Magnetic resonance imaging approaching microscopic scale: Maturation stages of Acetabularia mediterranea reproductive caps

Proton magnetic resonance imaging has been carried out in vivo on many biological samples (Mansfield & Morris, 1982) and is coming into increasing use in medical diagnosis of soft tissue defects in man (Budinger & Lauterbur, 1984). Spatial resolution from a whole body scanner, quoted in 1984 as at best 5 mm³ on a volume basis, today can sometimes be better than 1 mm in-plane in a slice thickness of about 2 mm. Improved spatial resolution could lead to an increasing range of biological applications. But the spatial scale into which such improvements are now reaching, hundreds and tens of micrometres, is somewhat of a no-man’s-land between the millimetric requirements of medical diagnosis and the submicrometre expectations of the modern biological microscopist. Hence, instrumental advances of one or two orders of magnitude do not evoke from experimental biologists in general a response matching their stature as engineering achievements.

In the field of developmental biology, however, the scale of hundreds to tens of micrometres is much more significant than in many other fields, because morphological differences between parts of an organism commonly arise on just that spatial scale, in both plants and animals and in embryos ranging from insect to human. It is quite common in animals for a large unicellular egg, sometimes as large as the 1 mm diameter frog (Xenopus laevis) egg imaged by Aguayo et al. (1986) to divide repeatedly without feeding, so that cell diameters are no more than 10 pm when a sequence of major morphogenetic events begins at the onset of gastrulation. The developmentalist whose work requires the observation of individual cells or subcellular structures will at that point lose interest in a 10 pm resolution microscopic technique. But if major cytochemistry, such as state of binding of water or proportions of bound and free water, is becoming patterned on a larger scale than the single 10 pm cell, then a technique that can reveal such changes may be very powerful at a somewhat lower spatial resolution.

In multicellular plants, cells are commonly 50-100 ftm in diameter and often much greater in length. Thus the first observation of the existence of cells, by Hooke in 1665, was made in wood. For information on tissue patterns in development and in regeneration after wounding, including some mention of cell sizes, see Wilson & Warren (1984). There is commonly a marked contrast between different parts of a plant stem in respect of water content, which has been of assistance in obtaining good spatial resolution in magnetic resonance imaging (Eccles & Callaghan, 1986).

The sample used in the present study provides a good test pattern for spatial resolution in the range of tens of micrometres; but it is a marine organism that consists chiefly of water throughout, with no dry parts, and it thereby challenges the contrasting ability of proton resonance imaging. Acetabularia mediterránea (Dasycla-dales; Chlorophyta) is a single cell that, over a period of a few months of vegetative growth, develops into a cylindrical structure about 4 cm long and 400 ftm in diameter, mainly by activity at a dome-shaped growing tip at the opposite end of the cell from its single nucleus. At the culmination of its life cycle, the cell produces a reproductive structure at the growing tip, the ‘cap’, which consists of about 70 compartments (‘rays’) joined together and radiating spokewise to form a circular structure about 7 mm in diameter. This can be turned upwards or downwards from its junction with the main stalk of the cell, to form a cap that is cup-shaped or umbrella-shaped. Within this range, one can find many samples that are more or less flat and suitable for twodimensional imaging without slice selection.

The main stalk consists of a cell wall at least 5 μm thick (mannan) and from 20-30 μm thick at the bottom or rhizoid end, which contains the nucleus and is furnished with rootlike processes that, in the wild, attach the cell to the sea floor. Immediately within the wall, the cytoplasm is in a layer that is also only 5-10 μm thick, arranged annularly around the central vacuole, which takes up most of the volume of the cell. The cytoplasm contains a total of about a million spherical chloroplasts, of about 1 gm in diameter, easily visible separately at high magnification but giving the cell a uniformly green appearance at low magnification. The same arrangement of cytoplasm and vacuole is found in each cap ray (for general accounts of Acetabularia, see Gibor, 1966; Puiseux-Dao, 1970; Berger et al. 1987).

When the cap is fully formed in respect of its external morphology, the nucleus at the other end of the cell divides and numerous daughter nuclei travel up the stalk by cytoplasmic streaming and enter the cap rays. There, they become arranged in a fairly orderly array, essentially a hexagonal packing, being visible under the optical microscope first as very small colourless spots amid the general green colour. This ordering is probably guided cytoskeletally. A very large number of microtubules appear in the cap ray cytoplasm at that time (Woodcock, 1971; Menzel, 1986). The living contents of the cell then gather into bundles around each nucleus, become separated from each other by plasma membranes and are finally encysted in heavy cell walls (cellulose). During this sequence the appearance near to each nucleus changes from light to dark, and the mature cysts appear black by transmission at low magnifications, because of the high concentration of organelles within them. During the maturation process, the vacuole is expelled from the cap rays as the daughter nuclei enter and form an orderly array, but aqueous solution later re-enters to surround the bundles of cysts.

Two features of the Acetabularia cap make it a particularly interesting object with which to test the capabilities of proton magnetic resonance imaging at high spatial resolution: first, the taper of the rays from hundreds to tens of micrometres, as a test pattern for resolution; second, the changes during maturation in respect of macromolecule-associated water versus ‘freE′ (including vacuolar) water. The biologist will judge genuine resolution from what features can be seen, not from calculated pixel volume; and will judge the value of ‘bound water’ information from relaxation data if it leads towards knowledge of the nature of the binding molecules.

Stock cultures of A. mediterránea, originally from the Bay of Naples, have been maintained through the 1970s and 1980s in the laboratories of B. R. Green and in this laboratory, according to published methods (Green, 1977; Shephard, 1970), in artificial sea water (ShepharD′s medium), in growth chambers at 18-22°C lit at 300-500 ft-candle (1 ft-candle ≈10 ·76 l ×) on a 12h: 12h, light/dark, cycle. For imaging, fairly flat caps of not more than 7 mm diameter were selected, severed from the stalks of the cells, and kept in ShepharD′s medium for up to a few days when necessary. For imaging, a cap was inserted into the nuclear magnetic resonance (NMR) spectrometer in a flat-bottomed sample holder (made of delrin, a hard plastic amenable to precise machining), closed to prevent escape of water vapour. Acetabularia inhabits the immediate subtidal zone and is not adapted to resist removal from sea water for long periods. In saturated air, caps do not deteriorate significantly in a few hours.

The imaging spectrometer used an Oxford Instruments 54 mm bore superconducting magnet giving a field of 6 ·3 T, 9 mm saddle-shaped transmitter-receiver coil for the 270 mHz proton resonance radiation at that field, giving a 90° pulse length of 14 fis, a Nicolet 1180 computer and a Nicolet 293B pulse programmer. A probe with a Maxwell coil for the s-gradient (not used in this work) and Golay coils (gradients 4 ·8G (gauss) cm⁻¹ Å⁻¹; 1 gauss = 10⁻⁴T); for x and _v was designed and built here. See Hall et al. (1986a) for description, eddy current problems, etc. Gradients of 15 G cm⁻¹ were commonly used, and up to 60 G cm⁻¹ in a few experiments. The normal imaging procedure was Fourier transform spinecho (‘spin warp’) method, Edelstein et al. 1980). Gradients are G_y, frequency encoding, and G_x, phase encoding. Both were applied square profile, and off during 90° and 180° pulses. Two sequences were used, as follows: I (in most experiments; Figs 1, 2, 3): 90°-G_x, G_y together-180°-G_y; II (early experiments; Fig. 4): 90°-G_y-180°-G_x-G_y.

Our imaging experiments used from 1536—2560 repeated applications of the pulse sequence. Total duration of the experiment is determined chiefly by the ‘relaxation delay’ permitted for the sample to return to equilibrium magnetization between successive pulse sequences, because this delay (usually 1 ·5 ± 0 ·5 s in our experiments) is much longer than the pulse sequence (≈4ms). In our procedure, acquisition of spatial information requires the use of a large number of increments of one of the gradients (G_x). For each value used, information is being acquired from the whole sample. Nevertheless, to judge the signal/noise problem in relation to sample size, one may think of the procedure as if a narrow strip were being scanned for each value of the x gradient, and hence arrive at a number of repeat scans needed at each gradient increment to give sufficient improvement in the signal/noise ratio. For optimization of spatial resolution in a biological sample that is not geometrically static, the desirability of increasing the number of gradient increments and the number of repeated scans at each increment must be balanced against the possible changes and movements within the sample. In most of our experiments, 128 increments of G_x were used with from 12-20 scans at each G_x value, giving a total imaging time of the order of 40 min. For display, data were zero-filled to permit use of a 256 × 256 pixel screen output. For a 9 mm field of view, pixel size is then 35 μm in both x and y directions.

Four methods were used to try to improve contrast between different kinds of water in the sample: T₁, T₂, D₂O and Mn²⁺ contrasting. All except T₁ contrasting involved the same spin-echo pulse sequences used in ‘normal’ imaging. The amplitude of the spin-echo depends on echo time t (from 90° pulse to acquisition of echo, which is twice the time between 90° and 180° pulses) as exp(—t/T₂), where T₂ is the transverse or spin-spin relaxation time of the protons. T₂ contrasting was studied by varying /. D₂O and Mn²⁺ contrasting were both attempts to suppress the signal from aqueous solution surrounding the cysts of mature caps, by chemically altering that solution in such a way that, it was hoped, the change would not penetrate into the cysts during the time of treatment. For D₂O-contrasting, the caps were immersed for 5 min in ShepharD′s medium made with D₂O in place of all the usual H₂O. For Mn²⁺-contrasting, the caps were immersed for 1 h in 2mM-MnCl₂. For the D₂O work, the sample was mounted very quickly in the magnet and only four repeat scans were used, for a total time of about 20 min after removal from the D₂O.

An intense signal (highlight in an image) may come from water with very long or very short T₁. To resolve this ambiguity by acquiring the sign of M_z is very difficult in spin-echo technique and when any pixel may contain water with several different T₁ values. We therefore did a few experiments by a different T₁ contrast method, saturation-recovery, which highlights only short T₁. This uses the normal spin-warp sequences, but with relaxation delays between repeats too short for full relaxation of protons with long T₁. The energy levels of these protons become saturated (upper and lower state equally populated) and give little signal. This does not give images as good as those from the inversion-recovery method, but aids interpretation of the latter.

A perfectly formed Acetabularia cap, though aesthetically pleasing, is not the ideal test pattern. One can count total rays to ensure that all are resolved, but one cannot distinguish small-scale details from each other. Spatial resolution is more clearly tested when a sample has some small features of unique shape, so that one can state unequivocally whether an individual feature can be seen in the image. Therefore, to test resolution, samples lacking geometrical perfection were selected (Fig. 1A and C).

Fig. IB and D are the corresponding images obtained by using a spin-warp procedure with a short echo time. This minimizes T₂ contrasting. There may be a small T₁ contrast effect, because the repetition time between pulse sequences (2 or 3s) is two or three times T₁and 5 T₁ is generally regarded as necessary for complete elimination of T₁ saturation-contrast. With this reservation, the image may be considered as basically a colour-coded map of area density of protons in the xy plane. There is no slice selection in the z direction. The sample is effectively projected into the xy plane.

In Fig. IB and D, the cap rays appear to be resolved from each other down to about 50 gm. There is some indication of better resolution for immature rays (most of C and D, and two rays at ‘9 o’clock’ in A and B) than for mature rays (most of A and B). For those latter, the bundles of cysts within the rays are in no way distinguished from the rest of their water content.

For T₁ contrast studies, caps were selected for maximum diversity in their cytoplasmic content, i.e. with an assortment of mature, immature and partly mature rays (photomicrograph, Fig. 2A). This permits a direct demonstration, within one image, of different sensitivities for different types of biologically associated water; but the types of water are not very precisely characterized. The selection procedure carries a danger that the caps, being chosen for maximum lack of synchronicity in maturation of the rays, represent slightly abnormal development. Apart from asynchronicity, however, the maturation sequence appears fairly normal in Fig. 2A. Abnormal cyst formation, which can be induced by inhibition of microtubule formation with colchicine and which we sometimes see in few cells of normal cultures, has a quite different appearance (see Fig. 16 of Woodcock, 1971).

In the sample of Fig. 2, even the normal (sequence I) spin-warp proton density image (B) shows some contrast between different types of ray, but not a very precise correlation. But the inversion-recovery T₁\ contrast images (C-F) show distinct promise for this technique in monitoring major cytoplasmic change during development. At delay times (i) of 300 and 400 ms, especially the latter (E), there is a definite correlation between rays highlighted in the image and partly mature rays as seen in the photomicrograph. High sensitivity arises when T₁ is either very much shorter on very much longer than the null value set by the delay time (400 ms delay, T₁ (null) = 577 ms). A similar cap with some partly mature rays was therefore imaged with the T₁ contrast provided by short relaxation delay in the spin-warp sequence (down to 0·25 s between repeats). The images were similar to the ones obtained by inversion-recovery, highlighting the partly mature rays and thereby showing that they have a water component with a very short T₁.

A few T₂ contrast experiments were done, by using longer echo times in the spin-warp pulse sequence. Since signal intensity varies with exp(—t/T₂), where t is the delay time from 90° pulse to acquisition of echo, all proton signals are diminished by increasing t, and signal/noise decreases unless more repeats are used, leading to a longer total imaging time. No contrasts

In visual observation, one sees a few quantitatively minor components of a living system that happen to scatter or absorb visible light rather strongly; NMR imaging ‘sees’ the numerically predominant water molecules. It is thus quite remarkable that visible-light photographs and NMR images often look so similar. In our images of Figs 1 and 2, however, there is one obvious difference. The visual contrast between the dark bundles of cysts in mature rays and the colourless solution surrounding them is absent in the NMR images. Two techniques, both involving chemical modification of the water outside the cysts while the water inside them is unaffected for a long enough time, have enabled us to image the cysts. One of these is to immerse the cap for 1 h in a 2mM solution of MnCl₂-The paramagnetic Mn²⁺ causes very rapid T₂-relaxation of water and suppresses the spin-echo signal from water in which it is dissolved. Fig. 3 shows an image obtained in this way for a mature cap.

The other technique involves replacing the solution outside the cysts with one in which all the water is replaced with D₂O, which gives no signal under the conditions of proton resonance imaging. Transfer of water across cell surfaces is usually rapid, so that shorter times were used in this experiment.

Fig. 4 shows a photomicrograph of one half of a mature cap, and an image of the same region made (by our earlier technique of separate application of G_y and G_x, i.e. pulse sequence II) after a 5 min immersion in culture medium made up using D₂O for all its water. The shapes of many bundles of cysts are clearly resolved. As the bundles taper radially inwards and become ‘dotted lines’ of cysts, it can be seen clearly where they become lost in noise level. The cap used in this study was a healthy one. When a similar cap was used repeatedly for imaging work, the same D₂O treatment caused the image to vanish completely. Evidently, features of the physiological health of a cell, which are not visually apparent, can affect transport rates of water so that they can be monitored by D₂O contrast imaging.

For imaging technology as it has so far been envisaged through the 1980s, the practical limit of spatial resolution has generally been considered to be about 10 μm (Mansfield & Morris, 1982; Connelly et al. 1987). Pixel sizes of 10-20 pm in-plane (but with greater slice thickness in the z direction) have been used in a few reported high-resolution imaging experiments (Eccles & Callaghan, 1986; Aguayo et al. 1986). The 10 pm limit is not based on an obvious fundamental criterion analogous to the wavelength of light as an absolute limit for optical microscopy. The limiting criterion is more subtle; the signal/noise ratio in an r.f. (radio frequency) coil, the primary receiver of information in NMR. Various secondary criteria, as discussed below, must ultimately be related to r.f. coil properties to find the unsurpassable limit.

Pixel size might seem to be an arbitrary number, set to ensure that the display does not lose anything that is present in the data set, in the same way that one tends to carry through a calculation one or two more figures than one expects to be significant in the final quantity. In most reported work, however, the pixel size is set by optimization considerations such as those mentioned in Materials and methods, and therefore has some real significance as a resolution limit. Nevertheless, for achieved resolution of morphological features of a biological sample, the 50 pm shown in our images is at least the equal of any previously published images. The thickness of a ray roughly equals its width at the same radius.

Δx-_T;-represents the spatial uncertainty associated with the ‘natural’ linewidth 2/T₂ in angular frequency (rads⁻¹). T₂ is the spin-spin relaxation time from spin-echo with homogeneous B_o. A few measurements onAcetabularia caps gave the order of magnitude of T₂ as 25 ms, in conformity with values measured in other biological systems, usually some tens of milliseconds (Abetsedarskaya et al. 1968; Walter & Hope, 1971; Burnell et al. 1981; Eccles & Callaghan 1986, indicating 10 ms for plant stems versus 50 ms for animal tissues; Connelly et al. 1987, <10 ms for germinating seeds; Menon et al. 1987). Our value corresponds to a spatial uncertainty Δx_T=2/γ=GT₂=2/(2·6753×10⁴)(15)(25×10⁻³) where γ is the gyromagnetic ratio of the proton and G is the field gradient. This is not the most important limitation on spatial resolution. It is only 4% of the achieved limit in our images, and could potentially be reduced further by using higher gradients.

ΔX_B arises from spatial inhomogeneity of the main magnetic field B_o. It depends on how meticulously the shimming of the magnet is done, and can also be reduced to a value that is a very small part of the currently achieved resolution.

Δxs, with S standing for either susceptibility or surface, is also an inhomogeneity effect, arising from the change in magnetic field caused by changes in magnetic susceptibility wherever the main field crosses an interface between two materials, especially between air and a condensed phase. If the change in field were AB/B_O= Ip.p.m. (parts per million) or 0 ·063 G for our B_o = 6 ·3 T = 63,000 G, then at a gradient G = 15 G cm⁻¹ the corresponding resolution limit is Δxs =ΔB/G = 42μm. At an air-water interface, ΔB/B_O could be as much as 6p.p.m. (Hedges, 1984) and resolution could therefore be as poor as 250 μm from this effect alone. Our Acetabularia samples are a severe test of this effect because, unlike long unsectioned plant stems arranged parallel to B_o, the flat caps have air/water interfaces in the imaging region. Clearly, the resolution is not nearly as bad as the worst limit just mentioned. But it is also very likely that this effect is the main one limiting resolution to 50 μm, and that substantial improvement would require the absence of air/water interfaces in the imaging region. On a linewidth basis, 1 p.p.m. is 270 Hz for our 270 MHz proton resonance frequency. We obtained NMR spectra for both whole Acetabularia caps and water samples in glass NMR tubes arranged to have air/glass/water/air interfaces in the r.f. irradiation region. In both cases, linewidths enormously greater than that of a normal water spectrum were found: l·5-2p.p.m. (400-540Hz), corresponding to spatial res-olution limit of 60-80 μm at G = 15 gauss cm⁻¹. We believe that Δxs is the greater part of this broadening.

Δx_D arises from diffusion of water in the duration (t) of a single pulse sequence, Å4 ms in this work. Diffusion (D) of precessing nuclei in a field gradient leads to a scatter of the phases of their precessions, for which Carr & Purcell (1954) derived an expression. If this scatter were translated directly into an error distribution of initial positions of the nuclei, it would give a variance of (2/3)Dt. For resolution of two overlapping Gaussian distributions, they need to be separated by about three times the square root of the variance, i.e. (6Dt)^1/2. Two discussions of ΔX, however (Mansfield & Morris, 1982; Callaghan & Eccles, 1988), have given expressions that do not contain the time t, but contain the gradient G. The expression given by Callaghan & Eccles is (1-2JT/3*) (D/σGx)^1/2, and is based on a thorough analysis of how the Fourier transform technique perceives the scatter of phases. The Mansfield & Morris expression, based on a simple comparison of echo amplitude dependence on diffusion and on T₂ relaxation, is of the same functional form with a different numerical constant. In these two expressions, time (which would seem to be of the essence in a diffusion effect) has disappeared because optimization considerations link it to gradient.

As determined by NMR, cytoplasmic water commonly has a diffusivity, D, of about 50-65% of that of pure water (i.e. about 1·25×10⁻⁵ to 1·7 ×10⁻⁵cm²s⁻¹ at 25°C) (Walter & Hope, 1971; Clark et al. 1982). Out-samples probably contain both cytoplasmic and ‘freE′ water, so thatD = 2×10⁻⁵ cm²s⁻¹ should be a reasonable average. Put into the expression (6Dt)^1/2, this yields Δ,x_D = 7μm. Put into the expression of Callaghan & Eccles, with G_x = 15 G cm⁻¹, and magnetogyric ratio y for a proton, it yields Δv_D = 9·6μm. The two estimates could, in principle, be widely different. But it seems likely that optimization of the procedure will usually fix G so that both are similar. The expression of Callaghan & Eccles arises from the most detailed analysis of what is being done, technically, in the imaging spectrometer.

From these estimates, diffusion seems, in the present work, to be the second most important limiter of spatial resolution, accounting for about one quarter of the resolution limit.

Δx_F, where F stands for flow, represents the effects of all motions of or within the sample that are macroscopically concerted motions as distinct from diffusion. These could include movement of the whole sample on its support and distortions of its external shape if its condition deteriorates during imaging, which are difficult to estimate quantitatively. But in healthy Acetabularia cells, there is very rapid cytoplasmic streaming, which carries with it intracellular structures such as chloroplasts (diameter about 2μm, but they can move together in clusters of up to 50) and, during cap maturation, secondary nuclei (diameter about 4pn), at a distribution of velocities: 3-11 μms⁻¹, averaging about 5μms⁻¹. Koop & Kiermayer (1980) give detailed information on streaming rates of various structures, with a number of sequences of optical micrographs of strips of about 30 μm width at intervals of a few seconds, illustrating these movements quite dramatically. Most of the structures shown are smaller than our NMR imaging resolution limit; and the streaming is probably much slower, if not arrested altogether, in the severed caps that we image. In our double Fourier transform imaging procedure, information is acquired on every part of the sample throughout the 40 min duration of the experiment. A resolvable object moving at 5 μms⁻¹ would be imaged as a 12mm streak. Even if a procedure were devised to image a strip one pixel wide (say, 30 μm, the same width as the strips in Koop & Kiermayer’s pictures), for instrumental parameters otherwise similar to ours the complete procedure would take about 20s (i.e. 40min/128), and movements of 100μm could have occurred in that time. This streaming is among the fastest in nature. In many kinds of cells, streaming either does not occur or goes much more slowly. The time-scale of such movements would, however, become a problem in a wide variety of cell types if attempts to improve resolution limit below 10 μm were made by using increasing total imaging times. For work in vivo, microscopic scale requires rapid data acquisition.

In summary, where the achieved resolution is A.v = 50μm, we believe that A.Vg & 30um and Δx_D—-10 urn are the two major contributions. A.vg could potentially be made smaller by using higher gradients G for the same main field B_o. Δx_D could be improved only by devising pulse sequences with shorter total time. If such improvements led to resolution better than 30 μm, pixel size would become limiting for the imaging of a whole 7 mm sample in a reasonable total time. As in the optical microscope, to achieve very high resolution one must accept a smaller field of view than the whole sample. This can raise difficulties related to r.f. source, bandwidth, etc.

If T₂(bound) < 10 ms and T₂(free) Ål·5 s, the second term on the right of equation (3) is negligible, and one may approximate the fraction of bound water as T₂(bound)T₂(obs). A few measurements of T₂(obs), which we made on caps from immature through two intermediate (but not well characterized) stages of partial maturity to fully mature, gave the sequence: T₂(obs) changes from 39 through 30 through 25 to 23 ms during maturation. If, for example, we suppose that T₂(bound) Å5 ms, then the fraction f of bound water changes from 0·13 through 0·47 through 0·20 to 0’22. These values may be scaled up or down in direct proportion to the assumed T₂(bound). For a detailed analysis of T₂ during the drying of wood, see Menon et al. (1987). Since Acetabularia caps have a heavy mannan cell wall and during maturation a large quantity of cellulose is synthesized for the walls of the cysts, much of the bound water is probably wall-associated, and a comparison with wood is likely to be quite useful. The T₂ of this water is given as 7-4 ms for western red cedar (Thuja plica ta) and 2−9 ms for Douglas fir (Pseudotsuga menziesii). The fraction (1—f) represents the total of vacuolar and cytoplasmic free water. These might have somewhat different T₂ values; but both should be in the range l-5-2s typical of oxygenated solutions chiefly of diamagnetic ions, and therefore both would approximate out of the above calculation.

NMR relaxations are often multiexponentials showing at least three components, e.g. three T₂ components as found by Menon et al. (1987) in wood. Our 7’₂ measurements are of a preliminary nature, and are not clearly resolved into components. Images T₂-contrasted by using increasing times in the spin-echo pulse sequences gave increasingly low signal/noise, because all components decay more at longer times, and did not show any contrasts markedly different from those of the normal sequence with about 4 ms from the 90° pulse to the echo signal acquisition.

It has been pointed out by Berendsen (1975) that ‘bound water’ or ‘water of hydration’ may be a term having connotations of equilibrium properties, dynamic properties involving slowing-down of molecular motions, or structural aspects of the geometrical arrangement of water molecules. (These categories are the three main divisions of physical chemistry, which one of us has stressed as being the three main categories of mechanisms for biological development (Harrison, 1987).) Corresponding to this diversity of meaning of bound, various experimental methods can give widely disparate values for the fraction of bound water. Clark et al. (1982) found that, for fresh barnacle muscle fibres, diffusion data from NMR give the bound water as 0·65 g/g macromolecules, while relaxation (T₁) data give 0·07g water/g macromolecules, or about 2% of the total intracellular water bound (Burnell et al. 1981). The latter data illustrate that T₁ relaxation can pick up interesting behaviour of quantitatively minor fractions of the total water in a way that most other methods cannot.

This is most noticeable in the images produced by T₁ contrast procedures. Among our images, highlighting attributable to a small fraction of short-water appearing transiently during cytoplasmic reorganization is seen in Fig. 2D, E and F. Correlation of the rays that image most brightly with the partly mature ones is clearest for the rays between ‘9 o’clock’ and ‘12 o’clock’, but is quite good for the whole specimen.

During maturation of the cap, various new binding surfaces for water appear in the cytoplasm of the rays: the materials of the daughter nuclei, plasma membranes (and eventually walls) of the cysts, and microtubules. The last-mentioned are the only new structures that probably reach a maximum concentration at an intermediate stage of maturation and do not persist to the end of the process. Our data give an interesting indication that we may possibly be picking up the effect of these structures on the T\ of water bound to them. But the data must be regarded as preliminary. A clear confirmation of this effect requires T₁ measurements on caps with well-synchronized maturation of all the rays, so that the T) components of one maturation stage can be clearly established. And the spectrometer must be well adapted to the detection of short-T; components. We have attempted a few 7\ measurements, but the maturation stages were not clearly established and the data were not clearly resolvable into components. Mean values of T₁ (single exponential fit) gave values decreasing from immature through partly mature to mature caps: 1460, 1230 and 1030, respectively (averages of 3-5 samples in each case).

In very general terms, the contrast between T₁ of the order of 1 s and T₂ of tens of milliseconds is well understood for restricted rotational motion in water. Bloembergen et al. 1948 (see also Poole & Farach, 1971) gave a basic theory of relaxation times in liquids with isotropically tumbling molecules. According to this, T₁ and T₂ should have similar values of the order of several seconds in pure deoxygenated water. Experimentally, T₁= T₂ = 3·6s; and for oxygenated water with no other paramagnetic solutes, they remain equal at lower values, of order 1 ·5-2 s. If the rotational motion is restricted (but still isotropic), T₁ and T₂ at first both decrease together, but beyond a certain degree of restriction T₂ continues to fall but T₁ increases. In terms of rotational correlation time (3 ×10”⁻¹² s for pure water), the T₁ minimum should be at a correlation time dependent on precession frequency. For the 40-60 MHz machines in use at the time of the account of Bloembergen et al., the minimum should be at a correlation time of the order of 10^—8 s; for recent 270-500 MHz machines, it should be between 10⁻⁹ and 10⁻¹°s.

Even for a two-state bound-free model, the relaxation behaviour of water in living systems involves the complications of anisotropic tumbling in the bound state, and exchange between bound and free water. See the account of Burnell et al. (1981) for an analysis of relaxation data on the barnacle muscle: rotational correlation time was found to be about 10⁻⁸ s for bound water, and exchange correlation time about 10⁻⁵s.

Cytoplasmic reorganization during development clearly involves the appearance or disappearance of extensive binding surfaces for water, and the complication of more than one kind of bound water. Nevertheless, we believe that T₁ and T₂ contrast imaging, and the measurement of components of T\ and T₂, show distinct promise for the monitoring of such developmental changes in vivo by NMR where the spatial scale is accessible to current technology. The maturing Acetabularia cap remains an attractive sample for more extensive studies of this kind. If changes in relaxation times are indeed related to the transient appearance of large numbers of microtubules, the method could prove useful in meristematic regions of higher plants. The large size of many plant cells makes them attractive for applications at the currently achievable scale of spatial resolution. At lower resolution, and with slice selection, wood has given images in which the annual rings are very clear (Hall et al. 1986b). It should be rewarding material to study on the cellular scale by NMR imaging.

We thank Dr B. R. Green for the Acetabularia cultures; Mr E. Matter and Mr T. Marcus for extensive technical work on the imaging spectrometer; Drs N. Burlinson, E. E. Burnell and P. T. Callaghan for helpful comments on the manuscript; and NSERC Canada for financial support in the form of: an Operating Grant to L.G.H., the same and an Infrastructure Grant to L.D.H., and a Graduate Scholarship to S.D.L. Both NSERC and MRC of Canada are particularly thanked by L.D.H. for support in the “New Research Ideas” category.