Oliver Smithies:

[00:00:00] This is book capital F’ beginning October 22nd, 1996 and extending through December of the same year.  And it will begin with continuing the assembly of the IRES internal ribosomal entry site construct that we’d been talking about earlier.  This is straightforward continuation of making DNA plasmids of a desired type and continuing on for several days.  New candidates for example on Saturday, November 11th, page 11.  And checked on them thereafter.  [00:01:00] And so it goes –  standard.  Except if one comes across comments such as this on Sunday, November 10th, page 16.  Wasted one month.  Mistake.  Exclamation mark.  Looking back, etc., etc.  Quite a common event.  A mistake.  Back to step three on the following Tuesday, November 12th.

And then by November 14th, Thursday, page 25, we’re on step six in making this construct.  And so on.  The next pages.  Step six is still going on [00:02:00] Thursday, November 21st, page 33, with new candidates the following Friday.  That must have been Thursday, November 21st.  The following day Friday new candidates.  Nothing particularly notable.  Step six going on.  Step 6A now.  November 25th  and November 27th.  Thanksgiving Day.  And then Wednesday, November 27th and Thanksgiving Day.  Continuing working.  Step eight now.  Fresh start on fragments on Monday, December 2nd, page 51.  [00:03:00] Designing linkers on Friday, December 6th, etc.  Nothing particularly remarkable.  Just bang bang bang.

Christmas Day, Wednesday, December 25th, the usual.  Still working, page 67.  Checks on.  Etc.  Was unsuccessful.  Now then here I’ve got a [00:04:00] break, the last entry being Friday, December 27th, page 69.  And then up to Monday, January 13th, a break there.  Writing GM20069 competitive renewal.  Talking about failure of animals to change blood pressure on genetic variation of ACE.  Can be modeled in the computer.  And is due largely in this model to an increase in angiotensin I or a decrease, etc., etc.  Not quite the correct interpretation I now know, but a beginning of trying to interpret the results of the genetic experiments by [00:05:00] a modeling type of process which you’ll see more of.  As it says a long gap to write the grant renewal.  Work on the computer model.  Rebuttal presumably of something to do with the grant and several papers.

Attempting here Monday, August 18th, 1997 to record the computer work and get back to the bench.  So here’s a gap from January 13th to August 18th, one of the longest gaps that I have in experimental work, which was devoted to developing computer simulations of the behavior of the renin-angiotensin system.  [00:06:00] And I’ll go over it in a little detail in discussing the following pages that are an abstract of what went on at that time.

Starting on page 73 with the heading computer modeling chronology with August 1997 comments.  So as we’ve seen earlier, August 22nd, 1996 I sent a letter to Marshall Edgell outlining requirements initially aimed at reproducing what I called a path diagram for the angiotensinogen 1/0 copy and ACE 1/0 copy animals.  This type of diagram [00:07:00] is illustrated in the letter to Marshall.  First of all on — which is attached on page 72.  The most critical part is actually the diagram of the renin-angiotensin system sketched by me.  Which starts with angiotensinogen AGT and the renin genes Ren1 and Ren2.  And ACE gene and angiotensin receptor genes and so on.  And in fact that was a diagram that I drew to make sure that I understood what was happening.  But it turned out to be quite [00:08:00] important because Marshall then tried to model that diagram in a computer simulation using programs that were available online for making simulations.  This was a type of program that was available from a company, I think somewhere in Maine.  The program was Stella.  But before I go into that, I have to also look at the second diagram that I [00:09:00] sent to Marshall, which is what I called a renin-angiotensin system path diagram.  And I’d been reading Richard Feynman’s work with virtual particles you might say.  And so this in a way is a sort of path diagram trying to represent what happens to the level of different components in the renin-angiotensin system.  So for example the solid black squares represent measurements that we have made already and that we know are correct.  The white squares are what I think would be the level of the specific component if I could measure it.  Angiotensin I, angiotensin II, [00:10:00] and the compound of the receptor and the ligand, etc.  And so I have what’s called path direction.  And this is a flow or path diagram for a heterozygous plus minus angiotensinogen gene disruption.  All vertical scales are logarithmic relative to normal.  So quarter X, half X, one X, two X.  And it shows that if you have one X of angiotensinogen and then you have a copy of two X, that the renin level goes up was my thought.  In fact [00:11:00] it’s not correct.  The renin level actually goes down.  But this was the type of diagram made.

So the squares are hypothesized.  Solid.  Open squares.  And the solid ones are known.  And this is what I gave to Marshall.  And then we shall see what Marshall sent.  That is on page 73 on September 1st, 1996.  Marshall sent an initial model and equations and instructions on how to use this Stella program.  In this model angiotensin I was called AGTI and angiotensin II was called AGT Roman II.  ACE was specified as an amount of protein.  [00:12:00] And there were the initial equations.  In fact the program did not run correctly as it might be expected but it was the basis of what eventually did work because if you look at the Stella diagram there it’s almost a mirror of the pathways that I’d shown in the first of my diagrams to Marshall.  But it goes on and explains the difficulties in the following pages which are not worth going over in detail except to say a few things.  That at one stage I began to realize [00:13:00] that the steps that were taking place depended upon the concentration of either a substrate or a product.  And so there’s a little note on page 74 that a couple of lines are really quite critical in making the system work.  And I also began to realize a very important thing for later experiment and that is that if you try to vary the amount of a protein for example, this could be done by varying the rate of transcription of the corresponding gene, but that’s rather difficult since one doesn’t know [00:14:00] how to alter promoters in a predictable way to increase their efficiency.  But I realized that the other thing you could do is alter the rate of degradation of the message, not the rate of production but the rate of degradation.  So halving the rate of degradation would double the level of message just as would doubling the efficiency of the promoter.  And this was an important thing because it led later to a whole series of experiments which continue right to the present day in which we varied the effective level of a gene by altering the stability of the message.  And that can be done by changing the 3’ untranslated region.

But anyway [00:15:00] this modeling was difficult because it very rarely gave anything meaningful.  That’s to say when one ran the program it would come to equilibrium when either the substrate had gone to zero or something of the same sort.  And never seemed to produce anything valuable.  And then John Hagaman ran for me a series of simulations in which we changed the level of ACE.  This was October 15th, 1996.  And began to realize that one could find conditions where the final outcome of a simulation was a steady state.  So that [00:16:00] on the note on page 77 about halfway down it’s 10/24/9 looking at renin protein concentration shows steady state can be reached just by production equals inactivation.  You get to a steady state when the amount of something produced is equal to the amount of it being degraded.  But where that steady state is is what you’re trying to calculate.

And so on, it goes on.  Having to think about what’s happening to the concentrations with the Michaelis-Menten equation on page 78.  And so go page after page of a whole series of calculations made with this model on page 81.  And the model is beginning to look [00:17:00] more stable on page 80 with the program and so on, and continuing page 83.  More notes on what was happening.  Rather complex series of tests.  Really rather painful.  I didn’t very much enjoy them actually.  They lasted nearly six months.  And I have a drawerful of the programs that were used day after day after day with little changes to try to understand everything.  For example on 4/16, April 16th, 1997, it says a series of graphs [00:18:00] summarizing the effects of the above variable.  The data were reworked in various ways to try to superimpose the curve, etc., etc.  The conclusion, the renin direct effect is by far the largest on the blood pressure, 10-fold effect at least.  And angiotensinogen had about a threefold effect.  And the overall conclusion being need to measure these various substances.  Reaching what I called a standard model, which was published.  Preliminary account based on angiotensin I, angiotensin II with the February 6th standard model.

I think the overall conclusion of this was that the simulations were very valuable [00:19:00] because they made you think about all of the variables.  And I learned a great deal in making them.  And a whole series of experiments lasting a long time came out of them.  But the bottom of page 85, 5/20/97, shows what happens with the angiotensinogen gene copy number varying.  That there isn’t — that both angiotensin I and angiotensin II increase.  But that ACE gene copy number — that means that the blood pressure increased with copy number of angiotensinogen.

But the second diagram shows that ACE gene copy number doesn’t alter angiotensin II levels very much.  So blood pressure doesn’t change.  [00:20:00] But the substrate changes enormously.  Angiotensin I levels change.  And summarize that by conversion and inactivation, etc. on the following diagram.  And that’s the end of these simulations.  And back to experiments again on August 26th, Tuesday, page 87, with a review of the angiotensin receptor 1A coding region duplication construct.  And showing what the total construct should look like.  Getting back into step.  And designing PCR screens for duplication on Wednesday, August 27th, page 89.  Back to experiments.

[00:21:00] Nothing very remarkable about the experiments.  Just continuing in the ordinary vein to make these different constructs.  Sequencing nowadays — it was now getting possible easily to get DNA sequencing done.  So page 101 Wednesday, September 10th shows a whole bunch of sequencing results dup COD ATR duplication of the coding region of angiotensin receptor.  Quite a reasonable sequence.  And interpretation of the data on the opposite page.  [00:22:00] More sequence data on page 103, Friday, September 12th with different primers.

Just straightforward slogging away.  Now on Wednesday, October 15th, page 121, screening for angiotensin receptor 1A duplication.  Positive.  Rather disappointing conclusion.  [00:23:00] New project start on page 125, Friday, October 17th.  Aim to give an advantage to targeted ES cell in repopulating germ cell after blastocyst injection.  The idea being to overexpress the receptor for c-kit which is a growth factor for bone marrow cells and germ cells.  Overexpressed c-kit.  The receptor in primordial germ cells derived from ES cells.  So the thought that one might be able to get germ cells to proliferate more.  [00:24:00] So I requested c-kit cDNA from Alan Bernstein.  Received today the relevant plasmid.

Usual business of repeating work Thursday, October 23rd page 127.  One more ABI.  Can’t bear to abandon the duplication of angiotensin receptor, exclamation mark.  Going back to it.  Tissue-nonspecific alkaline phosphatase being considered on page 129.  TNAP.  [00:25:00] c-kit replacing TNAP.  What I was trying to do was to put c-kit into the tissue-nonspecific alkaline phosphatase gene and so have c-kit made where normally alkaline phosphatase was made in the germline cell.  So page 130 shows the rationale.  c-kit exchange with TNAP.  So Grant McGregor had made a targeting construct that depends on one that TNAP promoter, the tissue-nonspecific alkaline phosphatase promoter, is active in ES cell.  Etc., etc.  [00:26:00] So using some properties of that gene to control the production of c-kit.  Then the necessary slogging away to make the construct.  For example PCR of c-kit stop cycle one on Monday, November 24th, page 143.  Mistaken product.  But it was a good PCR.  Usual thing that sometimes you get a pretty result but you get the wrong answer.  Conclusion the same as one which is that it was a good PCR but not the right thing.  [00:27:00] And saved again for a rainy day, page 142.  Fresh start on PCR of c-kit stop Friday, November 28th.  Keeping going in the same vein.  December is approaching.  Still working at the same problem.  And the book ends with a design.  Saturday, December 13th, design of a blue PCR primers for the — I’ve got TPAP but I really mean TNAP c-kit fusion product.  So ends [00:28:00] this book on December 13th, 1996.  Book F’.  [00:28:09]