Oliver Smithies:

I’m looking at the first of four books from my graduate student research period within Sandy Ogston’s lab. It begins, interestingly enough, with a circuit for a thermostat. That’s quite surprising to a modern day scientist, that one would have to build one’s own thermostat to control the water bath but my first job was to make this circuit. I was quite familiar with electronic circuits, well, electronic patch is the modern word, but they were tube circuits or valve circuits, because as a child I’d been making radios for a long time. In fact there was a rather enjoyable period during WWII when I was still a teenager that we were required to carry gas masks in a cardboard box about 6 inches tall and maybe 8 inches wide and 5 inches deep. And I made a radio to fit in my gas mask box, so I didn’t have a gas mask, I had a little one tube radio in the gas mask box. But anyway so I was well familiar with circuits and this was not a surprise to me that I have to start by building a thermostat, the electronic part of a thermostat. The main part of the diagram is in ink but there are various additions on the side, because the circuit also had to be able to have the ability to control the motor for the stirrer and also a subsidiary heater, so that the main heat was on all the time, but the thermostat just controlled the additional heat needed to get to the final temperature. There are other details there. And on the second page is how I set out the circuit on a board with motor resistance and the 40 watt lamp and the tube, vacuum tube KT71, and the heater, mains, resistance, etc. it’s all set up. The note that the motor supplies on the separate switch from the main switch, but supplied by the same plug from the mains, all other supply to the thermostat passes through the main switch. Heaters and regulators have additional series switches. Then there’s a blank page where presumably I made the circuit. And that already October 11, ’48, I’m rather surprised to see that I already have osmotic measurements here. So I must have been beginning to measure osmotic pressure already. I don’t remember what particular apparatus I was using at that time, but I do know that the final instrument that I made was quite complicated, and I don’t know where notebooks are, even if they exist, that detail its construction. My thesis has those details. I think I’ll stop this recording of this book and go back to my thesis before continuing this book, because it doesn’t make much sense without having the details of what the apparatus was, so I’ll stop here.

I don’t have a very good record of the experimental part of my thesis work in the form of a notebook, but I have a good account in the thesis itself, and it’s necessary to look at that a little in order to understand what was going on. The thesis itself is divided essentially into two parts, the one part being a rather complicated set of calculations and derivations of equations to describe the osmotic pressure of solutions and correct for the interaction between protein solutions and the surrounding electrolytes, in this case sodium chloride and so forth. I know now that the theoretical section has a serious error in it, because when I tried to write this up for publication, after I had finished my thesis and was a postdoctoral fellow in the University of Wisconsin, I had a very good roommate in the laboratory, a laboratory roommate in Wisconsin, Dick Goldberg, who was a fine theoretical physical chemist. When I attempted to write up the theoretical part of my thesis, I found that I’d made an assumption somewhere in the thesis of saying a certain term was negligible, and it turned out that the term that I’d neglected was approximately equal in magnitude to the term that I  later was considering theoretically important, which basically made that all of the theoretical part of the thesis was pretty well faulty. So, nowhere is the theoretical part published in the scientific literature. The experimental part was published, because the measurements were of high precision, the measurements of osmotic pressure, that is, were of high precision, and the method was interesting, though it has a rather sad history, in the sense that although the work took two years to accomplish experimentally, and I enjoyed this very much and published it, nobody ever quoted the paper for the method and nobody ever used the method subsequently, and I never used the method subsequently. So I often tell students, here is an example of a thesis problem which has a record of never being quoted by anyone and never being used by anyone, but having taught me how to do good experimental science, and enjoying it. So that was the point of the experimental part of the thesis. So in many ways, the thesis is completely unimportant. Nonetheless, it’s interesting to look at it and see how not dedicated, but compulsive behavior of Smithies in trying to get things to several decimal places of precision when other people had them to much lower level of precision. So it is worth looking at the experimental part of the thesis first, because it makes some sense of the recorded observations which are in four experimental books, 1-4, that I’ll talk about later. So let’s go back to the thesis and look at the experimental part of the thesis.

The experimental part of the thesis begins on page 51, talking about the experimental work falls mainly into three, naturally into three main sections: preparation of lactoglobulin, the protein that was being studied, the estimation of its concentration by refractometry, and the measurement of osmotic pressure is the third section. So the first part is talking about the preparation of lactoglobulin, which also comes repeatedly in the notebooks that I have 1-4 and will talk about later. It starts off on page 52 by talking about the prep of lactoglobulin solution, by a method described by Rupert Cecil and Sandy Ogston, with a few minor modifications. Starting typically, in this case May 31, 1950, with two gallons of milk, direct from the milking sheds. Remembering that you could in Oxford, as anywhere else, you could go to a farm and they would give you milk. Straightforward, getting rid of the cream and then precipitating the casein with ammonium sulphate and so on. So starting on May 31, by June 4 I’ve got 2 to 5 ccs of oil present in the dialysis sac, because dialysis was needed to get rid of the ammonium sulphate used to precipitate the casein, and so on. And then June 5 the oily material and supernatent were removed from the dialysis sac and put into the refrigerator, and the next day the preparation was seeded with crystalline lactoglobulin, which is interesting. So I had crystalline lactoglobulin available from previous experiments of other people, of Cecil and Ogston, and I used that to seed the material that I had and the crystallization was complete the next day. So it goes on and describes that. And then solutions of lactoglobulin were made in fairly concentrated sodium chloride, 1.19 molal sodium chloride, which was a theoretical concentration that was the outcome of the error prone calculations made in the theoretical part. But it was a high concentration of salt. And I made some solutions of .5, 1, 2, 4, 6, 8, and 10%, and they were dialyzed against a buffer all at the same time.

And they were rotated in this solution. And then the proteins were freeze dried, the crystalline protein was freeze dried. Remember it had been dialyzed against distilled water, so it didn’t have any salt in it anymore. It was freeze dried by a rather complicated apparatus that is described and diagrammed.

Then going further on I did various tests on the material that I freeze dried and came to the conclusion that the freeze dried material is not sensibly altered by the drying, so I don’t know what I did to make that statement.

I was really concerned in making the protein crystals free from water, and an apparatus was used to get them really dry so I could measure their concentration. I must have known from other people’s work that it was possible to dry crystalized beta 2 lactoglobulin without it being denatured. That’s rather unusual and I do remember doing tests where after they’d been dried, I heated it to 105 degrees as a I remember to make sure that it was dry, and it still was perfectly soluble in water, in a salt solution thereafter. So it was a very stable protein.

Here I even made a special dish for drying things in. And then there’s a bit you might say, not frustrated, but didn’t like the method of dialysis so I tried to make an electro dialyzer that would take out salt by passing an electric current instead of mechanical. So I make the comment that a mechanical dialyzer was used for most of the work, designed by Ogston, which enabled the dialysis sac to be spun clockwise and anti-clockwise, therefore mixing inside while the dialysis was taking place. But for reasons that I mentioned later in the thesis, I wanted lactoglobulin completely free from salt, so I tried an electro dialyzer made out of Perspex. I like making things so I didn’t have any trouble making this apparatus to control the removal of salt by electrolysis. And I have diagrams of the thought flowing water through the machine while the electrolysis was occurring at the same time. Awkward machine but I enjoyed doing it.

And I measured the current during the procedure so I could get an idea that I’d gotten rid of all the salt, because initially it was quite a large millium per current going through the machine, but as it was washed out by the flow of liquid, the current through the machine decreased more and more and washed out so that the final solution was essentially free from any salt.

Not satisfied with the arrangement of the water circulation in the machine, I messed around with that and have several pages describing the direction/flow of the liquid in the electro dialyzer. I’m sure not really very important but it was a sort of compulsive behavior of Smithies in trying to understand the experiments or to improve them.

Then we go to the section on measuring the concentration of lactoglobulin by an optical system. This was a system that had already been described, originally by Philpot, John Philpot, who was one of Sandy’s earlier students. Or I think he might have been a contemporary of Sandy’s rather than his student. But it had been designed and made by Sandy Ogston and Rupert Cecil. It was their machine but slightly modified to enable more accurate determinations to be made when the solutions were concentrated. And so it was a fairly sophisticated machine for measuring refractive index. Because you had a very fine slit, and an image of the slit, and then with a couple of lenses the optical path through the refractometer combined again to give an image where the slit image was now a pattern that could be measured. The diagrams are fairly self-explanatory. It’s a little difficult to verbalize it, but I was convinced that it could be made very accurate, and was writing down deflections and concentrations with 4 decimal places. .3599, .3599, at different times, .3599, three measurements on page 78, and then the blank was set up and the diffraction was then .0187 twice. So I was getting things reproducible to 4 decimals. Quite nice.

For example the blank was a measurement, but in practice it was found that this value was sufficiently constant to make it unnecessary to determine the blank every time, because the variation was less than the accuracy that I was aiming for, which was plus or minus .1%. Then I have the section beginning on page 82 on the measurement of the osmotic pressure, and it just shows a general solution initially. Having an external pressure on the higher on the – I’m looking at this – I’ll backtrack on that. On page 82, two solutions are being talked about. What’s surprising me is it’s talking about a pressure on one side A and a pressure on the other side B. And so the pressure side B being the side where the protein solution is, and the side A being the side where the solvent is, the sodium chloride solution. PB would have to be greater than PA for us to oppose the osmotic pressure due to the concentration of water in the protein solution being lower than the concentration of water in the solvent. It just sets up a few little equations indicating that you also have to worry about the surface tension of the solutions, because the surface tension of water is different from the surface tension of a protein solution. The surface tension – I tried I think material to coat the tubes but it didn’t really alter the surface tension. I tried a substance called “Teddol” and silicon varnish, which is a silicon varnish but it didn’t really alter the surface tension, it didn’t reduce it to zero. Another type of tubing “Veridia.” So I have an example of Veridia tubing. So I’m concerning myself about the surface tension. In the end the solution was really quite simple. I made a very thin latex, you might say like a little condom, and attached this to the bottom of my machine and allowed it to collapse. And it was very thin so that it essentially imposed no pressure difference between the inside and the outside, and I was able to directly measure the osmotic pressure using tubes that were in the final apparatus there. Page 89 shows this machine. A collapsed impermeable rubber sac, is what I used to call it. But my fellow students didn’t refer to it in that way, they called it a rubber johnny, was their expression for it. And it required making rather nice sacs of this sort. Quite a bit of effort in making them. On page 93, the sac’s openings were beautifully clear and uniform but showed pinholes when distended with about 15 ml of water. The pinholes were removed by double coating. So coat the glass rod, dry for half an hour, coat, dry for half an hour, then steam ten minutes and wash thoroughly. And these sacs made with latex partly diluted, were a little thick. The most dilute latex which could be used was 4.5 mils of latex to 1 of water, and this gave 4 sacs free from any leaks out of 6 that were made. And you had to steam them. Quite involved in making these sacs.

Then I determined the collapsed pressure of these rubber sacs, because as I said I assumed they had no difference in pressure between the collapsed sac and a little more collapsed versus a little less collapsed. And I found that the difference in pressure on the two sides was an average of about .015 milliliters of water, which was smaller than the error that I was prepared to accept which was .1 milliliter of water, so that they didn’t have any problems. It was clear that with an allowance of approximately .015 centimeters, for the collapsed pressure of the rubber sac, this U-tube method can be used to give a direct correction for surface tension error present in the osmometer. And then the experiment of actually measuring the surface tension of the different solutions of lactoglobulin, ½, 1, 2, 6, 8, and 10% is listed. And a rather nice graph of the surface tension correction. And the maximum still being only about .2 centimeters of water, but with a good correctional variable. So I was happy with being able to correct the surface tension differences.

And then the next thing was the preparation of the semi-permeable membranes themselves. These are membranes made out of collodian, which is I think a cellulose acetate if I remember rightly, it might have been a different form of cellulose, but you can get a solution of this and then coat a glass rod or whatever, and allow it to dry and you’ll get a thin membrane basically of reconstituted cellulose, or a cellulose derivative. And so I have this statement on page 102 that the dynamic osmometer used in this work required the preparation of membranes as permeable as possible to water and sodium chloride but completely impermeable to lactoglobulin and how to make them. And I tried different methods to make them. Trying to use something on which they could be attached to the apparatus, because they would be too fragile to insert onto the machine, onto the osmometer, but have to be made in situ. I tried various things. I tried a form made out of Wood’s metal, which is a metal that melts at 70 degrees. But eventually I found out that sodium thiosulphate could be used. It’s a rather interesting substance. If you warm it, it melts in its own water of crystallization. So sodium thiosulphate, I don’t remember how many molecules of water per molecules of sodium that’s probably something like ten, when you warm it up it will dissolve in its own water of crystallization. But when it cools again it sets into a solid. And I was able to make these tubes, these thin membranes, by having a layer of  thiosulphate, and seeding the layer of  thiosulphate solution with a crystal at the bottom. So that the crystals formed beautifully uniformly, and they could be made round and pretty. And there is on page 107 a little diagram showing how to take off the plastic tube that was used to contain the thiosulfate. And it’s a thumbnail being used. My co-graduate students used to enjoy this because they called it a thumbnail sketch, and it was really a sketch of a thumbnail, on page 107 of the thesis. So thumbnail sketch really is a thumbnail sketch. Making those crystals. But they were quite nice because once made, those membranes you could easily see if they had a leak, because the sodium thiosulphate was high concentration and as you washed out the sodium thiosulphate out of the thin membrane, if there was a hole in the membrane, you could see the thiosulfate coming out by refractive index so you could observe any leaks. So it was quite possible to make very good membranes that way.

And then I have the making of a constant temperature water bath, which is actually the first page of my notebooks where the circuit diagram of the constant temperature water bath is given. So page 114 of my thesis talks about the circuit diagram of the constant temperature bath that is the beginning of, the first page of book 1 of my experimental notes. So it begins, the experimental notes, begin in a sense after all of the experimental procedure was worked out. So that’s why I have had to go over the thesis bit by bit, otherwise the machine is not understandable. And it was the dynamic osmometer means that I was using it in such a way that I could tell whether the pressure on the protein side was too high, in which case liquid would flow from the protein solution into the solution of the dialysis solution. So liquid would flow from the protein side to the solvent side, and if the pressure was too low, liquid would flow from the solvent side into the protein side. And I followed this by using a fine capillary with a little oily droplet in the capillary, which could be observed with a telescope. And one could then vary the pressure and find out what the pressure was when there was zero flow by having first a positive flow, and a negative flow, and drawing a line between them. Where it crossed zero is the actual pressure. It was highly sophisticated and in the end quite unimportant.

I know the pressure accurately that I was applying, and that was done, with attached to the system was a manometer, with a bromo chbr3 on one side, and mercury on the bottom, and the pressure could be altered by a control knob. And the pressure difference was measured by looking at the manometer. So page 128 of the thesis shows on the left hand side the osmometer itself and the middle diagram is the surface tension apparatus. On the right hand side is the diagram of the manometer itself. All these were highly precision made glass constructs which were made in those days by the glass blowers in the department. It was so important being able to make glass things that the department of biochemistry, in which this work was being done, had professionals who would make anything you wanted out of glass. So all the apparatuses were actually made by professional glass blowers. It was rather beautiful stuff. The diagram on page 128 is well worth looking at.

Incredibly detailed of using the apparatus, really quite amazing to look at to me now. The final result is shown later but an example of where the thermostat temperature was varying is given on page 142, where one can see that the applied pressure it changes for getting zero flow, because the thermostat was not set correctly. And then the real results are presented on page 145, where all the theoretical, where the osmotic pressures are measured to 2 decimal places in centimeters of water. 7.04 was the lowest osmotic pressure and the highest was 91.51. These are all recorded to 2 decimal places of precision and the refractometer measurements are again 4 decimals and maybe plus or minus .1 percent. And the data are therefore with a precision of the order of .1%. The experiment points are so close together that it’s not possible to draw a line through them without covering the points up. So the final publication had to interrupt the theoretical line to show the experimental points. I was very proud of this. Quite rightly, for precision, but as far as usefulness in science, only useful to me as good training to do experiments. So the apparatus is made and calibrated, and we can turn now to the actual record of the osmotic pressure.

So book 1 starts off with a diagram of the circuit diagram for the thermostat.  I think I’ve talked a little bit about that already in the recording. I won’t go over that again. But then from there on, there are just a series of actual measurements, though the pages are not numbered but the first page after the diagram for the thermostat circuit is measuring osmotic pressure. So I began to measure osmotic pressure on October 11, 1948, the protein level was so and such and the buffer level was so and such and from it I measure the osmotic pressure. Note, the collodian in use was made by A.G. Ogston. What was called pyroxylin 3%, ethanol 40%, ether 60%, and ethylene glycol, 3 ccs. Presumably per 100 mils. Membrane was made by a single coat, dried for 6 minutes in the vertical dessicator, tested with T.C.A. (I don’t know what T.C.A. is here), buffer showed slight protein permeability, repeat and dry the membrane for 12 minutes. So membranes dried for 12 minutes showed very small or zero protein permeability. So T.C.A. must have meant that I tested whether I put a high pressure on the protein side, collected the material going through it, and precipitated it with tri-carboxylic, with tri-chloral acetic acid, to see if there was any protein coming through. And working that way to get the membranes impermeable.

For example on November 8, I said the membrane made two days ago with collodian 1a twice was doubly coated and tested with hemoglobin preparation and it shows whether there was any permeability. So I’m making good membranes. And here’s an example: repeat the experiment with the apparatus, and a leak developed in the water bath, I don’t know how that was possible. But it says a leak developed. I’m looking at hemoglobin solutions at the moment. So just testing things out with hemoglobin rather than with lactoglobulin.

3.1% hemoglobin, doubly coated membrane and measuring what’s happening. The movement of my little droplet in the osmometer. I didn’t like the droplet I was using, it tended to be not round, but I used chlorobenzene to reduce the flattening of the droplet. I don’t remember the final one was chlorobenzene or not. I have an idea that it was something different, but it’s recorded in the paper. I tried experiments with a sac empty of protein. So I measured the osmotic pressure by having a little potassium chloride on one side of the membrane versus water on the other side of the membrane. And I have 5 centimeters of osmotic pressure in it initially, and then in four minutes it’s down to 3.5, in 12 minutes it’s down to 1.8, in 30 minutes it’s down to .5 centimeters. From this I worked out that the solution would come to equilibrium with a half-life of 8 minutes, so that if the apparatus was set up in the morning or in the afternoon and one waited until next day, the solutions would all be at equilibrium as far as salt was concerned. So this was part of the comment once made about the apparatus in the literature, which was that it was – I guess it must have been just one place that described it as not working very well – it said that the apparatus required waiting overnight to come to equilibrium which is in the thesis here, so somebody did actually comment on the method, but nobody ever used it.

Fresh hemoglobin here on one page, 9.1 centimeters of osmotic pressure. Here I am getting used to the method. And then as always throughout these notes on the experiment, every now and then there is making a new preparation of lactoglobulin. So 1-12-48, the 12th of January 1948, making/recrystallizing lactoglobulin. Buffer made for lactoglobulin, osmotic pressure made, preparation of stock lactoglobulin, and then osmotic pressures and so on. The membrane seemed a great improvement. Conditions for making the form of the membrane, of the sodium thiosulfate membranes: clean and dry the glass parts and the polythene tube. Assemble with a small crystal. Prepare iced water, heat the thiosulphate, boil off a little of the water, allow it to cool to about 80 degrees centigrade, and then pour into the polythene, place in position. Immerse the lower end in ice water and let the crystals come. Finally as the crystals approach halfway, fully immerse it in the water. A detailed description of how to make the crystals form on which the semi permeable membranes were made. With a comment on the next page that a ten minute membrane, early type pyroxylin etc. etc., used etc. and with a big comment that it does not leak. And some more making new collodian solutions to make these membranes, and so on it goes. Pages and pages of detailed work on trying to measure these osmotic pressures. Here’s one previous preparation left overnight, no sign of a leak, mixed and repeated.

A comment here: it’s quite clear that leakage if any is very slight, however, a used membrane tested at 10 centimeters of mercury with protein solution.  The filtrate gave very slightly positive tests for protein with trichloracetic acid. And I even determined the bursting pressure of the membrane was 52 centimeters of mercury. So they could stand quite a big pressure. A 20 min membrane was made and tested with 17 centimeters of mercury. And that was 52 centimeters of mercury, not of water. So that would be like 500 centimeters of water or thereabout. So a 20 minute membrane made and tested at 17 centimeters of mercury with protein, a complete absence of cloudiness, that’s trichloracetic acid cloudiness, after ultrafiltration of about .3-.5 cubic centimeters of filtrate. And here is the measurement of osmotic pressure now, 12.18 the following day.

“Close the door on the window”, I’m concerned about temperature varying. Here’s an example left over the weekend. That’s slightly too low a pressure mixed and repeated. Another lactoglobulin preparation starting with 7.2 liters of separated milk. Capillary measurements, surface tension, and so on until the end of that book. Nothing very striking. Here I’m cleaning the apparatus with various things, in this particular case, remove the silicon with aqua regia. 1-2 hours of boiling. Incredible to think of doing that. And then we dilute sodium hydroxide and water to clean the apparatus to get rid of the grease that have been introduced in previous runs. So we use extremely harsh methods to clean the apparatus. And a test of the electro dialyzer that I talked about, on the 25th of March, 1949. Conclusions that the majority of sodium chloride carried out by the current, increase the current by stepping up the voltage, etc. etc. We wanted to see there wasn’t any hyperchloride made there, so I tested for whether there was any OCL minus made. So I redesigned the flow because I didn’t like the pattern that I could see. So the system was changed. Comments on the freeze dry machine. Again, how to freeze dry the protein as it referred to in the thesis. And this book ends with dismantled and cleaned again more extensively, reassembled etc. etc. and that’s where this book ends. Book 1.