The Cancer That Changed Type

Part three takes the blood fluke and the question underneath the whole season. Every textbook says schistosomal bladder cancer is squamous cell carcinoma. At the National Cancer Institute in Cairo, across 9,843 patients in three windows between 1970 and 2007, squamous carcinoma fell from 75.9% of bladder cancers to 28.4% while urothelial carcinoma rose from 16.0% to 65.8%, egg positivity in the specimen fell from 82.4% to 55.3%, and the median age rose from 47.4 to 60.5 years. Three more Egyptian series show the same flip, and by 2019 one of them put squamous carcinoma at 13.8%. A sentence in the present tense turns out to be a 1970s fact — and the fact that it stopped being true is better causal evidence than any odds ratio in the subject. Then we measure the effect honestly: the best-adjusted case-control study in the endemic population returned 1.72, with a confidence interval touching 1.0, against 6.6 for current smoking in the same study, and tobacco explained roughly 75% of male cases where the worm explained about 16%. There is no pooled estimate for a Group 1 carcinogen, because no meta-analysis exists. A Danish study of 12,271 bladder cancers, in a country with no schistosomiasis, found ten or more courses of urinary-infection antibiotics associated with squamous bladder cancer at an odds ratio of 11.4 — chronic irritation selects the histology, and the worm is one way to get decades of it. Then the mechanisms, each with its evidence class: N-nitroso compounds made largely by bacteria the infection recruits, with O6-methyldeoxyguanosine measured in 44 of 46 Egyptian bladders against 4 of 12 European ones; oxidative and nitrative lesions in human tissue, and why the antioxidant inference died in a 35,533-man trial; a worm protein that carries a nuclear address into human bladder cells, at the strength its own authors claim; and the mouse experiment in which eggs alone produced nothing and eggs plus a nitrosamine produced squamous metaplasia. The second half answers the question people actually mean. Four routes from a parasite to a human mutation, in descending order of evidence: host chemistry, the host's own DNA-editing enzymes, a parasite protein that holds the host's controls off — which is real, is reversible, and is a cattle parasite — and parasite DNA entering the human genome, which rests on a paper Cell retracted in September 2005 over its authors' objection, and on thirteen reported schistosome transfers that a reanalysis could not substantiate. No study has reported parasite DNA integrated into a human tumor genome. What does cross into human cells is protein and RNA in extracellular vesicles: regulation, not inheritance.

Show notes

Begin in a record room at the National Cancer Institute in Cairo. In 2007 Gouda and colleagues published what was in it: 9,843 bladder cancers treated at that one institute between 1970 and 2007, in three windows. Series A, 3,212 patients, 1970 to 1974; Series B, 3,988, 1985 to 1989; Series C, 2,643, 2003 to 2007. Three photographs of the same disease, in the same hospital, across thirty-seven years. The disease in the third photograph is not the disease in the first.

Here is what moved. Squamous cell carcinoma — cancer arising from flat, skin-like cells, the kind that covers your skin rather than the kind that lines your bladder — made up 75.9 percent of those bladder cancers in the 1970 to 1974 series. In the 2003 to 2007 series, 28.4 percent. Over the same span transitional cell carcinoma, which arises from the bladder's own lining and which pathologists now more often call urothelial carcinoma, rose from 16.0 percent to 65.8, becoming, in the authors' words, the most common tumor type. The two curves crossed.

Four more numbers moved with them, and they are why the first two matter. Bilharzial egg positivity — the pathologist finding Schistosoma eggs in the removed tissue — fell from 82.4 percent to 55.3. Median age at diagnosis rose from 47.4 years to 60.5. The male-to-female ratio fell from 5.4 to 3.3. Bladder cancer's share of all cancers at the institute fell from 27.63 percent to 11.7. Egypt's signature cancer became rarer, older, less lopsidedly male, and changed cell type, all at once (Gouda and colleagues 2007).

Be exact about that study. It is a retrospective review of one tertiary referral institute's case mix, and case mix is not incidence: who gets referred to a national cancer institute, and how a pathologist classifies what arrives, both moved over those decades. Schistosomiasis status was judged by egg positivity in the specimen, a test that misses infections rather than inventing them. The paper appeared in the Journal of the Egyptian National Cancer Institute, whose record carries no digital object identifier — a small, local, little-cited journal, which is the real point. The authors attribute the decline to better control of bilharziasis in rural Egypt. That is their inference, not a measured effect of any program.

And with every one of those caveats standing, this is the strongest causal evidence parasite oncology has ever produced. What you can get, if a country drives an exposure down across a generation, is the next best thing: the exposure falls, the signature disease falls with it, on the right clock, and the exposure marker falls in the same specimens. That is a reversal, and it is worth more than any case-control study: an association can be manufactured by bias in a dozen ways; a reversal has to be manufactured twice, in opposite directions, by the same bias.

Which means the textbook sentence has quietly expired. Open most pathology references and you will read, in the present tense, that schistosomal bladder cancer is squamous cell carcinoma. A 1970s fact. In Cairo, in the most recent window of the largest series anyone has assembled, most bladder cancer is not squamous — and the one series that cross-tabulated histology against the worm, Amin and colleagues 2019, found only one of nineteen schistosomiasis-confirmed cancers squamous. The error has a name: reading a historical case mix as a current fact. The sentence ceasing to be true is the field's best evidence.

So this hour has two halves. The first is the bladder: how strong the link is when somebody measures it, what the four Egyptian series show, and a control experiment in Denmark nobody designed. The second is chemistry — how a worm reaches human DNA at all. That half exists to test a premise: that Schistosoma haematobium gets its own genetic material into a bladder cell and rewrites it. There is no peer-reviewed evidence that it does. Hold the premise.

The animal, briefly: the Atlas page for Schistosoma haematobium carries the life cycle in full. It is a blood fluke: a parasitic flatworm whose adult male and female live clasped together inside human veins. Freshwater snails — for this species those of the genus Bulinus, which that page carries — shed a free-swimming larva called a cercaria, and the larva burrows through unbroken skin: no bite, no wound, no swallowing (Buonfrate and colleagues 2025). The pair settles in the vesical venous plexus, the net of small veins around the bladder, which is where the eggs have to go. The adults, says Colley's 2014 Lancet review, live in human blood vessels for years, shedding hundreds to thousands of eggs a day.

Now the sentence this whole part turns on. Those eggs, says Colley, must either leave the body in excreta or become trapped in nearby tissues — and trapped eggs induce a distinct immune-mediated granulomatous response. A granuloma is a tight ball of immune cells the body builds around something it can neither digest nor expel. It is the eggs that do the damage, and the host's reaction to them that does most of it. Not the adult worms. And not dying worms either, an error worth naming: the story in which the parasite dies and dumps a storm of inflammation and DNA damage into the bladder. Worm death does release antigen, which is why treatment can provoke a reaction. But the engine is a stationary egg and the scar around it, for decades.

How many people carry this is worth doing honestly: the field quotes one confident number and has none. At least 230 million infected worldwide (Colley and colleagues 2014). 210 million in 76 countries, says the S. mansoni genome paper (Berriman and colleagues 2009). More than 100 million for S. haematobium alone (Stroehlein and colleagues 2019). Over 250 million afflicted (Park and colleagues 2021 — four of whose authors are employed by Merck KGaA, which manufactures it). More than 290 million threatened, per the World Health Organization as quoted by Weglage and colleagues in 2020. Different years, different species, and in the last case a different thing entirely: threatened, not infected.

And treat modeled estimates with caution: somebody checked one against the ground. Yang and colleagues, in the BMJ in 2025, compared reported real-world data from China against the Global Burden of Disease 2021 estimates for six neglected tropical diseases, 2004 to 2020, in disability-adjusted life years — a unit that adds years of life lost to years lived with impairment. For schistosomiasis the modeled estimate ran about 1.5 times the reported figure: the second closest of the six, with only rabies, at 1.4, tracking reality more nearly. For leprosy the ratio was 17; for visceral leishmaniasis, 280.

Then the tissue, where the squamous answer comes from. Ahmed and colleagues, in 2017, examined bladder biopsies from 54 patients at Sohag University Hospital in Egypt, with mid-stream urine checked for eggs. Squamous metaplasia — the bladder's proper lining replaced by flat, skin-like squamous cells in response to long-term irritation, a change of cell type and not yet a cancer — was present in 38 of 54, 70.4 percent. Non-keratinizing in 20, 37 percent; keratinizing, the cells producing keratin, in 18, 33.3 percent. Invasive squamous carcinoma in 11, 20.4 percent. Now the weakness: one hospital, six months, no uninfected comparison group, a small local journal with no identifier on its record. A referral series, not a population prevalence.

The fuller ladder comes from a weaker source still. Sabe and colleagues, in 2008, in the same obscure Egyptian parasitology journal with no identifier on its record, stained 140 previously diagnosed schistosomiasis bladder lesions — 124 precancerous, 16 cancers as controls — and ran them through an image analyzer. Their flat, non-papillary sequence: reactive atypia, flat hyperplasia, keratinizing squamous metaplasia, glandular metaplasia, dysplasia — cells that already look abnormal — and carcinoma in situ, cells that look cancerous but have not yet invaded the tissue beneath. Non-diploid DNA was found in 53 percent of the carcinoma in situ lesions, while the three rungs below read as diploid. Nobody followed a bladder up that ladder. But its first rung is why the tumor at the top came out squamous.

How strong the link is, and the control nobody designed

So how big is the risk? Somebody measured it. Bedwani and colleagues ran a case-control study in Alexandria from January 1994 to July 1996 — a design that starts from the disease and works backward: gather the people who have the cancer, comparable people who do not, and compare their exposures. The cases were 190 subjects with incident, histologically confirmed invasive bladder cancer; the controls, 187 people admitted to hospital for acute, non-neoplastic, non-urinary-tract conditions. Eighty-six of the cases, 45 percent, reported a history of urinary schistosomiasis; 69 of the controls, 37 percent, reported the same.

Forty-five against thirty-seven. Everything follows from how small that gap is. The headline figure is a multivariate odds ratio: how many times more common the exposure is among the sick than among the well, adjusted for age, sex, education, smoking, other urinary infections and occupation. It came out at 1.72, 95 percent confidence interval 1.0 to 2.9 — the range the data cannot rule out, its lower edge sitting exactly on 1.0, which means no effect. The association Ferguson described in 1911, which Part 001 tells in full, got its best-adjusted test eighty-seven years later, and that is what it returned. The authors' own sentence: a clinical history of urinary schistosomiasis is significantly, but modestly, associated with increased bladder cancer risk, explaining some 16 percent of cases in this Egyptian population — a population attributable fraction, the share that would not have happened without the exposure. One case in six.

The same paper holds a result that cannot be true. Intestinal schistosomiasis — the sibling species that settle in the veins of the gut rather than around the bladder — came out at an odds ratio of 0.22, interval 0.1 to 0.4: a different worm in a different vein cutting your bladder cancer risk by four-fifths. What it reveals is how exposure was measured — not by finding eggs in urine, but by asking people what they had once been told they had. The error has a name: taking a protective-looking odds ratio out of a self-reported exposure and calling it a benefit.

Now the companion paper: same study, same city, same investigators. Bedwani and colleagues 1997 took 151 male cases and 157 male controls and asked about tobacco. Against never-smokers, ex-smokers gave an odds ratio of 4.4, interval 1.7 to 11.7; current smokers 6.6, interval 3.1 to 13.9; more than forty years of smoking, 16.5. Their conclusion: cigarette smoking could explain 75 percent of bladder cancer cases among males from Alexandria. Sixteen percent for the parasite. Seventy-five for tobacco. One difference between the papers runs in the direction that helps that comparison, so say it: the tobacco paper's controls were admitted for conditions that were not only non-cancerous and non-urinary but non-smoking-related, which pushes a smoking odds ratio up. The gap is real; it is not quite measured on the same instrument.

Then the absence nobody mentions. Search the systematic-review literature for schistosomiasis and bladder cancer and you find reviews. Not one pools a risk estimate for infection against cancer. A meta-analysis is the arithmetic that combines separate studies into one number; for this pairing nobody has run it. The nearest thing runs it on something else: Koonrungsesomboon and colleagues 2015 pooled odds ratios for markers inside the tumors across 63 articles — p53 expression at 9.46, interval 1.14 to 78.55, a range that declares the data nearly empty. And it compares schistosomal bladder cancer against other schistosomiasis patients, never what the infection does to your risk. A Group 1 carcinogen with no headline effect size behind it. The Atlas page on the World Health Organization's cancer agency makes the distinction: its ratings say how sure the evidence is, never how much.

Back to the reversal, where the real evidence lives. Four series report it. Gouda 2007 was the first. Felix and colleagues 2008 abstracted all available medical records at the National Cancer Institute of Cairo University for six calendar years between 1980 and 2005, and found a significant association between the period of diagnosis and the histopathological type: patients diagnosed in 2005 had sixfold higher odds of transitional cell carcinoma than patients diagnosed in 1980, odds ratio 6.00, interval 4.00 to 8.97. One correction, mine rather than theirs: those two are not independent. Felix and Gouda drew on the same institute, so the same patients and years sit in both. Four series, three institutions.

The third is a different hospital. Salem and Mahfouz reviewed 1,932 patients at Kasr Al Aini Hospital, Cairo University: 1,002 from 2001 to 2005, and 930 from 2006 to 2010. Squamous cell carcinoma fell from 73 percent to 25, transitional cell carcinoma rose from 20 to 66, associated bilharziasis fell from 80 to 50. Mean age rose from 41 years, give or take 11.2, to 52, give or take 8.6 — eleven years older in nine years of calendar time. And no difference between the groups in tumor stage, grade or lymph node metastases: how far it had spread, how abnormal the cells looked, whether it had reached the draining nodes. It changed cell type and age without becoming milder.

The fourth says it loudest. Amin and colleagues 2019 report 79.3 percent transitional cell carcinoma, a further 6 percent with associated squamous features, and squamous cell carcinoma only 13.8 percent. And the detail that should stop a listener: schistosomiasis was histologically confirmed in 19 of the cancer cases, and only one of those 19 was squamous. Their conclusions are two. The old concept that schistosomiasis is associated with squamous cell carcinoma should be re-evaluated, as most cases are associated with transitional cell carcinoma. And relying on histopathology to confirm schistosomiasis appears to be non-accurate and leads to irrelevant results — which makes every historical count of these cancers uncertain, in both directions.

Now the discipline, because all four share a weakness. Three are institutional case mix over time — the proportions of tumor types arriving at a hospital, not the rate of cancer in a population — and the fourth, Amin, is a single recent snapshot read against the older three. Case mix moves with referral patterns, and the attribution everyone makes, that mass treatment with praziquantel did this, is inference. The study that would settle it does not exist: an interrupted time series, the same rate before and after a program starts, of age-standardized bladder cancer incidence from a population-based registry, set against documented coverage of mass drug administration — treating whole communities on a schedule without testing individuals first. And Amin's argument that smoking cannot explain the shift, because lung cancer is not rising proportionately, is an ecological inference: individuals judged from population totals.

And there is a reading that runs the other way, which has to be said on air. Tawfik, writing in 1987 from the pathology department of that same Cairo institute, reported that in Fayoum Province schistosomiasis was decreasing while bladder cancer was increasing — one sentence, no figures attached, in a symposium volume with no identifier on its record, which is exactly how much weight it carries, and that is not zero.

Which leaves one question standing. The answer arrived by accident, in Denmark, from people not studying parasites. Pottegård and colleagues 2020 ran a nationwide case-control study: 12,271 histologically verified bladder cancers diagnosed between 2000 and 2015, each matched to cancer-free controls of the same age and sex. Denmark has no schistosomiasis. What it has is a prescription registry, so the investigators used prescriptions for urinary-tract-specific antibiotics as a stand-in for having had the infections, ignoring the two years before diagnosis — a lag to blunt reverse causation. Of the 12,271 cancers, 333 were squamous. 2.7 percent.

Here is what those 333 showed. Against people with no more than one such prescription, people with ten or more had an odds ratio for squamous bladder cancer of 11.4, 95 percent confidence interval 7.6 to 17.2, with a clear dose-response pattern at p below 0.001 — risk climbing step by step as the prescriptions accumulated. Excluding patients with known urogenital disease barely moved it: 10.8, interval 6.2 to 18.9. For urothelial carcinoma, 11,029 of these cancers, the same exposure gave 1.13, interval 0.97 to 1.32. Essentially nothing. And phenoxymethylpenicillin, an antibiotic not used against urinary infections, showed no association — a negative control built into the study.

The weaknesses, stated: the exposure is a prescription and not a cultured organism, and there were no data on smoking, though the authors' bias analysis suggested that mattered little. The pooled literature is weaker: Bayne and colleagues 2018 pooled eight case-control studies and found urinary infection raising the risk of non-schistosomal bladder cancer at a relative risk of 1.33, interval 1.14 to 1.55 — which stopped being significant once they dropped the studies published before 2000 and those at high risk of bias. Which fits: bladder cancer overall is overwhelmingly urothelial, and urothelial is the histology Pottegård found nothing for.

Now the two numbers in one sentence, caveat attached. Chronic bladder irritation in Denmark: 11.4. Urinary schistosomiasis in Alexandria: 1.72. About six times larger for the exposure that has nothing to do with a worm. Not a head-to-head — two countries, two designs, and a prescription registry is a better instrument than a patient's memory. But the direction is not subtle. What selects the squamous phenotype is decades of chronic irritation of the bladder lining, and Schistosoma haematobium is an unusually durable way of producing it: a stationary egg, a granuloma around it, for twenty or thirty years. One route to a common endpoint; in Denmark, recurrent infection is another.

Keep the Western baseline in view. Maia and colleagues 2019 put non-urothelial histologies at less than 10 percent of all bladder cancers, with squamous cell carcinoma the most common of those at approximately 5 percent — and they divide into the schistosoma-related and the non-schistosoma-related, the second commoner in the Western world. Squamous bladder cancer is a small minority everywhere. The Egyptian series where it made up three-quarters of cases were not describing ordinary bladders, but what a bladder does after thirty years of injury.

The chemistry, the radicals, and what the worm itself makes

So the worm buys decades of chronic irritation. What changes the DNA? Three answers, best first — and the best-measured is not the worm but chemistry made by the patient's own bacteria. The canonical synthesis is Mostafa, Sheweita and O'Connor, 1999, in Clinical Microbiology Reviews — a review, not a measurement. N-nitroso compounds, a family that includes the nitrosamines, were at high levels in the urine of these patients. Bacteria capable of nitrosation, the reaction that builds them, were found in infected urine at higher intensities of infection than in normal subjects. In infected animals the enzymes that chemically alter foreign compounds, sometimes arming them, gain activity early and lose it in chronic disease, which could prolong exposure to activated nitrosamines. And host DNA damage above normal has been observed, specifically alkylation — a small carbon group stuck onto a DNA base — with inefficient repair.

The human numbers are rarely quoted. Abdel Mohsen and colleagues, 1999, collected twenty-four-hour urine from 61 Egyptians: controls, infected people, and bladder cancer patients with and without a history of infection. Figures run per milligram of creatinine, which corrects for dilution. Nitrite, the ion that feeds nitrosation, was below the detection limit of 0.015 micrograms per milligram in all but one control. Infection raised it to 0.9 plus or minus 1.16, p equals 0.001; in both cancer groups it was about twenty times that again. Total N-nitroso compounds tracked nitrite, correlation 0.71. N-nitrosodimethylamine ran 4.02 plus or minus 1.61 nanograms per milligram against 2.04 plus or minus 2.97 in controls, p equals 0.01. Sixty-one people, standard deviations as wide as the means: a dose signal, not a dose curve.

Then the lesion. Badawi and colleagues, 1992, used a radioimmunoassay to look for O6-methyldeoxyguanosine in bladder DNA from Egyptian patients with bladder carcinoma and concurrent schistosomiasis — alkylation made concrete, a methyl group bolted to the oxygen at position six of guanine. They found it in 44 of 46 samples, 96 percent — the 46 being 38 specimens of tumor tissue and 8 of uninvolved bladder lining — at a mean of 0.134 micromoles per mole of ordinary deoxyguanosine, up to 0.485. In normal bladder tissue of European origin, 4 of 12, mean 0.046. Twelve European bladders, not uninfected Egyptians, on 1992 methods. Tumor and uninvolved lining were methylated to similar extents, which the authors said suggests the alkylating intermediate may have been present in the urine. Not made by the tumor. Sitting in the urine.

Why does one methylated base matter? O6-methylguanine pairs with thymine during replication, writing a guanine-to-adenine change into the new strand, and cells left carrying that mispair normally kill themselves by apoptosis, which requires the mismatch repair complex that spots wrongly paired bases (Fujikane and colleagues, 2016). So this arm has a carcinogen in urine, a dose rising with infection and again with cancer, a lesion in human bladder DNA, and the mutation that lesion makes. Nowhere else do all four line up in human tissue. And the carcinogen is largely made by bacteria colonizing a damaged, obstructed bladder. The worm's contribution is the bladder.

A second bacterial arm. Gentile and colleagues, 1985, infested hamsters with S. haematobium: raised beta-glucuronidase, the enzyme that snips the sugar group off compounds the body has tagged for disposal, handing a carcinogen back in active form, and urine concentrates that were not mutagenic themselves but enhanced the mutagenicity of other chemicals on the Ames test. The infection need not supply the carcinogen to make whatever else you meet worse.

Mechanism two is oxidative and nitrative damage, measured in human bladder. Salim and colleagues, 2008, stained bladder carcinomas from Egyptian patients with and without infection. 8-hydroxy-2'-deoxyguanosine, a guanine carrying an extra oxygen and the standard footprint of oxidative DNA damage, was markedly elevated in both the squamous and the urothelial carcinomas associated with schistosomiasis. With it came overexpression of the two enzymes that cut that base out, 8-oxoguanine DNA glycosylase and apurinic/apyrimidinic endonuclease; more single-strand breaks; and higher inducible nitric oxide synthase, the enzyme inflamed tissue switches on to make nitric oxide. Antibody staining scored by intensity: direction without numbers.

Ma and colleagues, 2011, added the nitrative half. 8-nitroguanine, the footprint nitric oxide chemistry leaves on DNA, and 8-oxodG were both significantly more intense in the cancer and the cystitis tissue — cystitis meaning an inflamed bladder — of infected patients than in normal tissue. Inducible nitric oxide synthase sat in the same cells as nuclear factor kappa B, the master switch for inflammatory genes, and the damage was forming inside cells positive for Oct3/4, a protein stem cells make. Kawanishi and colleagues, 2017, report the same lesion across the Group 1 infectious carcinogens: in Oct3/4-positive stem cells in schistosomal bladder cancer, and in Oct3/4- and CD133-positive stem cells in the liver-fluke bile duct cancer of this season's second part. And one correction before leaving this arm: do not attribute an 8-oxo-dG finding to Mostafa 1999, whose damage type is alkylation. The error has a name: lesion drift, letting one kind of DNA damage stand in for another because both are called damage.

Now shut the door onto the supplement aisle. Reactive oxygen species, what people mean by free radicals, have no single direction of effect. Contingent upon concentration, as Hayes, Dinkova-Kostova and Tew put it in their 2020 review, they influence cancer in apparently contradictory ways, either initiating and stimulating tumor formation or killing cells. The largest trial built on that inference failed in the wrong direction. The Selenium and Vitamin E Cancer Prevention Trial randomized 35,533 men; Klein and colleagues, 2011: 620 men on vitamin E at 400 international units a day developed prostate cancer, against 529 on placebo. Hazard ratio 1.17, 99 percent interval 1.004 to 1.36, p equals 0.008. Inflammation damages DNA. Therefore take antioxidants is the sentence the data killed.

Mechanism three is the one people want to be true and the weakest of the three; label it that way. What does the worm itself make? The solid half is IPSE, the interleukin-4-inducing principle of Schistosoma eggs. Pennington and colleagues, 2017, cloned two S. haematobium genes, H03-IPSE and H06-IPSE, paralogs of a protein already known from S. mansoni. Expression is restricted to the egg stage and female adult worms. Each carries a conserved nuclear localization sequence at its tail, a short run of amino acids that works as a shipping label for the nucleus: SKRRRKY and SKRGRKY. Added to the medium of HTB-9 human bladder cells, both get inside and those labels carry them to the nucleus; delete the label and they do not go. In cultured cells.

Mbanefo and colleagues, in Infectious Agents and Cancer in 2020, went further. Eggs injected into the mouse bladder produced angiogenesis, the growth of new blood vessels, leaky vessels, and urothelial hyperplasia. Recombinant wild-type IPSE raised endothelial and urothelial proliferation and skewed urothelial cells into S phase, where DNA is copied; a mutant lacking the native nuclear localization sequence did not. The authors' own word for IPSE is candidate. Two co-authors work at a cancer diagnostics company and one at a drug company.

And a fabrication to kill by name. There is no locus called Sh-ipse-1 and no chromosome 3 assignment for IPSE anywhere in the indexed literature. The error has a name: a fabricated genomic address, a coordinate invented for a real gene, convincing because the gene exists. The S. haematobium reference genome of that era was a 385-megabase draft sequenced at 74-fold coverage, with no chromosome-level assignment of individual genes at all (Young and colleagues, 2012). Schistosomes do have eight chromosome pairs, males carrying two Z chromosomes and females a Z and a W (Stitz and colleagues, 2021). Not an impossible object. Simply not where anyone published IPSE.

Then the estrogen hypothesis at its real strength. Santos and colleagues, 2014, studied 93 women and girls in Angola: catechol-estrogens and their DNA adducts, estrogen breakdown products chemically stuck onto DNA, were associated with schistosomiasis at an odds ratio of 3.35, interval 2.32 to 4.84, and with self-reported infertility at 4.33, interval 1.13 to 16.70. Gouveia and colleagues, 2015, found seven estrogen-like metabolites in urine from 40 Angolan patients, absent from healthy human urine. And Botelho and colleagues, 2013, found soluble egg antigen at 6.25 micrograms per milliliter raising proliferation, reducing apoptosis and registering genotoxicity on the alkaline Comet assay in cultured urothelial cells. Then the limits. The depurinating estrogen-DNA adducts that initiate cancer come from estradiol-3,4-quinone, not the estradiol-2,3-quinone usually named (Cavalieri 2011; Cavalieri and Rogan 2016). The earliest tumor-like phenotype experiments were done in CHO cells, Chinese hamster ovary cells, not human urothelium (Botelho and colleagues, 2010). And every paper in this arm traces to one consortium, in Porto and Washington, whose own verbs are the honest ones: we speculate, Botelho writes in 2015; we postulate, Correia da Costa writes in 2014. No outside replication, and no experiment in which blocking the pathway prevents a tumor.

Now the animal experiments, and what they show the eggs cannot do alone. Fu and colleagues, 2012, microinjected purified S. haematobium eggs into the mouse bladder wall. The mice develop macrophage-rich granulomas lasting at least three months, pass eggs in their urine, and develop urinary tract fibrosis, bladder dysfunction and urothelial change resembling the human disease, with a Type 2-dominant response of high interleukin-4, eosinophils and IgE — immunoglobulin E, the antibody class worms provoke.

The most pointed one is Chala and colleagues, 2017. Mice were given eggs, eggs plus one of two chemical bladder carcinogens, and controls. Squamous metaplasia appeared in the eggs-plus-N-nitrosodimethylamine group at week 12 and, in the authors' words, not in other groups. Ki-67, a protein present only in dividing cells, rose significantly with eggs plus N-butyl-N-(4-hydroxybutyl)nitrosamine at week 20. Eggs alone did not do it. And the limit out loud: the abstract reports no nitrosamine-only arm, so this shows the eggs were not sufficient, not that the nitrosamine alone would have failed.

And much of what the eggs do is the host doing it. Mbanefo and colleagues, in Infection and Immunity in 2020, repeated the injection in mice bred without a working receptor for interleukin-4. Without that signal the granulomatous response was significantly less intense, and the urothelial proliferation, the hyper-diploidy and the cell-cycle skewing did not happen.

One last correction. The claim in circulation is that trapped eggs release metabolites that inactivate the p53 gene, switching off cell suicide so damaged cells survive. No published work shows that. The error has a name: an invented mechanism bridging two true facts. Egg products do suppress apoptosis and damage DNA, which is Botelho 2013, in cultured cells. And TP53 is mutated in these tumors: Gonzalez-Zulueta and colleagues, 1995, found p53 mutation in squamous bladder carcinomas at a frequency similar to the roughly 60 percent reported for invasive transitional carcinoma. Mutated by a chemical is not switched off by a metabolite. And in the mouse model p53 is manipulated by the experimenter: the protocol uses a genetic switch, Cre recombinase, to delete one copy before the eggs go in (Mbanefo and Hsieh, 2018).

Inside the tumor, the drug, and the rest of the genus

What is actually broken inside one of these tumors? Gonzalez-Zulueta and colleagues, 1995, looked at squamous bladder carcinomas from nineteen Egyptian specimens and twelve Swedish, and the headline is chromosome 9, not p53. CDKN2, the gene that makes the brake protein p16, carried a homozygous deletion or a sequence mutation in 67 percent of them, eight of twelve: three times the frequency reported for uncultured transitional cell carcinomas, uncultured meaning tumor tissue read directly rather than grown in a dish. Deletion of 9p, where that gene sits, appeared in 92 percent, eleven of twelve, against about 39 percent of transitional carcinomas, thirty-five of ninety. p53 mutation ran near the roughly 60 percent reported for invasive transitional carcinoma, at different positions; loss on 17p, 38 percent against 60. Those percentages rest on about a dozen specimens — and whole-arm loss is a karyotype-scale event a hotspot sequencing panel never asks about, which the Atlas page on karyotypes and gene panels carries in full.

Then the cost. Immune checkpoint blockade — drugs that block PD-1 and its partner PD-L1, a brake the tumor puts on the patient's own T cells — transformed advanced urothelial carcinoma. Madureira, 2022, states the consequence in one line: in general, divergent histologies are ineligible for therapy. Divergent, or variant, histology means a bladder tumor that is not conventional urothelial carcinoma, the class the schistosomal squamous tumor falls into. The same review, in a current immunology journal, still puts squamous carcinoma at 53 to 69 percent of bladder carcinomas in endemic regions — the 1970s number, in the present tense, in the literature of four years ago: the error this hour opened on, surviving in print. And Maia's Western figures put these patients under a tenth of bladder cancers, under-represented in trials, their treatment extrapolated from urothelial disease.

Finding the worm is a separate problem, and the specimen decides the answer. When the whole organ comes out, eggs are everywhere: El-Bolkainy and colleagues, 1981, found schistosome eggs in 902 of 1,095 Egyptian bladder carcinomas treated by radical cystectomy, 82.4 percent — but in an endemic population eggs sit in bladders with cancer and in bladders without it, so eggs in a tumor do not make that tumor schistosomal, and a negative urine is not an absent worm. The part on what reaches the glass takes it specimen by specimen. And one number in circulation belongs to the wrong specimen: 3 percent by cytology against 49 percent by crushed biopsy is Poggensee and colleagues, 2001, finding eggs in the cervix. Cervix, not bladder; a smear, not urine, and those figures belong on the new Atlas page for female genital schistosomiasis.

Then the drug. There is only one. Zwang and Olliaro, 2014, pooled 55 trials and 19,499 subjects. At the recommended 40 milligrams per kilogram the cure rate — the share of treated people with no eggs found afterward — was 94.7 percent for S. japonicum, interval 92.2 to 98.0; 77.1 for S. haematobium, 68.4 to 85.1; 76.7 for S. mansoni, 71.9 to 81.2; and 63.5 for mixed infection, 48.2 to 77.0. Adverse events in 56.9 percent. Name the conflict: the second author worked for the World Health Organization's tropical disease research program, whose dose the paper endorses. And both endpoints are counted in eggs, which miss most light infections.

The pharmacology holds its own reversal. Praziquantel has been in use for decades and, after all of them, nobody knew how it worked until 2021, when Park and colleagues showed that it opens a transient receptor potential melastatin channel, a pore in the worm's membrane that lets calcium through, by sliding into a greasy pocket inside the part of the channel that senses voltage. Calcium floods in; the worm is paralyzed. The drug does nothing to liver flukes of the genus Fasciola, and the reason is a single amino acid: change that residue to match the schistosome's and the channel becomes sensitive. Four of the authors work for Merck KGaA. And one repeated claim does not travel with it: that the drug misses juvenile worms and leaves a reinfection window. No abstract in the file says so. Its account goes on the new Atlas page for praziquantel.

Then the hole at the center of the field: four studies nobody has run. No study tests whether mass drug administration changed bladder cancer incidence; the Egyptian evidence is institutional case mix, and Fayoum ran the other way. No cohort has followed treated against untreated people to a cancer endpoint. Nobody has measured latency; the only proxy is Bedwani 1998, with odds ratios of 3.3 for first diagnosis before age fifteen and 3.0 for thirty-five or more years since. And nothing shows whether squamous metaplasia regresses once the infection is cured — where the first human molecular evidence now points the wrong way. Mertelsmann and colleagues, 2025, sequenced cervical RNA from 20 infected and 19 uninfected Tanzanian women, before and four to twelve months after praziquantel: nine differentially expressed genes with infection against without, twenty-three after clearance, twenty-nine in cleared women against uninfected — most associated with heightened oncogenesis during infection and after it. A pilot study, 39 women, a transcriptome and not an outcome, and the authors say it is unknown whether that resolves with time. Not resolved by cure: enhanced after it.

Now the rest of the genus. Schistosoma japonicum sits in Group 2B, possibly carcinogenic (Jain 2025), where IARC's 1994 monograph put it — Part 001 carries that citation. One correction, since Part 001 owns the rating: you will read that it is now a probable human carcinogen. Both halves are wrong — possibly, not probably, and since 1994, so nothing moved. The famous threefold colon-cancer figure is Qiu and colleagues, 2005, a matched case-control study in rural Sichuan: 142 colon and 127 liver cancer patients against hospital controls with a non-cancer illness. Colon cancer, odds ratio 3.3, interval 1.8 to 6.1, attributable fraction 24 percent. Liver cancer, 3.7, interval 1.0 to 13, attributable fraction 27 percent — and every analysis used only hepatitis-negative matched pairs, because hepatitis so overwhelms liver cancer in China that the authors dropped every pair in which it appeared. Where no written record existed, exposure came from interviewing the subject or a relative.

Inaba, 1984, followed 2,067 residents over thirty in an endemic area of Yamanashi Prefecture from 1958 to 1982: observed deaths against expected ran 4.05 in men and 5.53 in women for liver cirrhosis, 2.30 for liver cancer in men only, and 2.25 for colon cancer in women only — with the cirrhosis and the female colon-cancer ratios, and only those, rising with years of residence before 1957. Qin and colleagues, 2021, set 248 patients with chronic intestinal schistosomiasis at a Wuhan endoscopy center against 992 without: intestinal polyps in 64.5 percent against 42.8; colorectal cancer in women 13.8 percent against 5.4. Everyone in it was already being scoped. Pan and colleagues, 2020, found extra copies of the oncogene c-MYC in 50 of 354 colorectal cancers, predicting worse survival in the schistosomal tumors and not in the others: prognosis in people who already have cancer. Hamid, 2019, calls it a colitis-like subtype. No meta-analysis exists.

Then Schistosoma mansoni, Group 3 on that same 1994 list, not classifiable. Weglage and colleagues, 2020, showed eggs and soluble egg antigen switching on Wnt/beta-catenin signaling — a growth pathway cells use to decide when to divide — and the protooncogene c-Jun, with Cyclin D1 and the DNA-damage markers Parp1 and gamma-H2a.x in gut lining cells, then found the same hallmarks in colon biopsies from infected patients. Pathway activation in hamsters and in dishes, with no cancer outcome measured in anyone. Shousha and colleagues, 2021, found a history of S. mansoni in 688 of 1,446 liver cancer patients at a Cairo clinic; against the study's only non-cancer group, 313 patients with cirrhosis and no liver cancer, prior infection carried an adjusted odds ratio of 1.589, interval 1.187 to 2.127 — a comparison among people who already had liver disease. The infected also cleared hepatitis C less often on treatment, 60 percent against 84.3. Toda and colleagues, 2015, is seven patients in Sao Paulo, all of whom died; it remains unclear, they write, whether S. mansoni alone has carcinogenic potential.

The newest frontier is the same worm in a different organ, where the evidence turns on definitions. Sturt and colleagues, 2025, screened 1,170 publications to April 2024 and found six eligible studies covering 1,081 women in sub-Saharan Africa, each using a different case definition. Their three results on high-risk human papillomavirus ran from 1.9 at baseline for composite-defined disease in Zimbabwe to 1.0 for visually defined disease in Madagascar. Four of the six also looked at cervical pre-cancer: one found 6.08, interval 1.58 to 23.37, three found nothing. Then the largest test. Lamberti and colleagues, 2026, recruited 2,532 Zambian women aged 15 to 50. Molecular disease — Schistosoma DNA on polymerase chain reaction, which copies a target stretch of DNA until a machine can see it — was weakly associated with high-risk HPV, adjusted odds ratio 1.3, interval 0.9 to 1.9, crossing 1.0, and with types 16, 18 and 45 at 1.7. Visually defined disease, much the commoner finding at 35.2 percent against 6.5, showed none at all. Mukerebe and colleagues, 2026, sampled 96 Tanzanian women twice, nine to twelve months apart: infection was associated with persistence of high-risk HPV, odds ratio 4.7, interval 1.3 to 16.5. So the prevalence association is weak to absent in 2,532 women; the persistence signal, the better endpoint because persistent high-risk HPV is the prerequisite to cervical cancer, rests on 96. Not a second carcinogen: a worm that makes an established one harder to clear.

One non-finding to end on. No controlled human study links schistosomiasis to prostate cancer. What exists is a scatter of case reports — seventeen in all, collected by Figueiredo and colleagues in 2014 — and one experiment in a dish: Tuffour and colleagues, 2018, applied soluble egg antigen from egg-positive urine in a Ghanaian community to a cultured human prostate cell line, where 12.5 micrograms per milliliter raised proliferation, P .029, depleted glutathione and reduced apoptosis. We left it out of the case deliberately: one cell line is not a human, and seventeen case reports are not a rate.

How a parasite reaches DNA, and the thing nobody has ever found

Here are the ways a parasite could change a human cell's DNA, best evidence first. Route one is host chemistry the infection recruits: the nitrosamines and nitrated bases of the last two beats. Route two is the host's own DNA-editing enzymes, switched on by the infection and left running. Route three is a parasite protein that holds a host cell's controls off without touching a single base. Route four is parasite DNA entering the human genome. The folklore is almost all route four; the evidence is almost all one, two and three.

Route two's clean experiment is malaria, not worms. A germinal center is a lymph-node compartment where B cells deliberately mutate their own antibody genes at speed, and the enzyme that does it is activation-induced cytidine deaminase, or AID. Robbiani and colleagues, 2015, gave mice chronic Plasmodium chabaudi infection: it prolonged germinal-center expansion and AID expression, and the B cells raised during infection suffered widespread DNA damage leading to chromosome translocations. And infection did not change the overall rate of lymphoma; it shifted the spectrum toward AID-dependent lymphomas carrying translocations. The human version is Torgbor and colleagues, 2014, asking why endemic Burkitt lymphoma tracks the map of Plasmodium falciparum in Africa. The parasite hits germinal-center B cells three ways: deregulating AID, raising the chance of the translocation that switches on the c-MYC oncogene — a gene that drives cell division, where the genes that brake one are tumor suppressors; increasing the number passing through; and raising the fraction carrying Epstein-Barr virus, which shields a cell from the suicide program c-MYC triggers.

And the caution that travels with such claims. Both infections also switch on a second family of host mutator enzymes, the APOBEC3 enzymes. Summerauer and colleagues, 2022: Epstein-Barr virus strongly upregulated enzymatically active APOBEC3B and APOBEC3G in human B cells, and malaria patients had significantly raised APOBEC3A. Overexpressed in a human kidney cell line, APOBEC3A, 3B and 3G all produced c-MYC mutations. And yet in the endemic Burkitt tumors themselves, mutational enrichment appeared only in AID motifs. Upregulating a deaminase is not the same as leaving its fingerprint.

Now route three, the best image in this hour, and it is in cattle. Theileria annulata, a tick-borne single-celled parasite, infects bovine white blood cells and makes them behave like a cancer. Dandasena and colleagues, 2025, sequenced the whole genomes of six infected bovine leukocyte cell lines and found 7,867 exon-linked somatic mutations shared across all six — inside protein-coding stretches, acquired rather than inherited — enriched in the oncogenes FLT4, NOTCH2, MAP3K1, DAXX, FCGR2B and ROS1, and in the tumor suppressors BARD1, KMT2C, GRIN2A and BAP1. Crizotinib, a ROS1 inhibitor, killed the infected cells.

Two findings get misquoted. TP53 is reported as suppressed in those cells, not mutated. And the host's own APOBEC3H is upregulated, offered as a possible parasite-driven mutator mechanism: a host enzyme doing the editing. Cell-line work in cattle, with drug validation for ROS1 alone, and two of the eleven authors at a commercial bioinformatics company.

The cleanest image is older. Haller and colleagues, 2010, found that in Theileria-transformed bovine leukocytes the host's p53 protein sits largely in the cytoplasm, physically associated with the membrane of the parasite's schizont, the intracellular stage that does the transforming. Not degraded. Held. Cure the cell with buparvaquone, which kills Theileria, and p53 translocates back into the nucleus, the pro-death proteins Bax and Apaf-1 rise, Bcl-2 falls, and the cell kills itself. The machinery was never destroyed. It was held, and released when the hostage-taker died.

Marsolier and colleagues, 2015, showed the parasite secretes TaPIN1, a working prolyl isomerase — an enzyme that bends a protein at one amino acid — and a homolog of human PIN1. In the host cell it degrades FBW7, a ubiquitin ligase that tags proteins for destruction, so the growth-driving transcription factor c-JUN is stabilized. Buparvaquone inhibits TaPIN1 directly. Earlier, Heussler and colleagues, 2002, showed survival depends strictly on constitutive NF-kappaB activity, with the kinase complex that switches it on assembled into foci on the parasite's own surface, only at the transforming schizont stage.

A metabolic layer belongs to another hour. Medjkane and colleagues, 2014, found a Warburg-like phenotype in these leukocytes — raised glucose uptake and raised lactate, both lost when the parasite was eliminated, with chronically stabilized HIF1-alpha behind them. Whoever writes the fuel part should keep that distinction: the abstract does not say the HIF1-alpha stabilization itself reversed. The Atlas page on the Warburg effect carries it.

Now three corrections, by name. First: it is not only Theileria. Dobbelaere and Heussler's 1999 review states that both species do it — T. parva transforms T and B cells, T. annulata affects B cells and monocytes — and it supplies the phrase people reach for: Theileria-induced transformation is, in their words, entirely reversible. Giacomini and colleagues, 2026, out of the same Paris laboratory, pooled transcriptomic, proteomic and epigenomic data: both species repress innate and adaptive immune genes — Toll-like receptors, inflammasome components, major histocompatibility complex class II — with activating histone marks lost from the promoters of the repressed genes and the silencing mark H3K27me3 enriched on CIITA, the master regulator of those MHC genes. Theilericidal drugs, in their words, could partially rescue immune gene expression. Partially, against 1999's entirely. And nothing here edits a genome: what changes is which host genes get read.

Nor is it confined to the genus. Certad and colleagues, in two 2010 papers, gave Cryptosporidium parvum oocysts — the tough transmissible stage — to dexamethasone-treated SCID mice, immunodeficient animals given a steroid on top. All six of six given ten to the fifth through ten to the seventh oocysts developed high-grade intraepithelial neoplasia, or adenomas with high-grade dysplasia, in the cecum after day 46. And Benamrouz and colleagues, 2014, sequenced those lesions: no mutations in the selected loci of beta-catenin, Apc, E-cadherin, Kras or p53 — though the same paper found Wnt pathway components and p53 abnormally located by staining. No mutation in the genes anyone thought to check, and a pathway visibly deranged without one.

Second: it is not fully malignant. The careful wording is Tretina and colleagues', 2015 — phenotypes shared with some cancers, most notably immortalization, hyperproliferation and dissemination — and that review says the parasite proteins responsible remain unknown.

Third, a citation, because the failure is so ordinary. The memorable phrase — a reversible Warburg effect is induced by Theileria parasites to transform host leukocytes — is a real title. It belongs to a two-page editorial by Medjkane and Weitzman in Cell Cycle, 2013, with no abstract in the record, not to the five-author Oncogene research paper whose author list gets pasted onto it. A real title, the wrong paper underneath. And the sentence that closes route three: the one parasite shown to actively hold p53 off is a cattle parasite with no human disease. Parasites suppress your p53 is not a human claim. A bovine cell-line claim, and only partly reversible.

Which leaves route four, and it fails. The retraction is the story. In July 2004 Nitz and colleagues published in Cell that kinetoplast minicircle DNA from Trypanosoma cruzi — small DNA loops from the single large mitochondrion of the Chagas parasite — had integrated into the genomes of human Chagas patients at five loci, frequently targeting the beta-globin locus and LINE-1 retrotransposons, the most abundant mobile elements in our own DNA. They reported the same in rabbits and chickens, inherited through the germ line as far as the F2 generation, the grandchildren of the infected birds.

Cell's one-page retraction notice appeared on 23 September 2005, fourteen months later, and the authors did not endorse it. Three weeks after it the journal's own editor published a separate editorial on the dispute, indexed under Retraction of Publication as Topic and Editorial Policies (Marcus 2005). The reasoning everybody quotes — that independent experts found the integration-site sequence analyses did not provide strong evidence for the central hypothesis and were open to alternative interpretations — sits in neither indexed record, since neither carries an abstract, so which document actually says it has to be read off the journal before air.

The same Brasilia group kept publishing, and those papers are not retracted. Hecht and colleagues, 2010, reported minicircles integrated mainly into LINE-1 in the founders of five naturally infected families; the chicken model followed in 2011 (Teixeira and colleagues) and 2014 (Guimaro and colleagues). And a 2006 paper tracked a minicircle integrated into a LINE-1 element in clonal human macrophage lines, relocating over three years, with the p15 locus — CDKN2B, a tumor suppressor — altered and its messenger RNA eliminated (Simoes-Barbosa and colleagues 2006): the closest thing in the literature to parasite DNA mutating a human tumor suppressor. No laboratory outside that one has replicated any of it, and all of it rests on one group's targeted-primer TAIL-PCR, precisely the assay class that stitches together DNA never joined inside a cell. A chimeric artifact.

For worms there is a systematic test. Wijayawardena and colleagues, 2015, took thirteen published reports of direct horizontal gene transfer between schistosomes and their hosts — DNA moving sideways between organisms rather than down a family line — and reanalyzed every one. They could not substantiate a single one, concluding the cases may be technical artifacts rather than biological reality. And the burden of proof, they write, is high.

What real transfer looks like runs sideways, or involves a bacterium. Brann and colleagues, 2026, found two transposable elements in the Schistosoma mansoni genome, Perere-3 and Sr3, that appear to have moved horizontally between the fluke and its intermediate snail hosts; the same elements turn up in turtles, fish and mollusks schistosomes cannot infect, and the authors cannot explain it. Wolbachia insertions are confirmed in four insect and four nematode genomes (Dunning Hotopp and colleagues 2007) and make up over two percent of the Drosophila ananassae genome (Klasson and colleagues 2014) — contested as contamination until Tvedte and colleagues, 2022, long-read sequenced a cured fly line and assembled 4.9 megabases of insertions under 8,000 years old. That is how these claims fail, and how one stuck. Gladyshev and colleagues, 2008, found bacterial, fungal and plant genes in bdelloid rotifers, microscopic freshwater animals.

One more, because peer review changed it. Wendt and Collins found a cyclin-dependent kinase inhibitor homolog, cki, throughout parasitic flatworms and absent from their free-living relatives — and the transfer language softened between the 2024 preprint, whose title led with horizontal gene transfer, and the 2025 Science Advances paper, titled unusual inheritance, whose abstract says the gene may have come from horizontal gene transfer. Brindley's 2026 commentary summarizes it; the worm biology belongs to part five.

Then the plain sentence, and the two claims that die with it. No study has reported parasite DNA integrated into a human tumor genome. Not one — and be precise about why: the experiment that could find it, long reads of tumor against matched normal tissue, the design that rescued the Wolbachia case, has not been run. That is an absence of evidence, not evidence of absence. So: Schistosoma haematobium does not incorporate its own DNA fragments into host bladder cells; the documented mechanisms are the inflammation and the nitrosamine chemistry already laid out. And parasites do not generally splice DNA into host genomes: the founding paper is retracted, thirteen worm reports could not be substantiated, and the one well-supported schistosome transfer goes into snails.

There is a route that does reach into human cells, and it carries no DNA. Chaiyadet and colleagues, 2015, found liver-fluke extracellular vesicles — tiny membrane-wrapped packages cells release and other cells swallow — in culture medium and in bile from infected hamsters and humans; taken up by human bile-duct cells they drove proliferation and interleukin-6 secretion, and antibodies against a fluke tetraspanin blocked the uptake and the interleukin-6 response. Pakharukova and colleagues, 2023, showed Opisthorchis felineus vesicles entering human bile-duct cells by clathrin-dependent endocytosis. Zhu and colleagues, 2016, found 403 proteins and small RNAs in Schistosoma japonicum vesicles, with microRNAs transferred into mammalian cells; Marcilla and colleagues, 2012, described the first helminth exosome-like vesicles, in the thirty-to-hundred-nanometer size class. And Lishai and Pakharukova, 2026, predicted 1,299 human targets for the eight most abundant O. felineus transfer-RNA fragments; of 1,484 genes whose expression changed in treated THP-1 cells, a human monocyte line, 159 were predicted targets. Regulation, not inheritance.

Heneberg, 2026, supplies the frame. He rejects the premise that chronic inflammation is intrinsically carcinogenic and proposes a threshold: carcinogenesis emerges only where sustained inflammatory intensity, prolonged exposure, genotoxic stress, tissue-specific vulnerability and permissive cofactors converge. Most helminths stay below it, he argues, because selection favors host survival, they carry no direct genotoxin, and they are effective immunoregulators. What puts this one worm on the Group 1 list is not that it is a worm. It is that it lays eggs in an organ that holds a chemically hostile fluid for hours at a time, for decades, in a host whose own bacteria make nitrosating agents.

So the verdict, in both directions. The link is real, measured in human tissue down to a specific DNA lesion and the mutation it produces. And it is smaller than its folklore: 1.72 against tobacco's 6.6 in the same Egyptian population, sixteen percent of cases against seventy-five, and no pooled estimate anywhere (Bedwani and colleagues 1998, 1997). The cancer it causes changed type when the parasite was controlled — squamous carcinoma from 75.9 percent of bladder cancers down to 28.4, transitional from 16.0 up to 65.8, across 9,843 patients in Cairo (Gouda and colleagues 2007) — which is the strongest causal evidence parasite oncology has produced, and is almost never told that way. And the mechanism stops at the host's chemistry and the host's enzymes, because nothing in the record shows a parasite editing a human genome.

Sources

Every paper referenced on air, in the order it comes up. Links go to the publisher via DOI.