Skip to Content

Bad Cells. So Many Bad Cells.

Let’s file this one under “We’ve seen this before, and I’ll bet we’ll see it again”. Anyone who’s worked for some years in cell culture (or with people who have) should appreciate the dangers of cell line contamination. You can get mycoplasma, you can get other cell lines entirely (particularly others that are more vigorous and hardy than the ones you’re trying to grow), and you can get viral thingies that cause you so much trouble that your entire company gets bought by someone else.

Here’s a new paper in PLoS ONE that tries to get a handle on the problem. The real kicker is that some of these cell lines became contaminated along the way, so that earlier papers and later ones in the field are actually referring to different cells. And others became contaminated (or mis-identified) so early that basically all of the literature on them is mistaken. Warnings have taken place about this stuff again and again, and the current literature is surely cleaner than the older papers. But how bad is it in the published record?

By correlating the literature with a list of known contaminated cell lines (many of them invaded by HeLa cells), the authors estimate a lower bound of over 32,000 papers that have worked on the wrong cells, compared to what they report. In turn, these papers are cited by at least 500,000 more articles, and that total excludes self-citations. And as the authors note, they were quite conservative with their name strings in the searches, so although there are also still a few false positives in those numbers, they are surely tiny compared to the false negatives – the mistaken papers that haven’t been flagged yet. A representative example:

In a 1994 report, the establishment of a group of novel thymic cell lines (F2-4E5, F2-5B6, P1-1A3 and P1-4D6) [39] was announced. In a report by MacLeod et al. [40], the cell lines were found to be misidentified, having been derived in fact from a liver carcinoma. In total, 69 articles were found that refer to these cell lines, in turn cited by 2092 articles. Of the primary articles, 43 were published after the report by MacLeod et al. and the most recent one was published only in late 2016 [41]. Of the fifteen most recent articles referring to the 1994 report, thirteen actually refer to it because they use the cell lines, all thirteen reporting research on thymic cells, without mentioning any knowledge of the misidentification of these cell lines. The other two articles refer to the establishing article for the sake of the method used in it to establish novel cell lines.

So yeah, there’s a lot of crap out there. All you folks who are trying to machine-learn your way through the medical literature, you now have a half million more papers to flag: and remember, there are plenty more where those came from. And as the authors note ruefully, just flagging a cell line as misidentified is not enough to stop people from using it. The “Chang Liver” cell line was established in the early 1950s, but as early as 1967 it was suspected to actually be yet another culture of HeLa cells. Ten years later, more evidence was presented, but the cell line originator argued at the time that these were real liver cells that had changed in cell culture, not HeLa. The question was resolved beyond doubt in 2001, so that should have been that. Right?

Wrong. A search through the literature will show “Chang liver” cells are still being used as if they were liver cells, with no mention of the misidentification. Hundreds of papers in the last fifteen years have done so, and their appearance in the literature shows no signs of going down. If anything, it may be slightly increasing. This new paper finds the same trend. The number of papers citing cell lines that are known to be wrong is increasing – perhaps not as a percentage of all scientific papers, but it sure isn’t going down, despite numerous warnings and exhortations.

You’d think that this would largely be a problem for areas where research has not been as well established, or where appropriate safeguards haven’t yet been put in place. For example, a 2015 report – from China – suggested that 85% of the cell lines established there were wrong, and were almost entirely HeLa cells. This latest work finds that China’s share of the contaminated cell line literature is indeed rising rapidly (like, congratulations, guys), but that the majority of such papers are still from the US, Japan, Germany, and the like. No one has any room to feel superior, for all have HeLa-ed.

What to do? At the very least, the authors suggest, we could flag these papers in the databases with a note that they used a cell line that is known to be misidentified. Later readers can then deal with them as they will. But for papers going forward, I’m stumped. Nothing seems to have worked. Just as with bad chemical probes, people just plow right ahead no matter how many warning flares you send up. Any ideas?

41 comments on “Bad Cells. So Many Bad Cells.”

  1. Tony Verow MD says:

    I am not an academic physician (not even close!) but has this issue been studied from a Quality Control mindset? Would it be possible to establish rigorous standards for typing/categorizing cell lines prospectively(perhaps even de novo cell lines) to avoid future errors? Obviously such a rigorous “reset” would have to be funded generously and it simply may not be possible right now.

    I believe that you have pointed out a major flaw in using such contaminated cell lines to do research. I have no simple answers.

    1. anon says:

      We run a large scale cell line profiling operation and that is what we do- fingerprint before and after each run. Companies (e.g. Fluidigm) market solutions to this but I’m sure underfunded academic groups aren’t excited to add another routine cost to every experiment.

    2. johnnyboy says:

      Well, you’d have to introduce the ‘quality control mindset’ to every academic lab in the world (ie. explain what ‘quality control’ actually means to every single MSc, PhD student and post-docs, the ones who actually do the work). Myself I had never even heard the words ‘quality’ and ‘control’ until I got out of academia and joined industry, so you have a ways to go.

    3. JB says:

      Yes, it isn’t hard. The NIH has already made a big push for validating all reagents, cell lines etc. Journals need to stop accepting manuscripts without cell line authentications. It only costs a few hundred dollars and most universities are capable of doing it rapidly.

  2. Project Osprey says:

    Surely the journals could/should be doing something about this. It can’t be beyond the wit of man to fix

    1. ScientistSailor says:

      It’s not a matter wit, but will and resources. At the Big Pharma I worked for, cell lines were routinely sequenced before and after each experiment.

  3. publish & perish says:

    Because NIST is all about standards they’ve looked into this issue and surely it’s been a topic at internal and external NIH meetings. Aside from recommended quality control protocols, particularly in the context of publications, it’s not clear how QC has actually been translated or enforced. For that matter, do publishers/reviewers typically enforce currently recommended protocols or standards? How are European and Asian institutions addressing this – it seems it will take global alignment, much like ISO standards (but hopefully more transparent and less bureaucratic)

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4907466/

  4. Emjeff says:

    The scope of this problem is shocking. Derek, do you think this is one of the contributing factors to the problem of irreproducibility of lab results?

    1. kriggy says:

      I would think so. Suppose you want to run a chemical reaction from paper you find, but you use wrong starting material and wonder why it is not reproducible

      1. Hap says:

        And you can’t just run an NMR on the starting material to tell what you started with, and the starting material changes between lots and over time.

  5. Bla bla says:

    I see why people still use cell lines, hard to do some stuff without them, but I’m a firm believer in primary cells only. Disadvantages to those as well, but at least they vaguely resemble the cells you’re interested in. Rarely the case in cell lines.

    1. greasypocket says:

      Misidentified/contaminated cell lines goes hand-in-hand with s*it antibodies – not to mention outright fraud. To be frank, I estimate that >90% of cell/molecular biology papers from academic labs is absolute garbage. And this garbage just grows exponentially as more and more self-serving trash gets poured into the cesspit. Thankfully, in the pharmaceutical industry we treat academic papers with (bucketfuls) of salt, and nothing is correct until it has been repeated in our labs with validated cell lines, orthogonal assays, etc. By contrast, in academia it doesn’t matter if the research is correct when their goal is to lengthen their CVs in order to go cap in hand begging Uncle Sam for more tax payer money to waste.

      1. Greg says:

        No really, tell us how you really feel.

      2. Bla Bla says:

        Someone didn’t get tenure and is a bit bitter….. It if wasn’t for academia you wouldn’t have any disease mechanisms to work on, no tool to work on them with and no chemistry to make drugs with.

  6. Uncle Al says:

    If my discipline were discovered to have the empirical validity of psychology, I’d flee the lab to management. Management makes decisions, labor makes mistakes.

  7. milkshaken says:

    I was working within a fairly large biology academic group as the only chemist there. A presence of mycoplasma in their cell cultures came as a sudden, startling revelation to them – and not that they were eager to do anything about it, tell their collaborators about it or try to reproduce their published previous work with clean cells. (Grantsmanship and the realties of foreign postdoc labor encourage publishing maximum positive results.)

    1. Electrochemist says:

      An interesting paper from 15 years ago found that (at that time) only about 60% of the leukemia-lymphoma cell lines they had access to were authentic, and that 50% of contaminated cell lines were contaminated with mycoplasma.

      So clearly these issues have been known for a long time.

      Reference: Cytotechnology 39: 75–90, 2002

  8. Anon says:

    “85% of the cell lines established there were wrong, and were almost entirely HeLa cells”

    Currently it is estimated that Henrietta Lacks (the patient from whom HeLa cells were derived) is the heaviest human ever, weighing in at over 20 tons. But now that looks like that is a gross under-estimate!

    1. a. nonymaus says:

      I’d rather say that she is the most prolific mother. Her untold myriad billions of descendants are genetically and spatially distinct from her, live separate lives across the globe in modest comfort, and regularly produce new generations to go out into the world to make their own way. Of course, given the obvious phenotypic and genetic differences, I prefer to think of her unicellular lineage as the model organism Homo Hela, with its unwanted presence in many labs as evidence for its survival in the wild (viz., independent of human whim) as a commensal organism.

  9. Peter S. Shenkin says:

    As far as what to do about it is concerned, I think the journals could have a role to play.

    Just as organic-chemistry journals used to (and to some extent still do) require an elemental analysis of new compounds reported prior to publication, perhaps (starting with the high-ranking journals) research involving cell lines should require a report of a cell-line profile for any cell line used in the research.

    This assumes that the profiling services of such companies as Fluidigm (referred to in an earlier compound) exists for the cell line and makes sense for this purpose. I’m not a biologist, so I don’t know how reliable or how comprehensive such coverage is.

    But assuming there is a way for a cell line’s origin to be validated, it would seem that the journals could play a major role in ensuring that such validation is performed and reported, by requiring and publishing the validation.

    1. Peter S. Shenkin says:

      P.S. I said “earlier compound”. I meant “earlier comment.” (See what happens when you’re a chemist?)

  10. tlp says:

    I think all this originates from the same premise that makes people value action over (perceived) inaction i.e. ‘getting things done’ attitude with implicit ‘no matter how’.
    And to be frank, quite some of those results obtained in HeLa cells instead of whatever else would be roughly the same in other cell lines. That’s probably why the whole science edifice is not falling apart after sudden insights that cell line contamination/passage number/inappropriate statistical analysis/antibodies/animal’s microbiome/circadian rhythms/technician’s sex can drastically influence experimental outcomes.
    Solution? Probably once sequencing is dirt cheap, journals can require some kind of cell ‘passport’ submitted with SI. But then again, this would work with high-rank journals, where there is competition to get published in. In journals having troubles filling their TOC with new papers this probably won’t work.

  11. Desiree says:

    In my PhD lab we instituted mycoplasma screening of all active cell lines every 3 months and whenever we got a new cell line. We stopped taking cells from other labs, both to reduce the risk of introducing mycoplasma in to our lab and to try to make sure that we have the correct cells. We ordered new cells from ATCC if we needed new cell lines. However, when we looked (~2015) at methods to make sure the cells were the correct cell type there were not a lot of affordable options so we trust that ATCC sent us the correct cells and that our freezer stocks are correctly labeled.

    This came about because we found mycoplasma in some of our cells. Fortunately at the time we were a new-ish lab with a handful of cell lines. After we decontaminated the room and replaced ALL of the cells (ensuring that they were clear) we took the opportunity to put in some procedures to try to prevent this problem in the future or at least improve containment. Fortunately, when we did some spot-checking by running previously completed experiments we found that we had similar results, but we weren’t doing hard core cell bio experiments. The plan we put into place isn’t perfect, but at least it offers bracket to our experiments, if your cells were clean at the last screen and they are contaminated now, you have a window of work to repeat vs doubting all of the work.

  12. Roger says:

    Sequencing at a fingerprint level is dirt cheap, fast, and accurate. Whenever I’ve been in a position to make it happen I’ve implemented these quick and easy QC checks. Yes, they made some people’s lives miserable, but we were accurate in describing the lines we were working with. Most people are simply not interested in knowing what they are working with because it could quickly become inconvenient and slow the rush to publish. Only publishers and granting agencies (and company scientific advisory boards) have the power to change this. It’s trivial to do, what is lacking is will.

  13. Anon anon anon says:

    It’s hard to tell one’s intended tone from text. That said, I assume “All you folks who are trying to machine-learn your way through the medical literature, you now have a half million more papers to flag” was intended pejoratively.

    In reality, it’s the most hopeful. After all, it shows a path to solving this problem, whereas the rest of the post demonstrates that *people* have consistently failed to flag these papers for the past fifty years.

    1. tangent says:

      Yeah, this. I don’t think we have long at all before machines — as dumb as they are — understand aspects of the literature more wisely than humans do. Because god, humans are dumb sometimes. Machines don’t have their motivations.

  14. Dr CNS says:

    So, should the solutions of the analogs we test using a cell line be analyzed for contamination with microorganisms?

    1. Cialisized says:

      Compounds that are kept as stock solutions in pure DMSO are assumed to be bug-free.

  15. MikeAndrews says:

    My wife the biochemist was working with a cell line that came pre-contaminated with a fungus, all the way from Australia. Because it was fungal, the antibiotics she was using in her project weren’t working. The prof chewed her out for bad technique, so she invited him to watch; he couldn’t fault her technique, so he tried himself, and got the same failures.

  16. petros says:

    Some years back Genzyme had a major problem of viral contamination at its manufacturing plant in Allston Landing.

    This meant it couldn’t produce some of its enzyme replacement therapies posing considerable problems for the patients

  17. steve says:

    I had a different experience than your wife. During my post-doc years we started having yeast contamination in all our cultures. After cleaning out incubators, hoods, etc. it kept coming back. I couldn’t believe my technique was that bad so I had a private meeting with the PI (who was female) and suggested that possibly one of the lab technicians might have a yeast infection. She looked at the contamination and said, no, it wasn’t that kind of yeast (Candida is long strings and these were individual cells). Stimied, she asked during a lab meeting if anyone had changed any of their habits lately, either inside or outside of the lab. After pretty extensive discussion, one technician finally volunteered that she had been taking a baking class after work. Case solved – she covered her hair from then on in the TC room and no more baker’s yeast.

  18. Anonymous says:

    “Betrayers of the Truth” by Broad and Wade includes the case of John Long at MGH in 1979. Long thought he had made a breakthrough discovery but then had to cover his tracks by knowingly falsifying data when it turned out that his human Hogdkin’s cell lines had been overtaken by monkey cells. (I have been told that it should have been absolutely trivial to see that the cells were not Hodgkin’s and possibly not even human under the optical microscope.)

    “The Immortal Cell; Why Cancer Research Fails” by Gerald B. Dermer in 1994 made the case that cell based research is seriously flawed for many reasons. From an Amazon review: “It’s not just another angry soapbox rant about how doctors and researchers don’t really want to cure cancer because that would put a stop to their profit flow. This book is based on detailed observations that Dermer made over his years as a pathologist. It is written in clear, non-technical terms and can be absorbed in one or two sittings.”

    I know of more recent cases of botched assays. When assays are reduced to “kit” form, anybody thinks they can do it: buy a kit, splosh around some stuff from bottles A, B, and C, put it in an expensive machine, press a button, and – whoosh! – data for a paper! (I know of a group that left out a key step in an assay kit because it seemed unnecessary to them and it saved time. Wrong!!!)

    As Peter, Paul and Mary used to sing, “When will they ever learn? When will they ehhhhhhh-ehhh–ehh-ever learn?”

    Off topic: The Captchas are getting harder! I had one that needed a two digit answer!

    1. Wile E. Coyote, Genius says:

      Sorry, but if someone tells you that you can distinguish monkey cells from human cells (or any other mammalian species) under the microscope, they are gravely mistaken. Requires stains for specific markers. If they were distinguishing Hodgkins lymphoma cells from fibroblasts, yes, that you could do.

  19. yf says:

    I worked in both academia and industry. Research in academia and industry have different focuses. In academia, the idea is most important. As a result, academia labs usually use many different approaches (cell line is one of the approaches) to prove one point. Flaws in one approach ( e.g. cell line) has limited impact on the final conclusion, since results from multiple approaches has to converge. On the contrary, in industry, multiple scientific ideas tend to converge into one standardize platform or one approach. Since this platform or approach will be the product companies sell, it needs to be consistent whereas flaws in individual scientific ideas is not a big concern. For examples, Promega developed some popular robust luciferase assays for many Immunology-oncology candidates. I am sure the quality of Jurkat cell line are rigidly controlled. In academia labs, these assays are highly contrived and their results are highly non-physiological. When I was in academia, we don’t want to see results from cell lines, we want least manipulated animal models and human PBMCs. Now, I am in industry, I prefer a robust cell line to PBMC to give me consistent results. However, I want my antibody candidates to be as diverse as possible.

  20. dearieme says:

    “Research in academia and industry have different focuses.” Yes, industry is about profit just as universities are meant to be about truth. Unfortunately the latter ambition seems increasingly to be watered down.

  21. JB says:

    Yes, pile on the hate on poor ‘ol academia, because industry time and again has proven that it is much better at doing unbiased and good science. Oh wait, that’s right, the biopharmaceutical industry has literally done things like sold known HIV contaminated blood to the market, Avanir continues to push their little red pills onto elderly patients who don’t even need it in order to rake in hundreds of millions in profits, and multiple studies, including one in JAMA, have shown that when for-profit companies run clinical trials they’re nearly 3x more likely to recommend an experimental drug when compared to completely unbiased trials run by non-profits. What do they say about those who live in glass houses? Let’s stop pretending like industry has always done the best science possible because all too often the $ corrupt their data integrity.

    >90% of papers in biology and molecular biology from academic labs is garbage, eh? Wonder if the same could be said about the quality of data for most clinical trials run by pharma. When academics do something wrong they get a retraction and deservedly get scrutinized by their peers when they get caught. When pharma does bad science they literally kill people or push their drugs on the market that have little or no benefits at all while burdening the system with billions of dollars in costs.

    1. Harrison says:

      There’s a lot of accusations packed into there, but if you deconstruct it, it really comes down to the motivation behind the research.

      For academic basic/pre-clinical research, one needs to get published to get grants to get tenure, promotion, etc. Novelty and publications are incentivized, and the fact that a good amount of the research is not reproducible and/or failures to reproduce are not published leads to the current state. For industry pre-clinical researcher, the goal is to get a small molecule or biologic to the clinic that has the smallest chance of failure. Being sure something is not a waste of time and money for a clinical trial is critical. There is no need to publish, and secrecy may be valued for a time, so failures to reproduce are not published, also contributing to the current state.

      The relationship is turned on its head for clinical research. While some of the same caveats apply, academic clinical research is more about patient outcomes. They want to identify the treatments that work best. For industry, they want to get a drug to market. Obviously a company would like to have the best product on a market, but after all the R&D cost, most would settle for a marketable product.

      It seems as if industry researchers police the pre-clinical research, and academic researchers police the clinical research.

    2. anon says:

      As a scientist in biotech, I get a lot my ideas from published papers. I am grateful for this “free” data. I am also keenly aware of that what I got is what I paid for. I do my own due diligence to validate published data. personally, I do not see many companies that will offer their data for free.
      Considering how american universities are managed, academia have the same incentive as industry — grant and money. Given the drastic pay difference between industry and academia, I don’t think scientists in industry deserves a higher moral ground.

  22. nonny says:

    My favorite story- we collaborated with an academic lab who sent us a cell line to run an assay they had published. As is our SOP, we tested for mycoplasma and found it contaminated. When we told the professor, he said he knew and that the assay only worked when the cell line was mycoplasma positive. Ended that collaboration in a hurry.

  23. Ted says:

    I wonder if PLoS ONE was their first choice for publication? I mean, why not… “Cell”?

    -t

  24. Dave says:

    The only talk I remember from my undergraduate days was a 1979 talk on Hela cells and how they had taken over so many cell lines in so many labs. I remembered it because it was such a striking talk about such a huge problem.

    It makes me wonder if in 2055 (i.e., another 38 years from now), we’ll still be writing about Hela cells taking over cell lines.

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.