The HeLa problem: What a retraction says about whether cancer researchers can trust their cell lines

Retraction Watch readers who’ve read Rebecca Skloot’s bestseller The Immortal Life of Henrietta Lacks may remember that decades ago, scientists began realizing that Lacks’s cells, now known as the HeLa cell line and used in labs around the world, were so good at proliferating that they had taken over many other cell lines researchers use to study human disease.

Such readers would have been nodding their heads at a front-page Wall Street Journal on Saturday. As Amy Dockser Marcus (no relation to Adam) reports:

Cancer experts seeking to solve the problem have found that a fifth to a third or more of cancer cell lines tested were mistakenly identified—with researchers unwittingly studying the wrong cancers, slowing progress toward new treatments and wasting precious time and money.

In hundreds of documented cases that undermine a broad swath of research, cancer samples that were supposed to be one type of tumor have turned out to be another, through either careless laboratory handling, mislabeling or other mistakes.

It’s not just HeLa cells that have been found as contaminants; a melanoma cell line is thought to have been mixed up with a breast cancer line, for example. As has become a common refrain on Retraction Watch, however, when scientists tried to sound alarm bells, they went unheard:

But researchers who yelled loudest were mostly ignored by colleagues fearful such a mistake in their own labs would discredit years of work.

Leaders in the field say one of the biggest obstacles to finding a cancer cure may not be the many defenses nature affords malignancies, but the reluctance of scientists to address the problem.

The WSJ study leads with an anecdote about a retraction in Oral Oncology. The retraction notice in question is quite detailed, which we appreciate, but what’s particularly puzzling is the reason for the retraction:

The cell line ACC3 which was used in the study reported in the above named paper was reportedly found cross contaminated with HeLa in several laboratories world-wide (Phuchareon et al., PLoS One 2009;4:e6040). This point was raised during the peer-review process of the manuscript. The authors obtained the cells directly from one of the original sources (Dr. Saku, Niigata University School of Dentistry, Japan) and observed a TP53 mutation in the ACC3 cells, which led them to believe that the cell line was distinct from HeLa because the latter is known to carry wild type p53.

Following acceptance of the paper for publication, the authors were unable to confirm the TP53 mutation during a repeated sequence analysis. A short tandem repeat (STR) analysis, which is recommended for cell line verification, was then performed. The results found that the ACC3 cells used in the study correspond to the ACC cell lines cross contaminated with HeLa as reported by Phuchareon et al. The authors communicated this new information to the Editors of Oral Oncology.

The scope of Oral Oncology covers only head and neck cancers. As the findings of this paper no longer refer to a cell line that is implicated in these tumours, the paper has been retracted from the journal, since it would not have been acceptable for publication during peer-review based on the information now available.

The major conclusions of the manuscript regarding the influence of SIVmac239-Nef on tumor cell proliferation and migration in CXCR4 expressing tumor cells, as well as on angiogenesis, are not affected.

In other words, the real reason the paper is being retracted isn’t because the results were no longer valid, but that they were no longer valid for a head or neck cancer.

Retraction Watch friend Jeff Perkel wrote last year in BioTechniques about another case of mistaken cellular identity:

In a January 2011 article describing how estrogen metabolites influence the release of inflammatory molecules from human amnion-derived cells, Italian researchers took advantage of a cell line called WISH, which they describe as “constituting a model for in vitro studies of amnion functions.”

There was just one problem: WISH cells have been known for more than 40 years to be highly proliferative HeLa cervical cancer cells. The authors did acknowledge this fact in their conclusions, albeit obliquely and only briefly: “… considering WISH cells as a suitable model for human amnion cells, our results suggest that active estrogen metabolites may finely modulate the onset of human labor. Yet, looking on recent reports, by which WISH cells should be more meaningfully be likened to neoplastic cells, our results could support estrogen metabolite relevance in neoplastic competence via inflammation pathway activation.”

That “caveat,” Perkel notes, “makes the study’s scientific relevance open for debate.” We particularly like this quote from Perkel’s piece, which we’d recommend reading in its entirety:

“What really frustrates me is that Stan Gartler showed that a whole series of lines were cross-contaminated with HeLa cells, and yet those bloody cells are still being used now in their false guise,” exclaims John Masters, professor of experimental pathology at University College London. “For nearly 50 years, people have been using falsely identified cells totally unnecessarily because they haven’t checked.”

The Wall Street Journal story — which we’d also recommend reading in its entirety — also quotes Masters, and highlights another issue. We’ve actually been aware of the Oral Oncology retraction since January, and have been meaning to write about it, but it got lost in our steady deluge of great tips from helpful readers. That’s just another reminder of just how many good retraction stories there are nowadays. (And that we’re two guys with day jobs, of course.)

31 thoughts on “The HeLa problem: What a retraction says about whether cancer researchers can trust their cell lines”

  1. This is waiting to happen one day. I had been questioning in all the seminars and talks in our institute whenever someone presents a paper on anticancer properties of a novel compound in HeLa or similar cancer cell types. It is hard to find out whether the same wonder compound will work in vivo or whether it has any effect on normal cells.

    The other worrying part is most of the gene identification and gene interaction studies are mostly done on HeLa cells where investigators overexpress or downregulate a certain gene and publish the data and results in high profile journals such as Cell, Nature and Science. Once such observations are published in BIG journals by BIG people, it is highly impossible to refute the studies if those data turned out to be not correct. Once role of a particular gene/protein has been identified in such an abnormal cancer cell line, it is expected that these genes will have analogous role or function in normal cells. Then follows 100s of papers on this topic proving a similar role in other immortal cell types and the story continues…..

  2. “Leaders in the field say one of the biggest obstacles to finding a cancer cure may not be the many defenses nature affords malignancies, but the reluctance of scientists to address the problem.”

    Name one.
    I dare say someone will have said it, probably after having had a few too many drinks.

    @Ressci Integrity – everyone knows the limitations of cell culture models, everyone also knows their advantages also.

    1. @littlegreyrabbit: Agree that everyone knows the limitations. I have seen papers/people working on a mouse cell line thinking that it is a human cell line – they don’t look at the karyotype (See comment from David below). How do you consider this?

  3. Yes, this is a major problem — we constantly nix papers in review because of this. Even some of the primary ENCODE lines, when passaged, evolve an abnormal karyotype, yet rarely do you see info about passage and karyotype (but, slowly this is improving). When you do NGS on a cell line with an abnormal karyotype you need to make certain your reads (whether genomic, RNA-seq or ChIP-seq or …) aren’t biased by the potential multiple copies. I’m not aware of any current alignment algorithm which is able to take this into account (would love to hear of one).

    Often this doesn’t impact the research question, but often it does: one can’t be naive about it. An uninformed reader would have no background allowing them to discern when it truly made a difference, that’s the sad part. It’s also sad that journal editors (whether big or small) allow these to pass into review.

    1. It’s not the cell lines I am most concerned about in the ENCODE ChIP-Seq’s, rather the antibodies they used. In any case if the cells are purchased at low passage from a reputable source, it should be easy to enforce regulations, i.e. keep passage number below 30 etc etc. I think it is important not to get TOO bogged down with some of this though. I mean, seriously, are you really concerned about multiple copies in ChIP-Seq data? Seems a little pedantic to me. You may have more of a case with RNA-Seq but, again, with the right controls, I also don’t see this being a major sticking point.

    2. You aren’t seriously suggesting you think HeLa cells taking over other cell lines is a major problem??!!! Sure if the thousands of labs and many of hundreds of cell lines doubtless it has happened on occasion – but a major problem?

      If you mean that immortal cell-lines change in culture, can be unstable and have severe limitations compared to the original tissue or cell type they derived from , well obviously – but everybody knows that. This is beginning to have a “should we abolish western blots” feel about it

      As for NGS on a cell line before allowing a paper published – seriously!???? After whistle-blowing on misconduct and being obliged to leave science I am doubtless out of date now, but it still seems highly unusual

      1. I’m sorry, I’m not sure I understand your comment. You seem a little agitated for some reason. Just to clarify, yes I do think it cell culture contamination, whether it be caused by other cells or micro-organisms, is a major problem in biomedical research. Of course it is. i don’t see how you anyone can deny that. Does it invalidate the majority of cell culture work? OF COURSE IT DOESN’T. You should read my other post below if you have any doubt about my feelings on this debate. We should keep this conversation in context and not take it to the level where we are questioning cell culture studies. I am certainly not willing to do that.

        Your last point about NGS made no sense to me at all. I don’t know what you are trying to say.

      2. @Dave “I’m sorry, I’m not sure I understand your comment. You seem a little agitated for some reason…..Your last point about NGS made no sense to me at all. I don’t know what you are trying to say.”

        That is because the comment was not remotely directed at you, rather to the preceding post from “Dave”

        The very close time stamp between our 2 posts might have given you a clue and the fact that “Dave” mentioned asking labs to do NGS on cell lines before permitting their papers to appear could also have been considered something of a give away.

        He hasn’t been back to clarify this.

    3. @littlegreyrabbit I am getting the impression that you leaving science was not entirely caused by whistle-blowing… reading and comprehension skills are quite important for scientists, and I can’t see any mentioning of “NGS required” in David’s(!) post. He only says that abnormal karyotypes lead to changes in read counts, which is absolutely correct.

      @David: There is no need for alignment algorithms to take care of this. They are agnostic to genomic structure, as they should be. It is up to the bioinformatician/statistician to account for bias in the down-stream analysis, using for example tools for CNV analysis. If this is done correctly, is – of course – a completely different question.

      1. Sebastian, I bow to your superior reading and comprehension skills. I dare say I am not up to it intellectually as far science goes – fortunately a juicy piece of blatant scientific falsification came along that allowed me to conceal my manifest deficiencies. Lucky me! I always secretly dreamed of a career in data-entry anyway.

        Can you explain what you think David meant by “nixing” papers because of this? I have never heard of anyone being asked to prove the identity of their cell lines during the review process.

        Anyway to the subject of the post. I really don’t think HeLa contamination is anything other than a very rare problem.
        1. Cell lines have quite distinct morphologies, while there a number of cell lines that appear HeLa like, the majority don’t. Hence there is only a subset of cell-line that HeLa could potentially colonise
        2. You have to be pretty careless to cross contaminate.
        3. Most labs don’t grow HeLa cells anyway. The only time I used them was at my own request mainly because I wanted the experience of using a bit of bioscience history myself, than expecting it to provide any results. And I am quite certain – due to distinctive cell morphologies – they didn’t colonise anything.

        We were trying to replicate a very exciting new anti-proliferative factor recently described by a very prestigious US group – very good at pulling grants and getting into high ranked journals. I was using the same cell-line, recently acquired from the ATCC and totally failing to replicate the results – not even the tiniest effect when throwing micrograms at them. So just on the off chance I suggested at throwing them at HeLa cells – not surprisingly that didn’t work on them either. At the next conference my supervisor casually mentioned as an aside we were finding it difficult to replicate the results, at which a couple of other labs piped up and said they hadn’t either. Where upon the very prestigious US group said they had pulled up new stocks from liquid nitrogen and had found that the results didn’t reproduce – but it didn’t matter because it worked in yet another cell-line. As it happened I had already tried this cell-line and had also drawn another blank.

        Anyway to cut a long story short, the prestigious US group is still prestigious, still pulling large grants and still getting into high ranked journals. We had put in large grant proposals which specifically hinged on this anti-proliferative factor actually being anti-proliferative, so rather than apply it exogenously we over-expressed it in a couple of cell-lines and said they showed lower growth than wide-type – completely dodgy reasoning in my opinion but it tided us through with publications to the next funding round.

        In summary, there are indeed problems with cell culture experiments, but widespread HeLa contamination is not one of them.

      2. Hi Sebastian,

        Thanks for helping to clarify.

        Yes, I agree with your assessment regarding alignment algorithms. What I often encounter, both in the review process and amongst collaborators, is the naive assumption that copy number doesn’t matter when working through a NGS pipeline. This results in a skipping of that important down stream analysis you mention. My desire for alignment to automatically take it into consideration arises from the realization that we’re still in a time where access to that important second step “bioinfo/stats analysis” is patchy, but access to NGS has blossomed.
        As I mentioned, and @Dave points out (I do also agree about the antibodies), whether copy number is impacting your results depends on your experiment. My experience has been that any consideration of copy number is rarely encountered — unless one’s question is specifically addressing it. The consequence leads to what I consider at worst dubious and at best hard to interpret results.

        In my opinion, the mining of the ChIA-PET data is an example of this (I apologize in advance for any offense! Singling out this is simply because of my recent rereading of a paper so it’s fresh in my memory — Li G, et al. Cell 2012 vol. 148, 84-98). The cell lines used have screwed up karyotypes (HeLa is included!!) and yet the data has been plunged a few times for fairly high impact publications with little consideration of how the altered copy number impacts what are major conclusions in regards to our model of transcriptional regulation. A naive reader may gloss over this.

        @littlegreyrabbit, by “nixing” I meant I consider it a major point that the authors need to adequately address before publication.

      3. @littlegreyrabbit

        1. Cell lines have quite distinct morphologies.
        You have excellent eyes that I envy.

        2. You have to be pretty careless to cross contaminate.
        Or be tired. Or use a bottle of trypsin/media that had been previously used by someone careless or tired. Etc. Etc. It sounds like you are excellent at tissue culture. But for most of the rest of us contamination of all sorts is a serious issue. It doesn’t take long for a HeLa cell or two to take over a culture.If HeLa cells are in a lab, they can be a contamination source. In my labs we avoid HeLa cells and take many precautions when we do use them, but never discount the possibility of their contaminating less aggressive lines.

        3. Most labs don’t grow HeLa cells anyway.
        There are over 10,000 citations of HeLa cells in 2011.

        Is HeLa cell contamination an issue? If you look at it compared to something line mycoplasma contamination, which I would suggest creates significantly more issues in the literature, than probably not. To discount HeLa cell contamination as a problem…I’m not so sure.

  4. This is an old, old, old story but change is slow unfortunately. The culprit here is inter-lab transfers and poor cell culture practices but, unlike some other methodological issues in research, there is an easy fix. Either buy cells from ATCC, which validates many human cell lines using micro-satellite genotyping, or perform the genotyping yourself and show the data to the journal. These are practical, simple and low-cost solutions. I think ATCC even has a database for researchers to use when checking their cells. The journals (cancer journals in particular) should have no excuse for not implementing these measures. (btw a related and potentially bigger problem is the issue of mycoplasma contamination in cell cultures).

    I cannot stand the tired old debate about “cell lines are rubbish and in vivo is everything”. There are just as many, if not more, limitations to human in vivo studies as there are in cell culture models. Many of the failures of clinical trials are not just related to the pre-clinical research but rather the limitations of human studies. Nobody is debating that the end point should always be human application, but cell culture studies are valid, useful and absolutely necessary. Let’s not turn this into another “can we trust blah blah blah….”

    1. I saw a recent high profile study in which a certain cell line was used. On the ATCC website they make it clear that this line is contaminated with HeLa cells. Presumably this work used the contaminated line. It is unlikely the reviewers or even the scientists were aware of this problem.

      1. Care to cite the paper?

        In general, they have lots of cell lines that are contaminated with various things. The point is they are transparent about it (if they know) and it is up to the scientists as to whether to use the line or not (and whether to disclose its issues in print). I don’t know about anyone else on this blog, but whenever I have brought purely technical issues up in a manuscript review, the editors of the journal were completely uninterested. I think this is a problem.

  5. And of course you don’t even mention the problem of cell lines contaminated with rickettsia, XMRV and its ilk, God knows what else, and everyone thinks this garbage is pristine. PUHLEEEZE!

    1. In reply to Dave April 23, 2012 at 2:03 pm.

      “whenever I have brought purely technical issues up in a manuscript review, the editors of the journal were completely uninterested. I think this is a problem.”

      Why not substitue “scientific” for “technical”, it is about the data?

      I agree wth what you write.

      I noticed that you can publish on odd things you find about your cells:

      An example:

      Placenta. 2002 Jan;23(1):32-43.
      ED(27) trophoblast-like cells isolated from first-trimester chorionic villi are genetically identical to HeLa cells yet exhibit a distinct phenotype.
      Kniss DA, Xie Y, Li Y, Kumar S, Linton EA, Cohen P, Fan-Havard P, Redman CW, Sargent IL.

      “Inasmuch as HeLa cells have never been grown in the laboratory (DAK), the only possible origin of HeLa cell contamination of ED(27) cells was the WISH cell line, and further PCR analysis revealed that this cell line was also genetically identical to HeLa.”

      You might expect that people writing about the amnion, not totally unrelated to the placenta, might read such articles:

      Curr Drug Targets Immune Endocr Metabol Disord. 2003 Mar;3(1):83-92.
      WISH cells as a model for the “in vitro” study of amnion pathophysiology.
      Pavan B, Fiorini S, Ferretti ME, Vesce F, Biondi C.

      1. @Dave,

        @conradseitz talks about the blogging about retractions which seems to be taking off. see Last sentence of the blog entry.

        And I agree.

      1. Perhaps, but this is a problem that comes up every couple of years, about the generation-time of a new PhD graduate. It bears repeating at such a frequency because each generation is often taught to disregard the HeLa contamination issue or say, “it couldn’t happen to us.” I’ve watched folks pooh-pooh this since I started doing cell culture in 1985.

        And while some may find the issue not being new, it continues to be disregarded as evidenced by the Oral Oncology retraction.

        By the way, for the cell culture and molecular oncology enthusiasts among you, search the WSJ archives for all work by Amy Dockser Marcus. She’s done some great writing on cell lines and rare cancers, particularly the story of a now-deceased Tulane graduate student who isolated his own sinonasal undifferentiated carcinoma line, and another on her own lovely mother who passed from gall bladder cancer a few years ago. That one was on my lab door for several years.

      2. Hi David,

        Thanks for your informative comments on cell line that this now deceased student generated. WOuld you please give some reference or article where I can read more about this cell line and gallbladder cell line as I am interested in gallbladder cancer and all the cell lines that have been generated from this cancer. Thanks again.


  6. In reply to John Travis April 24, 2012 1:53 pm and Dave at 5:12.

    I may not follow exactly what you are saying. WSJ = Wall Street journal, at the top of the post.

    The issue of cell contamination, or of anything not being what we think it is, is not new, but that does not mean it is not important. The problem is not going away any time soon. The problem is not going to simpler, or easier any time soon.

    It is topical becasue it came up in the retraction mentioned in this post, and will likely lead to many more quite soon (that is simply a prediction).

    I would say that something does need to be about the problem, not just writing about it (what I am doing now!). It is nice that a nice appeared in Science, but that was obviously not effective on its own, and is not the end of the matter.

    This may be melodramatic, but it is also obvious:

    “It is the duty of righteous men to make war on all undeserved privilege, but one must not forget that this is a war without end.”

    Primo Levi

    Science 16 February 2007:
    Vol. 315 no. 5814 pp. 928-931
    DOI: 10.1126/science.315.5814.928

    * News Focus

    Cases of Mistaken Identity

    1. Rhitu Chatterjee*

    + Author Affiliations

    Rhitu Chatterjee is a science writer in Washington, D.C.


    For decades, biologists working with contaminated or misidentified cell lines have wasted time and money and produced spurious results; journals and funding agencies say it’s not their job to solve this problem.

  7. The quite influential magazine (many people who are not scientists read it) “The Atlantic” is picking up on the story:

    ‘Oops, Wrong Cancer’: How Contaminated Cell Lines Produce Bad Research

    I don’t think it is meant a little jokey piece.

    It may be back to business as usual, but people with non-science degrees may start to notice, or might keep fussing about their museli mix (one is never sure how they will come down in an argument. Good science research, or museli mix? That’s a hard call).

    1. Where can I find a good museli mix? I’ve been eating great grains, which probably doesn’t qualify…

Leave a Reply

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

This site uses Akismet to reduce spam. Learn how your comment data is processed.