Measurement of β-hydroxybutyrate in acute hyperketonemia in human brain


Address correspondence and reprint requests to J.W. Pan, Department of Neurology, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, New York, NY 10461, USA. E-mail:


We report the measurement of d-β-hydroxybutyrate (BHB) in the brains of six normal adult subjects during acute infusions of BHB. We used high field in vivo1H magnetic resonance (MR) spectroscopy in the occipital lobe in conjunction with an acute infusion protocol to elevate plasma BHB levels from overnight fasted levels (0.20 ± 0.10 mm) to a steady state value of 2.12 ± 0.30 mm. At this level of hyperketonemia, we determined a tissue BHB level of 0.24 ± 0.04 mm. No increases in brain lactate levels were seen in these data. The concentrations of BHB and lactate were both considerably lower in comparison with previous data acquired in fasted adult subjects. This suggests that up-regulation of the monocarboxylic acid transporter occurs with fasting.

Abbreviations used



cerebral blood flow


magnetic resonance


monocarboxylic acid transporter


permeability surface.

It is well known that the human brain is able to utilize ketone bodies under numerous physiological circumstances including fasting and high fat diets (Robinson and Williamson 1980; Balasse and Fery 1985). However, the time course and molecular mechanisms by which brain shifts between glucose and ketone utilization are not well defined in humans. We have previously shown that after 2 and 3 day fasts, where plasma d-β-hydroxybutyrate (BHB) levels reach approximately 1.7 and 3 mm, respectively, the level of BHB in the human brain also increases, reaching approximately 0.6 and 1 mm, respectively (Pan et al. 2000a). The steady state level of BHB necessarily reflects net influx and clearance, both of which may change with fasting. In rodents, there is extensive data showing that the monocarboxylic acid transporter (MCT) is induced with fasting. Gjedde and Crone (1975) studied pentobarbital-anesthetized rats to demonstrate a five-fold increase in the Km, and a six-fold increase in the Vmax/cerebral blood flow (CBF) of ketone transport with starvation. Similar results have been reported by Hawkins et al. (1986). However, comparatively less data are available in humans. Hasselbalch et al. (1995) studied non-fasted and fasted subjects by the double indicator dilution method, finding that fasting did not significantly result in a change in the permeability surface (PS) area coefficient for BHB. Assuming a relatively small diffusional component to BHB transport, the absence of a change in the PS = Vmax/(Km + Pkb) was interpreted as indirectly supporting the hypothesis of transporter induction. The substantial fasting induced rise in plasma ketone levels Pkb would thus be compensated by alterations in the PS coefficients to result in no change in its apparent value.

Induction of ketone transport can be assessed directly by measuring tissue BHB levels in the non-fasted hyperketonemic state. In this study we used in vivo1H magnetic resonance (MR) spectroscopy to measure the concentration of BHB in the brain of normal adult subjects (n = 6) acutely infused with BHB. We found that with an increase in plasma BHB concentration to 2.1 mm, the tissue concentration rose to 0.24 ± 0.04 mm from 0.16 ± 0.07 mm. This is substantially lower than that detected in 48 h fasted subjects where tissue ketone levels reach 0.60 ± 0.26 mm with plasma levels of 1.67 ± 0.34 mm. This supports the belief that fasting results in a metabolically significant induction of BHB transport activity.

Materials and methods

Magnetic resonance

A 4T Varian INOVA whole body MR system and 7 cm surface coil was used for all studies. The surface coil was placed adjacent to the occipital ridge, such that the volume of interest was located within the occipital lobe. Scout imaging using inversion recovery gradient echo images was used to localize the sampled region to the occipital lobe. Brain BHB and lactate levels were quantified using an adiabatic homonuclear editing sequence (Sammi et al. 2001). Localization was achieved through a (coronal) paired slice selective adiabatic refocusing pulses, 10-mm thick. In-plane selection (4 × 4 cm2) was achieved through use of two dimensions of ISIS selection. Water suppression was provided by a pre-echo Gaussian modulated DANTE pulse. Homonuclear editing was performed using a Gaussian shaped DANTE pulse, placed at 4.15 ppm, so as to allow equivalent editing of lactate and BHB. Data were acquired using TR = 2 s, TE = 148 ms, with a scan time of 4.5 min.

Processing was performed using 2–8-Hz Lorentz to Gauss conversion and 50-Hz convolution difference. The frequency domain data were analyzed by phasing the individual data blocks. The difference (edited) spectra were curve fit using identical linewidths for NAA, lactate and BHB, a 7.4-Hz coupling constant for both BHB (1.2 ppm) and lactate (1.33 ppm, which coedits with BHB). These resonances were quantified using NAA (10 mm) due to its minimal concentration differences between white and gray matter (Pan et al. 1998).


All subjects were free of neurological or psychiatric disorders. Under an IRB approved protocol, six adult subjects were studied. After an overnight fast, subjects were catheterized in the antecubital fossae bilaterally. One catheter was used for the infusion, the other used for periodic blood sampling. After placement of catheters, the subject was positioned within the magnet and system optimization acquisitions were performed. After baseline MR data were acquired, acute hyperketonemia was induced using a variant of the protocol described by Veneman et al. (1994). A sterile, pyrogen-free solution of 200 mm pH 7.1 sodium d-BHB (Solvay Peptisyntha and Cie, SNC, Brussels, Belgium) was infused at a bolus rate of 80 µm/kg/min followed by an adjusted 20 µm/kg/min for the duration of the infusion study of approximately 75 min. Plasma BHB measurements were obtained online using an Analox GM-7 analyzer while plasma lactate assays were performed later. This infusion protocol typically raised and maintained plasma BHB concentrations at approximately 2 mm. The spectroscopic data were acquired during the final 30 min period of the infusion when plasma levels were relatively stable. Whilst in the magnet, subjects were monitored with pulse oximetry. The entire study duration was typically 100 min, and all subjects tolerated the infusion well.


Figure 1 shows the time course of plasma BHB measurements averaged over all six subjects. Plasma BHB concentration rose quickly with infusion, reaching the target 2 mm level within 15 min of the beginning of the infusion. As the infusion continued, there was a slight tendency for the plasma ketone levels to further increase (most likely reflecting a decrease in skeletal muscle consumption), requiring a downward adjustment of the infusion rates. The final plasma value was obtained after removal of the subject from the magnet (t = 76.7 ± 11.7 min). Plasma lactate decreased transiently during the infusion, but returned to normal levels (0.85 ± 0.31) by the end of the study.

Figure 1.

 Average time course of plasma BHB of all subjects during infusion, begun at time t = 0 min. The bars represent the standard deviations of the plasma measurements, the standard deviation of the time of each measurement is listed on the x-axis.

Figure 2 shows the region of interest, located in the occipital lobe. Such a centrally located mixed tissue voxel as has been shown by previous studies (Pan et al. 2000b) is typically 66% gray matter, 33% white matter. Figure 3(a and b) shows the performance of the editing sequence at baseline and the summed spectrum obtained between t = 45–60 min into the infusion from subject 3. The plasma concentration of BHB of subject 3 was 2.34 ± 0.21 mm over this period.

Figure 2.

 Inversion recovery scout image showing location of typical voxel studied in the occipital lobe. The selected voxel is 4 × 4 × 1 cm3.

Figure 3.

 (a) Spectra showing the spectral editing (subtractive) process, with the inset demonstrating the edited spectrum with lactate. The inset is plotted with 4 × amplitude compared to its base spectra. The edited spectrum is repeated on the right to facilitate comparison. (Because of the proximity of the glutamate and aspartate C2 resonances to the inversion pulse, the glutamate and aspartate C3 resonances co-edit but with lower efficiency, resulting in apparent differences at the 2.6 ppm position.) (b) Edited spectrum acquired at the end of an infusion study.

To demonstrate the stability of this study, Fig. 4(a) displays spectra of lactate and BHB over the duration of the infusion in volunteer 5. Averaging the data for all subjects (Fig. 4b), it is evident that the rise is prompt, with the majority of increase occurring within the first 30 min of infusion. Over the six subjects, the plasma values rose from 0.2 ± 0.1 mm at baseline to 2.12 ± 0.30 mm while the mean tissue BHB concentration rose from 0.16 ± 0.07 mm to 0.24 ± 0.04 mm (p < 0.05, unpaired two-tailed Student's t-test). Notably, lactate concentrations did not change with the infusion, 0.68 ± 0.09 mm to 0.72 ± 0.09 mm. The table quantitatively summarizes the steady state data from the six subjects.

Figure 4.

 (a) Time course of spectra from a single volunteer showing the small increase in BHB over the 1.25 h study period. The infusion started at t = 0 min. Each spectrum represents an accumulation time of 15 min. (b) Mean values of tissue BHB measured across all volunteers (n = 6) as a function of time into the infusion. The data were averaged over 30 min blocks, and all points into the infusion were significantly different from the baseline study (p < 0.05, paired t-test).


BHB rises with acute hyperketonemia

These data demonstrate that plasma BHB levels can be quickly increased with the acute infusion protocol used, rising from overnight conditions (0.20 ± 0.10 mm) to 2.12 ± 0.30 mm. These elevated plasma levels of BHB are higher than that reached after 48 h of fasting (1.67 ± 0.34 mm). During the hyperketonemic infusion, the tissue BHB increased by approximately 47%, from 0.16 ± 0.07–0.24 ± 0.04 mm. This may be compared with the 48 h fasted data (Pan et al. 2000a), which found the tissue BHB concentration to be 0.6 ± 0.26 mm, significantly correlated to the plasma BHB concentration with a slope of 0.26.

Methodologically, the present data acquired with the double adiabatic editing sequence represent an improvement over the previous method used to acquire the fasting data. This does not preclude comparison between the previous and present data; however, the present method has decreased variability, with the standard deviation in the (present) baseline lactate data at 0.09 mm, previously at 0.16 mm. However, a part of the variation in the 48 h fasting BHB data (0.60 ± 0.26 mm) is most likely due to inter-individual differences in the speed with which each subject adjusted to ketosis (by 72 h of fasting, the group measure of brain BHB reached 0.98 ± 0.16 mm). Thus a strict comparison of 48 h fasted and non-fasted data is complex. However, using an unpaired t-test, a direct comparison between the 48 h fasted and acute hyperketonemic data showed the latter to be significantly lower (p < 0.02) despite its higher plasma BHB level.

As measured by in vivo MR, the BHB signal represents both intra- and extracellular components. If the relaxation parameters and visibility of blood and CSF BHB are equivalent to that of brain, we can determine a lower bound to the brain BHB concentration. Using a CSF–blood partition factor of 0.15 (Lamers et al. 1987; measured under fasting conditions), at a concentration of 2.12 ± 0.3 mm and a non-brain volume contribution of less than 10%, at a minimum, the brain BHB would be 0.12 mm. This value is similar to the baseline BHB measurement at 0.16 ± 0.07 mm. However, if the relaxation values of blood BHB are shorter than those of tissue (due to the susceptibility effects of deoxyhemoglobin), and given the long echo time of this acquisition (Te = 148 ms), brain BHB will be higher than 0.12 mm. Thus the concentration of 0.24 mm represents an upper limit to the brain BHB.

Whether steady state was actually reached may be of concern in the physiological analysis of the data. Although there was some variability in the rapidity with which individual subjects reached steady state, all subjects reached their steady state BHB measurements within 45–60 min of the infusion start. This is consistent with earlier rodent data, which used ketone infusions of approximately 30 min to 1 h to reach steady state (Hawkins et al. 1971; Ruderman et al. 1974; Corddry et al. 1982). As demonstrated in Fig. 4(b), there is a suggestion of a slight rise in the brain BHB with time; however, given the error in the measurement, ad hoc comparisons of the intermediate time points to the final time point did not show any significant differences.

Transport models

To evaluate the acquired data in terms of fasting induced changes to the monocarboxylic acid transporter, it is important to consider that the steady state BHB concentration, whether non-fasting or fasting, necessarily represents a balance of transport and oxidative clearance (Eqns 1 and 2).


Jin is the plasma to brain BHB flux, Jout is brain to plasma BHB flux, CMRbhb is the cerebral consumption rate of BHB, Bbhb is the brain concentration of BHB, Pbhb is the plasma concentration of BHB.

If induction is hypothesized not to occur in this reversible symmetric Michaelis–Menten model (i.e. neither Vmax nor Km change with fasting), the inward flux Jin in acute hyperketonemia should be close to that in the 2 day fasted case since similar Pbhb values were obtained. Thus, in order to explain the higher brain BHB in the fasted case, the clearance of BHB (through oxidation CMRkb and/or efflux Jout) would have to be lower. However, CMRbhb has been reported to be elevated in fasting by Owen et al. (1967) and by Hasselbalch et al. (1994, 1996), with the latter studies reporting that ketone oxidation more than doubled with fasting (8.1 ± 4.7 µm/kg/min non-fasted versus 20.0 ± 15.7 µm/kg/min fasted). Thus, if induction does not occur, this implies that Jout should decrease in the fasted case. However this is also unlikely, since the higher brain BHB concentrations imply that Jout must be increased in the fasted case relative to the acute hyperketonemic state. Notably, inclusion of a diffusion (net) efflux Jdiff=D(Bbhb − Pbhb) would similarly increase with elevated Bbhb. Thus, the presence of a higher brain BHB level in the fasted case compared with the non-fasted case suggests that transport of ketones is up-regulated in humans with fasting. These considerations can be analytically described by solving Eqn 2 (assuming a reversible symmetric Michaelis–Menten facilitated carrier model) to give the concentration of BHB at steady state (Eqn 3):


From Eqn 3 it is evident that without any changes in CMRbhb, the transport parameters Km or Vmax, for a given value of Pbhb, Bbhb is fixed.

A decrease in Jout can occur without alteration in transport parameters if an asymmetric model of BHB transport is used. Physiologically, this describes the case where the affinity for BHB is characterized by two values, on the luminal side, Km,l and on the abluminal side Km,a. In this model, the concentration of lactate may affect the value of Km,a since lactate may serve as a competitive inhibitor of BHB. As we had reported (Pan et al. 2000a), by the second day of fasting, lactate nearly doubled from a baseline of 0.69 ± 0.17 mm to 1.31 ± 0.26 mm. Thus, with the competing presence of lactate, Km,a would increase from its original value Km,a,o as in Eqn 4.


where Ki is the affinity of the transporter for lactate.

Equation 5 describes the brain BHB with a reversible asymmetric Michaelis–Menten model:


Equation 5 indicates that in order to explain a 2.5-fold higher Bbhb in the fasted case relative to the non-fasted case (0.6 mm, 0.24 mm, respectively) without altering other components, Km,a must increase by 2.5-fold. However, with literature values for the Ki binding of lactate to the MCT of approximately 1.5–2 mm (1.8 mm, Oldendorf 1973), a two-fold lactate increase (from 0.7 to 1.3 mm) can at most result in a 24% increase in brain BHB. Thus, competition from lactate appears to be insufficient to explain the 2.5-fold BHB rise under the fasting conditions.

The rodent studies by Gjedde and Crone 1975 have shown that the Km and apparent Vmax for ketone transport increase, rather than decrease, with fasting. Taken together with studies showing increased ketone metabolism with fasting, our finding of elevated Bbhb for similar Pbhb for the fasted relative to the non-fasted state provides strong evidence that the transport activity for BHB is increased with fasting.

Notably, the ratio of the tissue BHB concentrations between the fasted and non-fasted cases is 0.60/0.24, or approximately 2.5. This may be compared directly with the ratio of cerebral ketone consumption rates as cited above (Hasselbalch et al. 1994, 1996), at 20.0/8.1, or approximately 2.5. Such a correlation suggests that under these conditions and at this level of ketonemia, cerebral ketone consumption is a function of tissue ketone levels. Greater information as to the contribution of oxidative flux from ketones can be obtained through the infusion of labeled glucose and ketones.

No significant differences in tissue lactate were detected with the acute hyperketonemia. This differs from that seen from the fasting hyperketonemia, where tissue lactate rose from 0.69 ± 0.17 mm at baseline to 1.31 ± 0.26 mm after 48 h of fasting. If, in the fasted case, the elevation of lactate were due to metabolic displacement of glucose by ketones, the absence of an increase in cerebral lactate in the non-fasted case suggests that the oxidative flux by ketones is relatively small, consistent with limited ketone transport.

Under normal conditions, elevations in plasma BHB are accompanied by increases in plasma acetoacetate. However, in this initial study of brain BHB metabolism, we did not simultaneously infuse acetoacetate. As measured in rodents, the relative rate of transport between BHB and acetoacetate is not significantly different between fasted and non-fasted states (Hawkins et al. 1986). Thus, although the interconversion of BHB and acetoacetate has been reported to be fast (Barton 1973), to account for both ketones, these BHB values may be scaled to provide an upper bound measure of their combined transport.

These data imply that the way the brain handles BHB in fasting induced ketosis is not identical to that in acute hyperketonemia. Given that it is unlikely that ketone body oxidation is greater in the non-fasted compared to the fasted state, it is likely that transport is up-regulated in the fasted state, consistent with rodent data and Hasselbalch et al. 1995. Additional studies over a range of plasma ketone levels in the fasted and non-fasted state will provide data as to the values for the parameters of transport.


This work was supported by the Charles A. Dana Foundation (JWP), the Albert Einstein College of Medicine General Clinical Research Center 1MO1RR12248 and NIH NINDS PO1NS-39092 (HPH).