current article.