Is myo-inositol a measure of glial swelling after stroke? a magnetic resonance study

Authors


Abstract

Purpose

To determine whether the hypothesis that the phenomenon of persistent cytotoxic edema in the subacute stage of ischemic stroke is in fact associated with the glial population. This is done by assessing the evolution of both the apparent diffusion coefficient (ADC) and the glial-specific marker myo-inositol (Ins) in a group of patients, and by comparing the results with the total cellular density by means of the creatine (Cre) level.

Material and Methods

Twenty-two patients with stroke in the territory of the middle cerebral artery were each examined once only at a time ranging from eight hours to six days following the onset of symptoms. Lesion-to-contralateral values of ADC were obtained based on diffusion-weighted echo-planar imaging. Short TE single-voxel proton magnetic resonance (1H MR) spectroscopy was used for quantification of cerebral metabolites in infarcted regions. Their levels were also compared with those in homotopic contralateral regions.

Results

In the stroke lesion, there was a significant correlation between ADC and the Ins level, albeit less pronounced than that for Cre. During different pathophysiological stages between 12 hours and three days, the Ins-to-Cre ratio increased by a factor of two and returned to apparently normal thereafter.

Conclusion

Our study provides the first demonstration of a relationship between persistent cytotoxic edema and the glial population in the context of cell swelling due to osmotic imbalance in stroke patients. J. Magn. Reson. Imaging 2003;17:11–19. © 2002 Wiley-Liss, Inc.

DIFFUSION-WEIGHTED IMAGING is a powerful tool for detecting early ischemic brain lesions, which are accompanied by a significant reduction in apparent diffusion coefficient (ADC) (1–2). Despite its indisputable role in mapping ischemic areas of cellular swelling called cytotoxic edema, there remain many questions about the actual pathophysiological processes that underlie the changes on diffusion-weighted imaging (3).

Although all arterial ischemic infarcts have a common etiology, the temporal rate of tissue evolution towards infarction varies between patients (4). This variability is possibly due to anatomical and physiological differences in regional blood flow, duration of vascular occlusion, location and type of brain tissue that is affected, and regional temperature. The mechanisms of temporal progression of ischemia to final irreversible brain damage are triggered by impairment of brain energy metabolism (5). The process begins with a simultaneous depletion of cellular energy stores and loss of ionic homeostasis, leading to intracellular sodium overload. The cell volume changes resulting from these altered intra-to-extracellular solute concentrations greatly affect the ADC. Subsequent lactate accumulation during anaerobic metabolism and release of monoamine transmitters into the extracellular space prolong the conditions favorable for glial swelling, also referred to as secondary cytotoxic edema (6). On the other hand, cells have mechanisms to partially restore their volume to normal (7). The cell volume is regulated not only by inorganic ions, but also by organic osmolytes, such as polyols, amino acids, and methylamines. For example, myo-inositol (Ins) contributes to osmoregulation in glial cells. There are at least two reasons why Ins may change in concentration after stroke. The first is because it serves as an osmolyte in volume regulation (8–9), and the second is because of proliferative changes of glial cells and reactive gliosis (10).

To obtain a clearer understanding of the brain response to ischemia, neurons should not be studied in isolation but in the context of surrounding glial cells. Proton magnetic resonance (1H MR) spectroscopy is unique in its ability to detect both the total cell pool and its individual glial and neuronal elements. While the total creatine (Cre) pool (phosphorylated and unphosphorylated creatine; in this text Cre refers to the combinded values) covers both cell types (neurons and glial cells) and is therefore a measure of cell density, N-acetyl-L-aspartate (NAA) is located in the neuronal population and Ins in glial cells (11). It has been recently reported that a relationship exists between ADC and Cre level in patients with early stroke (12). It has been shown that cases of near normal Cre level (i.e., no cell loss) have greatly lowered ADC values. Conversely, cases showing a reduced Cre level (i.e., cell loss) have higher ADC values (towards near-normal and above). To investigate this further, we designed the current study 1) to evaluate the Ins level from strokes up to six days, and 2) to compare the relationship of ADC to Cre with a possible—as we hypothesize—relationship of ADC to Ins. With this in mind, we investigated the concentrations of Ins and Cre in the infarcted region, as well as in the homotopic contralateral region, for a group of patients, each of whom was given an initial MR examination at times ranging from eight hours to six days following onset of symptoms. We used short TE 1H MR spectroscopy for metabolic quantification and diffusion-weighted imaging to measure ADC values.

MATERIALS AND METHODS

Inclusion Criteria

Out of approximately 175 ischemic stroke patients referred for routine MR scans, 25 suitable candidates were enrolled in the study over a six-month period. The selection was based on a case-by-case decision subject to the following criteria: 1) patients clinically diagnosed with ischemic stroke and referred for MR imaging within the first six days following onset of symptoms, 2) Patients demonstrating abnormality on diffusion-weighted imaging of at least two times in size, as used for 1H MR spectroscopy of 8 cc in the territory of the middle cerebral artery (MCA) in order to ensure a more homogenous patient population.

Study Protocol

MR imaging and 1H MR spectroscopy were performed using a 1.5-T whole body scanner (Magnetom Vision; Siemens Erlangen; Germany) with actively shielded gradients (25 mT/m). The head coil was circularly polarized. The routine clinical scanning protocol comprised diffusion-weighted imaging, T2-weighted imaging, and three-dimensional time-of-flight MR angiography.

The study protocol consisted of two series and was approved by the hospital's ethics committee. The first series involved ADC mapping. Diffusion contrast generation was based on a Stejskal-Tanner spin-echo diffusion-weighting segment of TE 136 msec (three b-values of 50 s/mm2, 500 s/mm2, and 1000 s/mm2 sequentially along the orthogonal directions Sx, Sy, and Sz) with an ensuing echo-planar imaging sequence. The acquisition parameters were as follows: TR 5100 msec, number of excitations (NEX) two, 96 × 128 matrix, 22 × 22-cm field of view, 5 mm slice thickness, and 2 mm slice gap. The acquisition time was 36 seconds. ADC of the trace of the diffusion tensor was calculated as the mean ADC in the three orthogonal directions (13). Series two involved 1H MR spectroscopy, using the stimulated echo acquisition mode (STEAM) method with TR 3000 msec, TE 20 msec, and NEX 128, and voxel dimension of 20 mm in all three axes. A set of two 1H MR spectra, one in the infarcted region and a second in the essentially normal tissue in the corresponding contralateral hemisphere, was acquired. For each location, a reference scan was acquired without water suppression. For spectral analysis and quantification, the LCModel routine was used, which analyzes an in vivo spectrum as a linear combination of model in vitro spectra from individual metabolite solutions (14). Stroke lesion and contralateral region were treated likewise: 1) the water signal served as an internal standard and was assigned 0.75 mL/g, 2) neither metabolite levels, nor unsuppressed water levels, were corrected for T1 and T2 relaxation because of the uncertainty of relaxation parameters in diseased tissue. The reliability of the estimated concentrations was analyzed by means of the SD for each single peak in the spectrum, which the program provides. For example, %SD = 25% means that changes of 50% in the concentration can be detected with high reliability. In this study, data sets showing Ins and/or Cre levels with %SD larger than 15% were excluded. This selection criterion takes into account 1) spectral quality and 2) that even if the spectral quality itself is good enough at low signal-to-noise ratio (S/N), the water content of the reference scan is highly uncertain due to accumulation of free interstitial fluid (12).

RESULTS

Twenty-two of 25 patient data sets were available for analysis, as three data sets were excluded because of high statistical uncertainty of the Ins level. The results of the entire patient group are summarized in Table 1. This lists the molar concentrations of Ins and Cre in stroke lesions (ischemic core), as well as in homotopic contralateral regions. Two Ins patterns can be discerned. The first group (patients 1–13) have concentrations of Ins higher than normal, and the second (patients 14–22) have lower lesion Ins levels. This is illustrated in Figure 1 by patient 4 with subacute stroke at 16.5 hours, and Figure 2 by patient 18 with established stroke at 82 hours. For each case, a representative T2 fluid-attenuated inversion-recovery (FLAIR) image, diffusion-weighted image, ADC map, and 1H MR spectra are shown. The ADC results are listed as mean values averaged over the volume of interest (VOI) as used for 1H MR spectroscopy. The ratios of ADC (ADCn)—meaning “normalized”—refer accordingly to the VOI on the lesion side related to the contralateral mirror VOI in the healthy hemisphere. These values are less dependent on the composition of the cortex to subcortical white matter.

Table 1. Compilation of MR Data
Patient no.Sex/ageTime after onset (h)Crea lesion (mM)Insb lesion (mM)Ratio Ins/Cre lesionCrec contra (mM)Insd contra (mM)Ratio Ins/Cre contraADC lesion (10−3 mm2/seconds)ADC contra (10−3 mm2/seconds)ADCn
  • a

    %SD increases from 5% for no. 1 to 12% for no. 22 (see Materials and Methods).

  • b

    %SD increaseses from 8% for no. 1 to 15% for no. 22.

  • c

    %SD = 5–8%.

  • d

    %SD = 8–10%.

  • ADCn-ratio of lesion to contralateral ADC.

1M/4586.435.540.868.005.060.630.320.890.36
2M/47116.896.210.907.404.640.630.300.820.37
3F/55126.655.600.847.605.200.680.330.850.39
4F/5516.56.209.701.566.835.810.850.330.940.35
5M/5826.55.286.401.217.504.440.590.380.870.44
6F/51314.936.361.297.334.890.670.330.860.38
7M/5732.55.206.611.278.495.100.600.420.830.51
8F/86334.406.801.556.804.400.650.400.990.40
9M/47384.975.941.207.245.380.740.300.830.36
10M/70525.757.401.297.674.860.630.451.180.38
11F/6857.55.646.851.218.174.930.600.430.920.47
12M/80594.236.471.537.804.700.600.550.910.60
13M/54764.405.121.166.134.600.750.441.020.43
14F/56823.863.961.037.674.860.630.570.910.63
15F/55843.683.100.847.304.900.670.490.940.52
16M/50904.324.521.057.455.100.680.621.060.58
17M/61936.404.950.778.055.720.710.350.910.38
18M/62934.654.500.976.606.050.920.440.990.44
19M/501022.902.100.726.806.140.900.600.880.68
20F/741244.903.900.806.305.480.870.400.810.49
21M/47126.52.902.800.976.905.280.770.460.820.56
22F/671372.462.200.897.305.300.730.560.890.63
Figure 1.

A 55-year-old woman (patient 4 in Table 1) with a right MCA infarct, 16.5 hours after onset. a: Axial FLAIR image through the lateral ventricles showing subtle gyral T2 hyperintensity and a few subcortical hyperintense foci in the right posterior frontal lobe. b: The infarct is made obvious by a corresponding diffusion-weighted image, as a well-defined area of increased intensity. Note the greater signal of the cortex compared with the subcortical white matter, with corresponding decrease in ADC values on the ADC map (c). This may be due to the higher density of neurons in the cortex, which are more sensitive to the ischemic insult, and therefore manifesting as more pronounced regions of signal change in diffusion-weighted imaging. d:1H MR spectra of the stroke lesion (top) and on the contralateral side (bottom) demonstrate a clear increase in Ins at 3.56 ppm in the stroke lesion, with a near normal Cre level at 3.0 ppm. Decreased NAA at 2.0 ppm and elevated lactate at 1.3 ppm are compatible with subacute ischemic lesions. MR spectra shown are the LCModel fits (bold lines) overlaid on the experimental data (thin lines).

Figure 2.

A 56-year-old woman (patient 18 in Table 1) with a left MCA infarct, 82 hours after onset. a: Axial FLAIR image showing gyral swelling and subcortical hyperintensity in the left frontoparietal lobe. b: Axial diffusion-weighted echo-planar imaging (trace image) showing the infarct as a hyperintense lesion, together with the corresponding dark area of reduced ADC on the ADC map (c). d:1H MR spectra of the stroke lesion (top) and on the contralateral side (bottom) demonstrate a decrease of all metabolites: Ins (3.56 ppm), choline (3.2 ppm), Cre (3.0 ppm), and NAA (2.0 ppm). Note that the Ins-to-Cre ratio is higher in the lesion compared to that in the homotopic contralateral region. MR spectra shown are the LCModel fits (bold lines) overlaid on the experimental data (thin lines).

As a point of reference, the mean values of the contralateral (normal) tissue levels (N = 22) are as follows: Ins, 5.13 ± 0.48 mM; Cre, 7.33 ± 0.60 mM; ratio of Ins to Cre, 0.71 ± 0.10; ADC, 0.91 ± 0.09 × 10-3 mm2/second. The “normal” value for Ins may be underestimated (15), as no relaxation corrections were made.

Figure 3 shows the time course of the ratio of Ins to Cre. The most striking observation is the plateau of twice the normal value between approximately 12 hours and 72 hours. Note that the statistical uncertainty for the data points increases from 9% to 20% (left to right) as the detectable signals become smaller.

Figure 3.

A time course of the ratio of Ins to Cre derived from the lesion. Error bars represent %SD from the ratios of the two independent measures in the lesion, Ins, and Cre (%SD increases from 9% to 20% from left to right). To avoid confusion, only error bars for points at the edge of the time course are shown. For reference, the mean value of the ratio of Ins to Cre on the contralateral side is shown as a solid line. Dotted lines depict 1 SD from the mean value.

Table 2 contains the results of the regression analysis for ADC vs. Cre, ADCn vs. Cre, ADC vs. Ins, and ADCn vs. Ins. Figure 4a plots the correlation between ADCn and Cre level, showing that the higher the Cre level, the lower the ADC values. This is expressed by the strength of relationship R2 = 0.65. The correlation of ADCn vs. Ins level is shown in Figure 4b. Notice that the correlation of the former is stronger than the correlation of the latter, as defined by R2 = 0.46.

Table 2. The Relationship Between ADC and Metabolite Levels
Type of correlationR2
ADC-to-Cre0.58
ADCn-to-Cre0.65
ADC-to-Ins0.31
ADCn-to-Ins0.46
Figure 4.

Correlation plot of ADCn vs. lesion Cre concentration (R2 = 0.65) in part a, and ADCn vs. lesion Ins concentration (R2 = 0.46) in part b. The arrow head indicates the contralateral mean Ins level. The solid line in both graphs represents the regression line, while the 95% upper and lower confidence limits are presented by dashed lines.

Decreased NAA levels and strongly elevated lactate signals were seen in all patients. There was no systematic relationship either between NAA level and ADC or between NAA level and Ins level.

DISCUSSION

This study shows that the ratio of Ins to Cre increases significantly between approximately 12 hours and 72 hours after stroke, and returns to levels that appear to be normal if cell loss is taken into account. The study confirms that there is a relationship between Ins and ADC. This is, however, weaker than that between Cre and ADC. The results are discussed in three different directions: firstly, the role of Ins in cell volume regulation (as a background review), secondly, the role of Ins in cell density, and thirdly, the role of Ins in cell swelling.

Cell volume regulation by Ins is a fundamental physiological process. Cell volume can be altered by changes in extracellular osmolarity or intracellular solute content. Separating these entities is difficult because of two issues: 1) complexity of ion homeostasis and 2) unpredictable quantity of ion shifts. The following examples attempt to shed some light upon this, in ascending order of complexity. First example: Osmoregulatory changes in Ins content have been extensively examined using cell lines, as they allow well-aimed extracellular osmolarity or intracellular solute content to be manipulated and, thus, conditions of uptake and loss (of Ins) can be mimicked (8–9, 16, 17). Second example: Accumulation of Ins has been found in chronically hypernatremic mice, and the converse (depletion of Ins) in hyponatremic mice. This suggests that intracellular changes in concentration of Ins serve to counteract extracellular changes in osmolarity (18). This has been demonstrated in an apparent increase in Ins level by 1H MR spectroscopy in an infant with severe dehydration (19). Third example: Cerebral hyperammonemia caused by chronic liver disease represents a unique, relatively simple in vivo model with which to study the counteractive role of osmotically active solutes, namely accumulation of glutamine and subsequent loss of Ins (20). This model has received much attention because it is 1) cell type-specific and 2) involves only one ion (ammonia) type that disturbs the ionic homeostasis. Glial cells seem to be the primary target (with neuronal dysfunction being the consequence) because ammonia detoxification depends mainly on glutamine formation in this cell type. Thus, an increase in glutamine and concomitant clearance in Ins have been reported in patients with hepatic encephalopathy by 1H MR spectroscopy.

Hypoxia-ischemia reveals a much higher order of complexity and therefore the general osmolytic stress is not predictable. This is because 1) both neurons and glial cells are targeted, 2) a large number of ion types are involved, and 3) neuronal-glial interactions cause additional ion shifts. A simplified scheme (Fig. 5) shows some potential phases of ischemic brain damage. In the first phase (sketch I), there is an influx of Na+, water, and Ca++ (not shown) accompanied by a large increase in extracellular K+ in neurons, while glial cells are known to be slightly more resistant to ischemia (21). In the second stage, glial cells, if activated, protect the neuron by taking up and buffering the excess K+ released by neurons (sketch II) (22). According to the principle of “ion control ranges over volume control” (23), neuron–glial interactions (glial cells are the intimate partners of neurons) do not respond to homeostatic mechanisms by acting against osmosis. After a while, however, the glial cell becomes overwhelmed either by the number of K+ ions or by the prolonged exposure time. Therefore, in time, the electrolyte is expelled from the glial cell to be replaced by Ins (sketch III). When brain glial cells are exposed to chronic hypertonicty (more than eight hours), this ionic electrolyte-to-Ins replacement has been reported (8). It should be noted that for reasons of simplicity, shifts of other electrolytes or osmolytes are not included.

Figure 5.

Schematic diagram of a glial-neuron unit. Phase I after stroke shows intracellular Na+ overload and an increase of extracellular K+. The normal level of Ins for glial cells is labelled INS. In phase II, glial cells take up and buffer the excess K+ released by the neurons. Phase III depicts a stage later than 10 hours after stroke showing intracellular K+ replaced by Ins in high concentration. In phases III and IV, the activated glial cells are swollen and depicted as enlarged. In phase IV, the Ins level returns to normal as far as the viable single cell is concerned. The cell density is, however, diluted, shown here as a lysed neuron (dotted lines). For the sake of simplicity, other electrolytes or osmolytes are omitted.

Given this background, we are now able to discuss the role of Ins in cell density. We separate the Ins spectroscopic findings of this study into two patterns: one with higher lesion Ins levels compared to normal tissue, and one with lower lesion Ins levels. For the first group, Ins levels are higher in the lesions irrespective of whether there is cell loss or not (patients 1–13 in Table 1). In the second pattern (patients 14–22 in Table 1), the observations are complicated by the fact that there is a net decrease in both the glial and neuronal cell density. This condition is supported by the considerable decrease in Cre levels and elevation of the ADC, reflecting cell lysis and enlargement of the extracellular space (12). Therefore, the observed decrease in lesion Ins levels may either be due to an actual reduction in the Ins levels in the remaining glial cells or an increase in the individual gial cell Ins against a background of a decrease in total glial cell numbers.

Brain cells begin to die after just a few minutes without blood and oxygen. Due to both the physical and chemical changes that occur in the brain with ischemic stroke, damage can continue for several days. Hence, the amount of residual viable cell density is an unknown variable. To put this question in cellular terms, what fraction of the ADC signal comes from neurons and what fraction from glial cells? Although there is a neuronal marker and a glial marker, no technique exists to directly determine the fractional densities of neurons and glial cells. It is in the nature of NAA and Ins that these markers themselves become changed in neuropathological events and are thus of no use as markers of fractional cell densities. Impaired synthesis and degradation are the most obvious additional changes for NAA (24), while, as mentioned, Ins functions as an osmolyte in glial cells by changing its concentration.

Does this work in an indirect way? Cre is a good indicator of the general cell density for two reasons. First, 1H MR spectroscopy detects the total pool of Cre, irrespective of how much phosphorylated Cre is converted into unphosphorylated Cre on the induction of ischemia (25). Second, Cre is found in both cell types. It should be noted that the Cre concentration in human grey matter exceeds that measured in white matter, suggesting a higher Cre level in glial cells compared to neurons (15). This therefore requires normalized Cre levels by means of lesion-to-contralateral levels. However, with regard to fractional cell densities, unknown differences in rates of disintegration of neurons and glial cells make it impossible to predict what proportion of the signal comes from each cell class. It should be noted that when neurons are lost, glial cells start to proliferate, and this is called gliosis. This process leads to partially compensating effects on the Cre level. The ratio of Cre (lesion) to Cre (contralateral) is therefore not applicable as a correction factor to account for fractional cell loss. For the same reason, the ratio of Ins (lesion) to Cre (lesion) does not standardize the Ins concentration. However, this quantity may give a solid clue as to how much Ins is produced per cell density. This is demonstrated in Figure 3. The values for the ratio of Ins to Cre were already slightly increased for < 12 hours; there then follows a plateau with twice-normal values and with large regular fluctuations (probably due to the diversity of ischemic thresholds) from approximately 15 hours to approximately 80 hours, and thereafter a drop to apparently normal values. This pattern most likely represents an upregulation followed by a downregulation of the Ins uptake as the osmotic conditions change over time. This is shown in Figure 5. It is suggested that the decreased Ins levels in established stroke (sketch IV in Fig. 5) result mainly from a cell loss rather than from an osmotically active depletion of Ins, as is the case in hepatic encephalopathy (20).

The analysis of the role of Ins in cell swelling extends findings of earlier research (12). The ADC-to-Cre correlation is expressed by a strong relationship, showing that the higher the Cre level the lower the ADC values (Table 2, Fig. 4a). This result is closely comparable to our previous study (12) based on different MR protocols. Our present result further indicates that there is a significantly weaker correlation between ADC and the Ins level after stroke (Table 2, Fig. 4b). We suggest three explanations. First, cell swelling occurs primarily when water enters the cell in order to decrease the electrolyte concentration. In cases where ions are replaced by Ins (8) in an attempt to create a less harmful environment to the cell, the concentration of Ins does not equate glial density only. In addition, accumulation and clearance of organic osmolytes are slow processes relative to electrolyte changes (7). Thus, any correlation in fact becomes less obvious. Second, all types of glial cells can proliferate in response to brain injury, including hypoxia-ischemia (26). As a result, the Ins level may change accordingly. As glial proliferation is normally noted from three days after the insult, only the data of later points in time are affected. Third, glycine contributes approximately 10% to the signal intensity of the Ins peak at 3.56 ppm (27).

A number of limitations (see reference12), such as uncertainties in brain water content in different stages of vasogenic edema and composition of the cortex to subcortical white matter, makes testing of the hypothesis difficult; the former can lead to errors in the LCModel analysis, while the latter affects ADC. This is clearly demonstrated by the comparison of ADC vs. ADCn in Table 2. Nonetheless, support for this model of interplay between the ADC and the Ins levels is given by investigations of intracellular metabolite diffusion (28), where changes of ADC were most pronounced for Ins.

In conclusion, our results show that the Cre level, used as a surrogate marker of cell density, correlates with ADC. In an analogy, using Ins as a point for glial density, it is apparent that there is a similar but weaker relationship. This is because the relationship is compounded by other superimposed mechanisms, the most likely of which is the accumulation of Ins for osmolytic needs. These findings are in line with a persistent cytotoxic swelling, attributed to the glial population, found in early subacute ischemic infarcts.

Acknowledgements

The authors acknowledge the dedicated patient care provided by Victoria Villanueva Cadevida and Leticia Delacruz Camua, Healthcare Assistants.

Ancillary