Contents
• Introduction
• Introductory story
• Summary of the conversation with Gemini
• The story I have arrived at
• A digression on counting viruses
• Conclusions
• References
Introduction
In the course of reading Benzer’s paper at reference 1, the Benzer who went on to become a scientific eminence of his day, I puzzled about how he could have made pure cultures of viruses – and, given the speed, in the right circumstances, that the viruses can replicate and mutate, about how pure these cultures needed to be for his purposes.
Reference 1 is a paper from the 1959, so written some five years after Watson & Crick wrote to Nature about the double helix in 1953. Prior to the 1950s, genes were thought to be the elementary particle of genetics – and if those genes were to be broken down into smaller entities how were these entities arranged in the molecules of genetic material concerned? Were they arranged in a one-dimensional linear array or were they arranged in some more complicated branching structure, possibly with loops, possibly two or three dimensional? Benzer showed that the small amount of genetic material involved in the many rII mutants of the T-phage he isolated (for which see reference 2) could be broken down into smaller entities, in a way that was consistent with their being coded as a string of nucleotides of the sort proposed by Watson & Crick.
I had come across this paper as a reference in the Wikipedia article on string graphs (reference 3), which were mixed up with the way one approached assembling all the overlapping short sequences one got from a genetic sequencer into a full-blown chromosome. An old paper. but one which appeared to provide an accessible introduction to the biology underlying these graphs – a sort of graph which, I may say, does not appear in my 2005 introduction to graph theory – although it does in appear in Wikipedia.
The beginning of this paper is snapped above. Noting in passing that it was communicated to Nature by his boss at the time, Max Delbrück. Perhaps that was the etiquette, the line of command in those days.
Not having any background in this sort of thing, I also went back to reference 4, a reference from the reference 1 which I started with.
In the course of looking at these two papers, it seemed to me that preparing pure strains of the mutant viruses involved must have been an important part of the work – despite the concepts of ‘mutant’ and ‘pure’ not sitting very well with each other: how mutated do you have to be to count as a mutant? At reference 5, for example, we are told that every human child carries about 70 mutations from their parents – mostly from the father – where I suppose they are mostly talking about point changes and deletions.

Digging into this, I found that growing viruses on a lawn of appropriate bacterial hosts in a small dish – perhaps a couple of centimetres or so across - is still an important feature of work in this field. Still very recognizably the same sort of thing that Benzer was doing back in the 1950s. I had turned up the paper at reference 6 and the manual at reference 7, both of which were helpful, but which did not seem to address the business of pure strains. So I thought I would ask Google’s AI product Gemini, to be found at reference 8. What follows is mainly a record of how we got on.
My first impression was that my (paying for) version of Gemini was convenient, accessible and helpful. But he was not infallible and he was not yet as good for this particular purpose as one would expect a good teacher to be.
A field for which a huge amount of material is available on the Internet for large language models to chew on, mostly good but some bad. It is far from clear (to me at least) how Gemini and his colleagues sort out the good from the bad.
I start with an introductory story then go to the conversation with Gemini and lastly summarize the story that I have arrived at as result. Plus a digression.
I might add that I have been using Gemini, and his predecessor Bard, at first reasonably regularly, now increasingly, for more than a couple of years now. They have come on enormously in that time: they have progressed from being a toy to being a tool. Also, that I am not here concerned with the wider merits or demerits of these sorts of tools – with, for example, the huge amount of electricity they are said to consume on their way to making lots of us obsolete, redundant or both.
Introductory story
Viruses are much smaller than the bacteria, the animal or vegetable cells that they prey on. The variety of present interest look something like the creatures in the snap above.
The idea is that the virus lands on the surface of its target and locks on. It then squirts its DNA down from the head, through the tail and into the target cell.
There, this DNA takes over some of the basic machinery of the host cell and replicates, perhaps yielding several hundred children. It then triggers the production of a special enzyme to destroy the cell wall, thus killing its host and releasing its children to the world to start over. A cycle which might take about half an hour.
A much more destructive parasite than a bacterium, which last might well be helpful rather than harmful.
See references 9 and 10 for a fuller story.
Google Images turns up lots of versions of the left-hand part of the snap above, a visualisation of a T4 phage, which reminded me of a moon lander or something of that sort, which probably accounts for its popularity, but I forget now where I lifted it from. The more florid version right comes from Facebook. Notice how solid geometry has got into the head.
Summary of the conversation with Gemini
Mainly in the form of my input, italicised in what follows.
In the 1950s, how would someone like Beezer have isolated a pure strain of a virus - in particular a strain of the rII mutant of T4?
Gemini was not bothered by my misspelling of ‘Benzer’, correcting it without comment
Lots of useful information here, including answering the question. Gemini has not just stuck to answering the question asked.
How big were these plaques?
Not being a laboratory person, I had little feel for this work and wanted more. The size of the plaques was a place to start. Gemini provides more information about plaques, without answering the question. Which prompted my next question, which he does answer.
But were they a few millimetres across or a few centimetres? Are the plaques sticky enough that they hold together if you pick up an edge?
And he goes on to explain that the plaques are not like a scab or the sort of thing I get with psoriasis, rather a circular patch of cell debris. For some reason, this took a while to sink in.
It sounds fiddly and time consuming.
This input was not a question, but Gemini provides a helpful response, with some good additional background, nonetheless. Quite like a conversation with a person in that way
But the video was a bit too technical for me. Too much information. I ought to say that, generally speaking, I do not respond very well to tuition by YouTube. Perhaps it is an age thing.
[some time later] Picking up on the T4 bacteriophage again, how long does a single cycle of replication for a phage take?
Gemini stores conversations and you can resume any one of them after a pause. A useful feature.
In this case, a simple answer, backed up with some explanatory material about the lytic cycle, for which see reference 9.
Afterwards, I got in a muddle with mental arithmetic. If one virus becomes 100 viruses in half an hour., what happens in fifty cycles, say a day’s worth? Is this really (10^2)^50 = 10^100? It does seem rather a lot. But today, Gemini agreed with my arithmetic and provided some background – both entertaining and relevant – including the fact that the number in question is called a googol. Bing agrees and takes me to Wikipedia which tells me that ‘… The term was coined in 1920 by 9-year-old Milton Sirotta (1911–1981), nephew of American mathematician Edward Kasner. He may have been inspired by the contemporary comic strip character Barney Google…’.
Gemini knows all about this too. And he did quite well on the follow up on the connection, if any, to the slightly later cartoon characters of Walt Disney,
[some time later] Going back to the plaquing, how could one be sure that at most one virus had infected any one bacterium?
A perfectly good answer, which I understand better now than I did at the time.
Do the plaques stop eventually stop growing? Or is the trick to stop the process before the plaques get too big and start to run into each other?
Another perfectly good answer. Including ‘… Plaques are typically counted within a specific time window—usually 18 to 24 hours for a fast-growing system like T4 and E. coli—to ensure the best results…’.
Sometimes, two viruses are going to infect the same bacterium. But if they are identical, does this mean that it makes no difference? The point of some of Benzer's experiments was that they were different.
Gemini opened with: ‘When multiple viruses, even identical ones, infect a single bacterium, it's called a mixed infection or co-infection. This is a crucial concept in phage genetics and was the very basis of some of Benzer's key experiments. It's not the same as a single-virus infection, because even with identical viruses, the events happening inside the cell are different’ and went on to tell about what happened when the two viruses were not the same. But he doesn’t explain the last sentence above, which may well be wrong. In any event, I maintain the view that in the unlikely event of a bacterium being infected with two copies of the same virus you would get the same result as being infected with one copy.
In your answer to my first question about this, you told me about a confirmation and purification step. Can you amplify what this means in the case that the plaque in question was actually generated from two, different viruses which could recombine in interesting ways?
Another good answer.
In this case, is not the main goal just to eliminate plaques generated by more than one virus?
I at first thought that Gemini was to be having trouble sticking to my point here, which was that one wanted pure samples of viruses to work with – regardless of how those viruses had been found or made in the first place. But once again, Gemini reads better with more knowledge.
From where I associate to FIL commenting on a pamphlet that I had given him about some aspect of mental health, probably to do with the old-style mental hospitals of his day, saying that it was a jolly good summary for someone who already knew the answer – but maybe not so good for someone who was looking for the answer. But then, he was a teacher by trade.
The story I have arrived at
The key to this work by Benzer is finding a group of mutants of the T-phage that can be distinguished by their response to plating on three different bacterial mediums, here B, S and K. The snap above being adapted from reference 2. With plating being explained at reference 3a.
The plaques are small clear, roughly circular areas on the plate, perhaps a millimetre or so across, where the virus has killed off the available bacteria. You get different looking plaques with different versions of the same virus and there is a good chance that, in what follows, you will pick the sort of plaque that you want.
Step 1. You mix a B strain of bacteria with a possibly mixed sample of T4 viruses. You plate it up and watch the plaques grow; a plating up which involves more or less fixing the viruses in the place where they arrive, they are not drifting about. Concentrations are adjusted so that it is very unlikely that any one bacteria will be infected with more than one virus and that the infected bacteria are well spread out on the plate.
Step 2. You identify a plaque of interest, in this case an rII mutant, and pick off a sample of its virus with a fine pipette. This sample is supposed to be a fairly pure strain, the product of just one virus The number of cycles of replication is small enough that the chances of significant mutation are small.
Step 3. You plate up your sample of virus with the B strain of bacteria. If you get lots of identical looking plaques, you know you have a viable, pure strain.
Step 4. You plate up your sample with the K strain of bacteria. If you get no plaques, you know that you have just one rII mutant. Mutant because you have no plaques and just one mutant because this test is very sensitive and any contamination of the sample with anything else – not least two different mutants recombining to revert to wild type – is going to result in plaques.
A digression on counting viruses
If one had a vial of liquid in which viruses were suspended, diluting that liquid in a controlled way seems straightforward enough, even when quite small quantities are involved, say millilitres rather than litres. So you might know that this sample has twice as many viruses to the millilitre as that sample, but how do you count the viruses in one of those samples to give you a baseline?
I was having trouble with this one too, so something else to try Gemini on.
He was good at explaining why counting them with an optical microscope, perhaps using the hemocytometers of reference 11, was not going to do, and although there are now more or less optical microscopes with the necessary resolution, they do not appear to be very well suited to counting.
One answer seemed to lie in counting the sort of plaques that we have at reference 1, but for some reason I did not press Gemini to explain why this might be, preferring to go my own way, and after a bit I turned up reference 12, the work of another (subsequently eminent) member of the Delbrück team. Where I found:
‘… From the fraction of live eggs in each batch the average number of infecting virus particles with which each egg was inoculated was determined by assuming a Poisson distribution of the particles per egg. This assumption is legitimate in view of the proof previously given in this article that infection of the cells is produced by one virus particle…’.
Poisson distributions ought to be something that I could deal with, so off to Wikipedia again, eventually landing up at reference l, where I find what I am looking for. An article which also loops back to Delbrück. All a matter of putting the question in the right terms.
So, if you randomly infect N cells with M viruses where N is a lot bigger than M, what can you say about the numbers of viruses in each cell? A first approximation is that M/N of them are going to be infected by just one virus, giving you M plaques. But if you need two or more viruses in a cell to generate a plaque, the proportion is clearly going to be very much smaller.
A first approximation which is confirmed by the Poisson theory at reference l, which gives us the required result that when just one virus in a cell is enough to kick off a plaque, relationship between the number of plaques and the dilution of the viral suspension is linear, which is what Dulbecco found at reference 12. Also by Kozikowski and Hahon at reference 13.
All you need to do is to dilute the viral suspension down to the right concentration to get the count in your infecting drop in the right range: small enough for M/N to be small, but big enough for the resultant plaque count to be statistically robust.
All very simple once one knows the answer, leaving aside all the lab work needed to do all this. An answer which I might have been able to get out of Gemini, but I chose to flog away at reference 12.
Along the way, learning about the natty device snapped above. An adaptation of the Carrel flask used in some of this work in the 1950s, adapted to better control the atmosphere above one’s Petri dish, while leaving the top accessible to one’s microscope. Lifted from reference 14.
Conclusions
The important thing is that, with Gemini's help, I now feel I have a good enough answer to both my primary (purity) and my secondary (counting) question for my purposes.
Gemini can do very well, but you have to hit on the right questions to ask him and sometimes you have to work a bit to get him to stick to the point. He can be repetitious, but that does not really matter.
He will sometimes give you the answer you want straightaway, and you have confidence in that answer, but more often you need to have a dialogue with him. He has got pretty good at following the thread of a conversation.
I have only picked up one possible mistake and one definite mistake.
Interesting that a lot of his material reads better now I have got a better understanding of the matter in question.
PS: I get the impression that this sort of thing is taking me a lot longer than it should, or at least a lot longer than it would have done a few years ago. Maybe it is just as well that I have now got Gemini to help me along.
PS 2: and let’s not forget Wikipedia, which also has a big part to play in all of this, not least because it is the source of a lot of Gemini’s material.
PS 3: along the way, I have learned about Microsoft's Markdown version of Notepad. Hmmm.
References
Reference 1: On the topology of the genetic fine structure – Seymour Benzer – 1959.
Reference 2: https://en.wikipedia.org/wiki/Escherichia_virus_T4.
Reference 3: https://en.wikipedia.org/wiki/String_graph.
Reference 4: Fine structure of a genetic region in bacteriophage – Seymour Benzer – 1955. Reference 6 from reference 1 above.
Reference 5: Evolution 101: Mechanisms: the processes of evolution – Understanding Evolution, Berkeley - 2020.
Reference 6: Viral Concentration Determination Through Plaque Assays: Using Traditional and Novel Overlay Systems – Alan Baer, Kylene Kehn-Hall – 2014.
Reference 7: Virology Guide – ATTC – 2025. According to Wikipedia: ‘ATCC or the American Type Culture Collection is a nonprofit organization which collects, stores, and distributes standard reference microorganisms, cell lines and other materials for research and development. Established in 1925 to serve as a national center for depositing and distributing microbiological specimens, ATCC has since grown to distribute in over 150 countries. It is now the largest general culture collection in the world…’.
Reference 8: https://gemini.google.com/.
Reference 9: https://en.wikipedia.org/wiki/Lytic_cycle.
Reference 10: https://en.wikipedia.org/wiki/Escherichia_virus_T4.
Reference 11: https://en.wikipedia.org/wiki/Hemocytometer.
Reference 12: Production of plaques in monolayer tissue cultures by single particles of an animal virus – Renato Dulbecco – 1952.
Reference 13: Plaque formation by psittacosis virus – Edmund H. Kozikowski, Nicholas Hahon – 1964.
Reference 14: Long-Term Cell Culture on a Microscope Stage: The Carrel Flask Revisited – D. J. Stevenson, D. J. Carnegie, B. Agate, F. Gunn-Moore, and K. Dholakia – 2008.