On-chip culture system for observation of isolated individual cells

Abstract

To investigate the properties of isolated single cells with their environment, we developed the differential analysis method for single cells using an on-chip microculture system. The advantages of the system are, (i) continuous cultivation of a series of isolated single cells or a group of cells under contamination free conditions, (ii) continuous observation and comparison of those cells with 0.2 μm spatial resolution by a phase-contrast/fluorescent microscopy system with digital image processing. The core of the system is an n × n (n = 20–50) array of chambers, where each is 20–70 μm in diameter and 5–30 μm deep holes etched into a biotin-coated 0.17 mm thick glass slide. The biotin-coated glass slide is covered with the streptavidin coated cellulose semipermeable membrane, which is fixed on the surface of the glass slide by streptavidin–biotin attachment, separating those holes from the nutrient medium circulating through a ‘cover chamber’ above. A single cell or group of cells can thus be isolated from environment perfused with the same medium, and the medium in each chamber can be changed within the diffusion time (<1/30 s). In addition, the microchamber volumes of specific cells or cell groups can be controlled by the sizes of the chambers. By using this system we found that the length of isolated Escherichia coli increased at 0.06 μm min⁻¹ between cell divisions regardless of the chamber volume, and that the cell concentration reached 10¹² cells ml⁻¹ under contamination free conditions. The system is thus particularly useful for one cell level analysis because the direct descendants of single cells can be cultured and compared in the isolated microchambers, and the physical properties of the cells in each microchamber can be continuously observed and compared.

---

Introduction

There are two different types of analysis methods for understanding the biological phenomena in cells. One is the batch culture based analysis, and the other is single cell level direct observation. In the former case, for understanding the phenomenon related to cell cycle, we need to obtain synchronously growing or dividing cells. Using the synchrony method of batch cultivation, the time-course of cell properties such as cell volume increase and DNA synthesis during the division cycle has been studied.^1–4 In the synchrony method, any results obtained in the first cycle could be due to a perturbation of the cells, which is the most common and most often overlooked problem with synchronized cells. The first generation results are commonly not reproducible in the second and succeeding generations without any stimulation such as heat shock method.

In the latter case, controlling the spatial distribution and interaction of each cell is most important in understanding the mechanism and function of each cell in detail. For example, in studies on the nervous system, the former type of study gives us no information about the network. Only the latter type of study shows us information on the cell-to-cell interaction. For time-course observation of cells by the optical microscopy system, we usually need to tether cells on the surface of substrates such as a glass slide or agar plate.^5–7 Recent works on topographic control of cells such as cell orientation, rates of movement, and physical contact of cells were achieved by microfabrication techniques.⁸ The effect on cells by the microstructures such as cliffs or grooves has been studied for understanding the orientation of cells.^9–11 Those results showed the potential of microstructures for cell position control, but they also indicate that cells attached on the surface of microstructures can move and pass through even when the height of walls is higher than 25 μm. We also have to mention that those topological techniques cannot control the swimming cells and cannot prevent contamination from the environment by themselves. Biological attachments such as biotin–avidin or antibodies were also applied for catching particular cells or proteins on the surface of structures.^12,13 By applying recent advances in microchip technology, several methods for handling cells in glass microchambers have been developed, such as those used to separate white blood cells from whole blood,¹⁴ and collecting cells and analyzing their DNA/mRNA content.¹⁵ Although microfabrication technology is one possible solution to the above problem, single-cell cultivation methods using microchips have not been exploited yet.

We therefore developed and examined an on-chip culture system for cell analysis.¹⁶ The advantages of the system are: (i) isolation of single cells or groups of cells from the external environment using a semipermeable membrane; (ii) 0.2 μm resolution for phase-contrast/fluorescent real time continuous observation by 0.2 mm thick glass slide; (iii) flexible medium buffer exchange; and (iv) flexible cultivation chamber size. We also tried to apply biological attachments such as streptavidin–biotin for the sealing of microchambers, the advantage of which is there is no need for any heating or any pressure to enable contact to be made between them. In this paper we examine the properties of this on-chip culture system by isolated Escherichia coli cultivation.

Materials and methods

Streptavidin decoration on a cellulose semipermeable membrane

First, a cellulose semipermeable membrane (Spectora/Por M_r 25000, Spectrum Medical Industries Inc., USA) was washed using pure water and was cut in proper sizes adequate for sealing the glass slide. Next, the washed membrane was incubated in 0.2 M NaIO₄ for 6 h to generate aldehyde functions by oxidation of cis-vicinal hydroxy groups of agarose. As aldehydic functions react at pH 4–6 with primary amines to form Schiff bases, the membrane was washed in 0.1 M phosphate buffer (pH 6.0) twice and 15 μg ml⁻¹ streptavidin hydrazide (Pierce, IL, USA) added. After 2 h incubation, the membrane was washed with water and stored at 4 °C.

Biotin coat on the surface of a microchamber array glass slide

A 0.17 mm thick glass slide was sonicated in 1 M NaOH aqueous solution, and was washed with water to clean the surface. After cleaning, the dried glass slide was coated with 100 nm thick chromium by sputtering, followed by spinning with a posi g-line photoresist, OFPR-800 (Tokyo Ohka, Ltd., Kawasaki, Japan). After baking at 85 °C for 15 min, lithography of the microchamber array patterns was carried out on a contact aligner with a broadband near-UV source (G, H, and I lines). The exposed glass slide was then developed and dried again. The exposed region of glass slide was etched using MPM-E30 solution (Intec Inc., Tokyo, Japan) for chromium, and then by HF (4.7% (w/v)), NH₄F (36.2% (w/v)) solution (etching velocity = 60 nm s⁻¹ at 25 °C) for glass. When the shape of the microchambers reached the desired size, the glass slide was washed with water to stop the etching, and dried again.

The dried microchamber array glass slide was next dipped in the solution containing 1% 3-(2-aminoethylaminopropyl) aq. at 25 °C to decorate the amino group on the surface of the glass slide. The amino-coated glass slide was dried for 30 min at 120 °C. Next, in order to decorate the biotin on the surface of the glass slide, 1 mg ml⁻¹ EZ-Link NHS-LC-Biotin (PIERCE) aq. was dipped on the surface of the amino-glass slide and incubated for 60 min. The biotin-coated glass slide was washed with water, dried at room temperature and stocked.

On-chip culture system

Fig. 1 shows a diagram of the on-chip culture system used to observe bacteria under cultivation. The system consists of 3 parts: a microchamber array plate, cover chamber, and phase-contrast/fluorescent video microscope with the digital image processor.

Fig. 1 Schematic drawing of the on-chip culture system. The cells are cultivated in individual microchambers etched on the microchamber array plate (Fig. 2). Phase-contrast/fluorescent microscopy, in which light beams with two different wavelengths are used simultaneously, is used to observe single cells in the microchambers. A fresh medium is supplied into the cover chamber, which covers the microchamber array plate, to add nutrients into the microchambers. The flow velocity of the medium can be controlled by a pump, with a range of 0.14 to 10 ml min⁻¹.

As shown in Fig. 2(a), the microchamber array plate includes an n × n (n = 20–50) array of chambers, where each chamber is 20–70 μm in diameter and 5–30 μm deep. The biotin-coated microchamber array is covered with a streptavidin-coated cellulose semipermeable membrane (M_r ∼ 25000) separating the chambers from the nutrient medium circulating through the cover chamber. The semipermeable membrane is fixed on the surface of the glass slide by using a streptavidin–biotin attachment, which is strong enough to keep the cells within the microchambers while preventing contamination from the external environment. An example of the spatial arrangement of the microchamber array is shown in Fig. 2(b), where 20 μm (diameter) × 5 μm (depth) microchambers are etched at 100 μm intervals in a 0.17 mm thick glass slide.

Fig. 2 (a) Schematic drawing of the microchamber array plate. An n × n (n = 20–50) array of microchambers is etched into a 0.17 mm thick glass slide. Each microchamber is covered with a semipermeable membrane separating the chamber from the nutrient medium circulating through the medium bath. A single cell or group of cells in the microchamber can thus be isolated from others perfused with the same medium. (b) Optical micrograph of the microchamber array plate. The arrow indicates the position of one microchamber. The bar indicates a length of 100 μm. In this case, each microchamber is 20 μm in diameter and 5 μm deep. (c) Schematic drawing of the two sizes of microchambers. The ‘small chamber’ on the left has a 20 μm diameter and 5 μm depth (shown in (b)), and the ‘large chamber’ on the right has a 70 μm diameter and 30 μm depth. Optical micrographs of (d) a small chamber, and (e) a large chamber. The arrows indicate the positions of E. coli in the chambers. The bar, 10 μm.

The cover chamber is a cubical glass box filled with a medium buffer, and it is attached to the microchamber array plate to allow the buffer in the microchambers to be exchanged through the semipermeable membrane. The volume of the cover chamber is 1 ml, and the buffer can be exchanged within 10 s by using the maximum flow speed of 10 ml min⁻¹, which is used only for changing the medium conditions. The temperature of the medium buffer is maintained at 37 °C by using a Peltier temperature controller. The temperature of the stage is also controlled by using a Peltier temperature control unit. The medium can be changed easily at any time during culturing by choosing stock solutions with different chemical and nutrient concentrations.

Phase-contrast/fluorescent microscopy (IX-70; with an oil-immersion objective lens, ×100, NA = 1.35; Olympus, Tokyo, Japan) was used to study the growth and division of the cells. The phase-contrast images and fluorescent images were acquired simultaneously by using a charge-coupled device (CCD) camera (Olympus, Tokyo, Japan). The images of cells were recorded by a VCR and analyzed using a video capture system on a personal computer. The spatial resolution of the images was 0.2 μm when the ×100 objective lens was used, which was achieved by the digital image processing of deconvolution method.¹⁷

Bacterial assays and growth conditions

We used E. coli strain JM109 (endA1, recA1, gyrA96, thi, hsdR17(r_k⁻,m_k⁺), relA1, supE44, λ⁻, Δ(lac-proAB), [F’, traD36, proAB, lacI^qZΔM15]; obtained from Toyobo, Tokyo, Japan), which was transformed with the plasmid pGFP (plasmid of green fluorescent protein (GFP) containing an ampicillin resistance region and pUC origin; Clontech Laboratories, Inc., California, USA),¹⁸ in these experiments.

Before the one-cell cultivation experiment, all cells were grown overnight in an orbital shaker at a speed of 250 rpm and temperature of 37 °C with 100 ml of M9LB medium, which consists of a minimal medium (M9: 4.5 g l⁻¹ KH₂PO₄, 10.5 g l⁻¹ K₂HPO₄, 50 mg l⁻¹ MgSO₄·7H₂O; pH 7.1) containing 1% (v/v) Luria–Bertani medium (LB) and 100 mg l⁻¹ ampicillin.¹⁹ When the cultured cells reached the stationary phase (>10⁹ cells ml⁻¹), they were fractioned in 500 μl Eppendorf tubes and maintained at −80 °C as a glycerol stock solution (50% v/v). In each experiment, we activated the stocked cells at 37 °C in the M9LB medium, using the orbital shaker again at 250 rpm. After 20 h of cultivation, the cultivated cells reached the stationary phase again and were diluted to 10⁷ cells ml⁻¹ and cultured one more time until the number of cells reached 10⁸ cells ml⁻¹, at which they were used as the sample for one-cell cultivation. In this cultivation, the number of cells was measured by using a counting glass slide (Erma, Tokyo, Japan) using optical microscopy.

Results and discussion

Single-cell cultivation in microchambers

For single-cell cultivation, prepared E. coli cells were spread on the microchamber array plate and sealed in with the streptavidin-coated cellulose semipermeable membrane. Fig. 3 shows the cell number in each microchamber after the lid was attached. As shown in the figure, about 300 (3/4 of total) microchambers were vacant and 50 (1/8) chambers were packed with isolated single cells. Cells entrapped in non-well regions were sandwiched between the glass slide and membrane, and were no longer moved or grown. When the concentration of cells in the dropping buffer was higher than 10⁹ cells ml⁻¹, the batch culture of cells stopped growing as the cells entered the stationary phase. On the other hand, when the concentration was less than 10⁷ cells ml⁻¹, cells were in the lag phase or early log-phase, before any cell-number increase or just their increase started. Thus the concentration of E. coli, 10⁷–10⁹ cells ml⁻¹, was appropriate for packing one cell into each microchamber. Concerning how the lid is attached, we usually used the following two ways; one way is that we incubate the membrane on the glass slide until they attach, the other is that we used centrifugation (×1000g) to remove excess water and attach the membrane on the glass slide. After the membrane was attached to the plate, no E. coli trapped in the microchambers could escape. The membrane-sealed microchamber array plate was then covered with the cover chamber and set on the stage of the microscope. The microchambers trapping single E. coli cells were chosen for observation. The time-course changes of the chosen isolated E. coli were continuously observed and recorded by the CCD-camera and VCR (max. 405 min recording).

Fig. 3 The number of cells in each chamber of the microchamber array (400 microchambers) after cells were spread and sealed with membrane. The concentrations of culture cells used for packing were 10⁷, 10⁸, and 10⁹ cells ml⁻¹, respectively.

After recording, the time-course growth of the single E. coli cells was analyzed using image analysis software (1/30 s and 0.2 μm resolution). We defined the length of an E. coli as the distance between its two ends (see Fig. 4), and the growth of an E. coli as its elongation rate. The length measurement includes 10–15% uncertainty, because we could only measure the two dimensional end-to-end length images of three-dimensionally tumbling E. coli.

Fig. 4 Time course of isolated single E. coli growth in a microchamber. The magnified micrographs at the top (a to g) show the time course of one of the microchambers in Fig. 2 (see arrow in Fig. 2(b)) at times of (a) 0 min, (b) 100 min, (c) 150 min, (d) 175 min, (e) 225 min, (f) 230 min, and (g) 15 h after the inoculation. The arrows in the micrographs show the positions of E. coli in the microchamber. The bar indicates a length of 10 μm. The graph at the bottom shows the time course growth of the individual E. coli.

In this system we used a 0.17 mm thick glass slide because the maximum working distance of the ×100, N.A. 1.35 object lens is less than 0.3 mm. It should be noted that, as shown in Fig. 4, no E. coli attachment on the glass surface in the microchamber wells was additionally observed during cultivation. Thus we can measure the time course change of E.coli swimming freely in the microchamber.

Observation of cell growth rate and cell division time in microchambers

We first examined the growth and division of E. coli under isolated conditions (Fig. 4). In this experiment we used the chamber with the size of the initial concentration for the isolated single cells being 1.0 × 10¹⁰ cells ml⁻¹. First, in this example, we observed that an isolated E. coli, 2.9 μm in length, maintained its length for 80 min (Fig. 4(a)). After 80 min of sleep (we call this ‘sleeping time’), the cell started growing at a rate of 0.04 μm min⁻¹, frequently bending around the septa. When the cell reached a length of 7.4 μm (160 min after inoculation), it divided into two daughter cells (Figs. 4(c) and (d)), each with a length of 3.6 μm. Following the first division, the two daughter cells started growing and tumbling again, as for the first ‘mother’ cell. The daughter cells divided again after they had grown for 60–70 min (Fig. 4(f)). The period of cell division after cell growth had begun showed exponential behavior, with a rate of 86 min cell⁻¹ division (data not shown).

This experiment demonstrated several advantages of this system. First, neither contamination from the cover chamber nor escape of E. coli from the small chamber was observed during the experiment. Second, the direct descendants of single cells could be cultured and compared in the isolated microchambers. Third, the physical properties of the cells in each microchamber could be continuously observed and compared. Finally, the resolution of cell length, 0.2 μm, was an order of magnitude smaller than that of conventional methods such as using the Coulter Counter (Coulter Electronics, Inc., Hialeah, FL, USA), which determines the sizes of cells electronically.

Growth and cell cycles of daughter cells after single-cell division

We can also compare cell growth and cell division times among isolated single cells and their daughter cells. Two daughter cells have the same DNA and cytoplasm as their mother cell, and in this experiment they grew in the same environment, including physical contact with each other. The graph in Fig. 4 is one example of such a comparison. An isolated mother cell grew at 0.04 μm min⁻¹ and divided 85 min after cell growth began, while its daughter cells grew at 0.06 μm min⁻¹ and divided after 65–70 min. In this case, although the growth rate and cell division time for the mother cell and daughter cells were different, those for the two daughter cells were almost the same. However, none of the results of eight other comparisons showed strong correlation between mother cells and daughter cells, or between daughter cells (data not shown). We also observed that mother cells sometimes divided into two daughter cells of obviously unequal lengths (data not shown).

Flexible medium buffer exchange through the semipermeable membrane

Controlling the medium condition is important for one cell analysis. The semipermeable membrane, which is attached on the microchamber array plate, exchanges the medium in microchambers within the time resolution of the VCR system, <1/30 s, which was checked by real-time confocal optical microscopy system (CSU21, Yokogawa Electric Corp., Tokyo). The result indicated no delay of diffusion in 5 μm depth with 1/30 s resolution. That is, there was no difference of increase of the fluorescent dye’s concentrations between the top and the bottom of the microchamber after the introduction of the dye into the system. Thus we concluded the diffusion time was in 1/30 s.

As shown in Fig. 4, after overnight culturing, the number of E. coli had fully increased in the microchamber (Fig. 4(g)). In this case, the concentration of E. coli corresponded to >10¹² cells ml⁻¹, which is more than a thousand times larger than the stationary-phase concentration of >10⁹ cells ml⁻¹ in batch cultivation using the same sample and medium.

We checked another use of the on-chip culture system, for changing the medium solution in the microchambers through the semipermeable membrane. Single E. coli cells were cultured in small chambers with M9LB medium. To express GFP in E. coli (JM109), we used M9LB-IPTG medium buffer, which is M9LB containing 1 mM isopropyl-β-D(−)-thiogalactopyranoside (IPTG; Wako Chemical Co., Tokyo, Japan), and 0.1 mM lactose.¹⁸ As shown in Fig. 5(a) and (b), no GFP expression was observed before IPTG induction. Forty min after the medium exchange from M9LB to M9LB-IPTG, GFP expression was successfully observed. Fig. 5(c) and (d) show the GFP fluorescence in the same E. coli 60 min after the medium exchange, which is almost the same as batch cultivation data under the same buffer condition at 37 °C.

Fig. 5 Green fluorescent protein expression in E. coli, shown before IPTG induction (a, b), and 60 min after IPTG induction began (c, d). (a) and (c) are phase contrast micrographs; (b) and (d) are fluorescent micrographs. The arrows show the positions of E. coli. The bar indicates a length of 10 μm.

These results show good system performance for solution exchange, which is difficult for batch or plate cultivation methods.

Flexible chamber size control of single cell cultivation

Using this on-chip culture system, we can easily control the environment for cultivation. For example, we can control the cultivation volume by using different sizes of microchambers. Fig. 2 shows one example of volume control by using different microchambers. The size of the chamber shown in Fig. 2(d) is 20 μm (diameter) × 5 μm (depth), which we call a ‘small chamber’, and that in Fig. 2(e) is 70 μm (diameter) × 30 μm (depth), a ‘large chamber’. The volumes of the chambers are 1.0 × 10⁻¹⁰ and 7.7 × 10⁻⁸ ml, respectively. If we culture single cells in these chambers, the estimated concentrations are 1.0 × 10¹⁰ and 1.3 × 10⁷ cells ml⁻¹, respectively.

Fig. 6 shows the chamber-size dependence of single E. coli growth. The average growth speeds of five samples were 0.06 μm min⁻¹ both in the small chamber and in the large chamber. The mean values of the cell division time were also almost the same for both chambers, around 45 min. The results showed no significant differences in mean values of growth speed and cell division time, even when the mean free paths were 10 times different.

Fig. 6 Chamber size dependence of E. coli cell growth. The circles indicate a small chamber, 20 μm in diameter and 10 μm deep (Fig. 2(d)). The squares indicate a large chamber, 70 μm in diameter and 30 μm deep (Fig. 2(e)). The plotted points are mean values of cell length at 10 min intervals, with error bars indicating the standard deviation of five cells. The arrow and bar labeled ‘a’ indicate the mean time and standard deviation of the cell division time for the small chamber; those labeled “b” indicate these values for the large chamber.

In conclusion, the on-chip culture system for analyzing single cells was examined, and several advantages of this system were demonstrated using E. coli as the example. The system enabled us to find that single E. coli enclosed in a chamber can be successfully cultivated. Such results obtained by the on-chip culture method might help to explain the cell growth and cell division mechanisms not only for single cell analysis but also for a group of cells such as nerve cells, and might show some unique usage for drug discovery studies.

Acknowledgements

We thank Professor K. Kaneko of the University of Tokyo and Professor T. Yomo of Osaka University for their valuable advice and discussion, and Professor M. Esashi and his staff at Tohoku University for their valuable advice on microfabrication techniques. This study was supported in part by Grants-in-Aids for Science Research from the Ministry of Education, Science and Culture of Japan (No.11CE2006 (Komaba Complex Systems Life Science Project) to K. K., and No.12206028 to K. Y.).

References

1. C. E. Helmstetter and D. Cummings, Proc. Natl. Acad. Sci. USA, 1963, 50, 767.
2. C. E. Helmstetter and S. Cooper, J. Mol. Biol., 1968, 31, 507.
3. H. E. Kubitschek, J. Bacteriol., 1990, 172, 94 .[*] This paper shows one example of synchronous cultivation of Escherichia coli using the velocity sedimentation method. Cell numbers and mean cell volumes of the synchronized cells were also measured electronically by the Coulter multi-channel analyzer system.
4. S. Cooper, Bacterial growth and division, Academic Press, Inc., California, USA, 1991.[*] This book is the summary of bacterial cell growth and division cycle studies from a biochemical viewpoint. It introduces both synchronization methods and nonsynchronization methods of cultured bacteria, explaining their advantages and problems. It also summarizes cytoplasm synthesis, DNA replication, and synthesis of the cell surface during bacterium division cycle.
5. J. A. Shapiro and C. Hsu, J. Bacteriol., 1989, 171, 5963 .[*] This paper showed one example of continuous microscopic observation of bacterial growth and cell divisions of E. coli attached on an agar plate. Though the mobility of bacteria was limited, microcolonies containing several hundred or more bacteria cultivated from single cells were visible until they were contaminated by other colony.
6. K. C. Neuman, E. H. Chadd, G. F. Liou, K. Bergman and S. M. Block, Biophys. J., 1999, 77, 2856.
7. W. D. Donachie and K. J. Begg, Nature, 1970, 227, 1224.
8. A. Curtis and C. Wilkinson, Biomaterials, 1997, 18, 1573 .[*] This paper is the review of literature on the reaction of cells such as orientation, rates of movement to their surrounding topography. Cells were captured by the particular feature of the structures on the surface of the plate.
9. P. Clark, P. Connolly, A. S. G. Curtis, J. A. T. Dow and C. D. W. Wilkinson, Development, 1987, 99, 439.
10. P. Clark, P. Connolly, A. S. G. Curtis, J. A. T. Dow and C. D. W. Wilkinson, J. Cell Sci., 1991, 99, 73.
11. D. M. Brunette, G. S. Kenner and T. R. L. Gould, J. Dent. Res., 1983, 62, 1045.
12. A. L. Grosvenor, A. Feltus, R. C. Conover, A. Daunert and K. W. Anderson, Anal. Chem., 2000, 72, 2590.
13. A. Tixier, Y. Mita, B. L. Pioufle, P. Surbled, Y. Murakami, E. Tamiya and A. Fujita, in Micro Total Analysis System 2000, ed. A. van den Berg, W. Olthuis and P. Bergveld, Kluwer Academic Publishers, Netherlands, pp. 123–126.[*][*] This paper showed a new non-lithographic technique for the patterning of antibody-linking surfaces, using a shadow mask technique. In this way, they immobilized cells specifically on the patterns of structures on the surface of the plate using antibodies.
14. J. Cheng, L. J. Kricka, E. L. Sheldon and P. Wilding, in Topics in Current Chemistry, ed. A. Manz and H. Becker, Springer-Verlag, Heidelberg, Germany, 1997, vol. 194, pp. 215–231.
15. J. Cheng, E. L. Sheldon, L. Wu, A. Uribe, L. O. Gerrue, J. Carrino, M. J. Heller and J. P. O’Connel, Nat. Biotech., 1998, 16, 541.
16. I. Inoue, Y. Wakamoto, H. Moriguchi, K. Onuma, A. Kashiwagi, K. Kaneko, T. Yomo and K. Yasuda, Biophys. J., 2001, 80, 166a (abstr.).
17. D. A. Agard, Annu. Rev. Biophys. Bioeng., 1984, 13, 191.
18. M. Chalfie, Y. Tu, G. Euskirchen, W. W. Ward and D. C. Prasher, Science, 1994, 263, 802.
19. T. Maniatis, E. F. Fritsch and J. Sambrook, Molecular Cloning: a Laboratory Manual, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA, 1982.

---