Epistemic status: fairly confident

Most cancer cells, instead of using cellular respiration, get ATP  from glycolysis. This is called the “Warburg Effect.”

Glycolysis is less efficient than cellular respiration, which is why cancer cells have higher glucose requirements and why interventions that improve insulin sensitivity (like metformin) have anti-cancer effects.  The high glucose consumption of cancer cells is so reliable an indicator of cancer that it is the basis of detecting metastases in PET scans.

Tumor progression is associated with the emergence of the Warburg effect.  One hypothesis for why is that cancer cells lose mitochondrial function (and thus the capacity for cellular respiration) as they mutate; another is that the hypoxic environment of a tumor makes cellular respiration impossible.

Because  reliance on glycolysis is such a general feature of cancer, a natural type of drug to try on cancers is a glycolysis inhibitor, which would cut off tumors’ energy supply.  To the extent that the Warburg effect is a result of the inability to engage in cellular respiration, glycolysis inhibitors should not fail as cancers continue to advance, but should, if anything, become more effective; cancer cells which have lost mitochondrial function are unlikely to regain it, while cancer cells which have some distinctive surface marker targeted by a drug are likely to lose it over time.

An early-stage drug called 3-bromopyruvate is a glycolysis inhibitor which has shown promising animal results.

3-BP is a strong alkylating agent that inhibits the glycolysis enzyme GAPDH.

In a rat model of hepatocellular carcinoma [1], rats were treated with 2.0 mM of 3-BP; all treated rats survived more than seven months, while all control rats died within days.  The 3-BP eradicated large bulging tumors in the treated rats. The cellular ATP level of healthy cells stayed constant while the cellular ATP level of cancer cells dropped to 10% of its former level.

In a study of rabbits with liver tumors [2], mean survival in the control group was 18 days while mean survival in the treated group (with intraarterial 3-BP) was 55 days.  The tumors in treated animals had not grown and were mostly necrotic; the livers in control animals was almost completely replaced by tumor, with extension into the diaphragm and lungs.  Death in the treated animals was believed to be due to extrahepatic disease (by the time of treatment, the tumors were already metastatic, both in control and treated animals.)  Three animals were treated “earlier” (one week after implantation): one survived, the other two survived 80 days and died of lung metastases.  Intravenous administration, by contrast, was not effective.

Another rabbit study [3] found that intravenous 3-BP doesn’t kill liver tumors but does eliminate their lung metastases.

3-BP  also enhances survival of nude mice injected with human mesothelioma.[4]  There was a control group, a 3-BP treated group, a cisplatin-treated group, and a combined 3-BP and cisplatin group.   In the control group, 3-BP only group, and cisplatin only group, all mice died before day 45. In the combined group, survival was better than control (p = 0.0021), 3-BP alone (p = 0.0024), and cisplatin alone (p = 0.0161).  

In a study of 3-BP on mice with lymphoma[5], a significant reduction in tumor activity was found among those given repeated treatments (p = 0.0043 at day 7) but in those given only a single treatment, the effect was only significant at the second day (p = 0.0152 at day 2) and afterwards the tumors returned to normal.  The only manifestation of toxicity was lower body weight in the 3-BP-treated groups.

Microencapsulated 3-BP, delivered systemically, into a mouse model of pancreatic ductal carcinoma[6], causes minimal to no tumor progression, compared to animals treated with gemcitabine, which had a 60-fold increase in BLI, a measure of tumor activity.

Glioma cells are refractory to most treatment. D-amino acid oxidase combined with 3-bromopyruvate [7] decreased proliferation and viability in rats.

Aerosolized 3-BP also prevents the development of lung cancer in carcinogen-treated mice, without causing liver toxicity.[11]

There has been one human case study of 3-bromopyruvate, [8] on a young man with terminal fibrolamellar carcinoma. At the time of treatment, he was on a feeding tube, he had ascites, and his spleen was swollen due to complete blockage of the renal artery. He had already been treated with chemotherapy.  The University of Frankfurt’s ethics committee permitted him to be treated with a 3-BP, through the TACE delivery method (transcatheter arterial chemoembolism). He began having symptoms of tumor lysis syndrome, a dangerous condition associated with tumor necrosis.  CT scans showed that the tumors were necrotic after administering 3-BP, and the tumor areas were encapsulated and showed fibrosis.   His mobility was limited due to pre-existing ascites and edema, but he began to go out in his wheelchair. He died of liver failure 2 years after his initial diagnosis.   A fluid sample from his ascites found some mesothelial cells, but detected no tumor cells.

There are also a few negative results about 3-BP, but these are explainable by differences in protocols from the successful studies. One study of rabbits with liver cancer found 3-BP didn’t significantly increase tumor necrosis relative to controls, but it only used a single administration rather than repeated administration.[9]  One study of rabbits with liver cancer found that 3-BP killed them, but that was with a dose of 12x the dose in the successful studies.[10]  

This is fairly strong animal evidence, across a variety of cancers, of the most confident kind (tumor eradication and prolonged survival, not just inhibition of growth).  The single human study is not strongly conclusive but leans towards a positive effect.

One major issue is that intra-arterial administration for liver cancer seems to work while systemic administration doesn’t, though other types of cancer do seem to respond to systemic, microencapsulated, or aerosolized delivery. Clearly, attention needs to be paid to optimizing dosage, administration schedule, and delivery method.

As well as being a glycolysis inhibitor, 3-BP is also an alkylating agent, like most cytotoxic chemotherapies; this suggests it may work along multiple pathways. It also suggests we might see similar side effects to cytotoxic chemotherapy in humans.

3-BP is not the only glycolysis-inhibiting drug that has been tried.

Imatinib, which is by far the most successful targeted cancer therapy, is, among other things, a glycolysis inhibitor.

Dichloroacetic acid had some promising animal studies and stabilized five patients with recurrent glioblastoma on palliative care[12] for fifteen months (note that median survival time after the first round of radiation + chemo is only 15 months itself). But DCA has carcinogenic effects in animals, so research was halted. It also lost credibility due to alt-med practitioners selling DCA without FDA approval.

Lonidamine, another glycolysis inhibitor, is currently in clinical trials for brain cancer, but its past clinical results have been unspectacular.  I would have to look in more detail to understand whether there’s any mechanistic reason why 3-BP could be expected to do better than lonidamine in human trials; but for now, the failure of an earlier glycolysis inhibitor is admittedly an argument against 3-BP.

The case for 3-BP may be somewhat weakened but not wholly dismantled by alternate hypotheses about what causes the increased rate of glycolysis in cancer.

The “Reverse Warburg Effect” is a theory that argues that in some epithelial cancers (such as types of breast cancer), it is not the cancer cells but a type of neighboring healthy cells called fibroblasts that have elevated levels of glycolysis.  Glycolysis inhibitors still work on such cancers, however; DCA blocks cancer growth in vitro in fibroblast-induced breast cancer tumor growth.[13]

The Crabtree effect[14] is the phenomenon that some tumor cells have a reversible, short-term shift to glycolysis in the presence of glucose; in such cells, glycolysis inhibitors might not work because the cells could switch back to cellular respiration.  The overall effectiveness of glycolysis inhibitors on cancer will depend on whether the shift to glycolysis is mostly a facultative or an obligate phenomenon.  We can expect glycolysis inhibition to slow growth even if it doesn’t kill tumors, because glycolysis allows for faster cell growth than respiration; but if tumor cells can switch back and forth from glycolysis to respiration, then glycolysis inhibitors wouldn’t be likely to eradicate late-stage tumors, but only to stall early ones.

Overall, 3-BP looks like an unusually strong early-stage cancer drug. It appears likely to be broadly effective among many types of cancer. It is simple, has dramatic results (in animals), and is upstream.

Mechanistically, glycolysis inhibitors seem to be a strategy that is robust to different hypotheses about the cause of the high rate of glycolysis in cancer, and glycolysis inhibitors that are also generally cytotoxic would be even more robust to mechanistic uncertainty.

While there are many preclinical drugs that never succeed in humans, 3-BP is fairly novel in mechanism and seems likelier to pan out than the average new targeted therapy, precisely because it relies on a strategy of attacking cancer that is quite simple, should be broadly applicable across cancers and shouldn’t change over time as tumors evolve.  In order for a tumor to “work around” 3-BP, it would have to create an alternate enzyme to catalyze the 6th step of glycolysis, which seems unlikely; while in order for a tumor to work around an angiogenesis inhibitor, for instance, it would only have to find a different way to stimulate blood vessel growth, which, given the wide variety of growth factors in the human body, is not too unlikely.


[1]Ko, Young H., et al. “Advanced cancers: eradication in all cases using 3-bromopyruvate therapy to deplete ATP.” Biochemical and biophysical research communications 324.1 (2004): 269-275.

[2]Vali, Mustafa, et al. “Targeting of VX2 rabbit liver tumor by selective delivery of 3-bromopyruvate: a biodistribution and survival study.” Journal of Pharmacology and Experimental Therapeutics 327.1 (2008): 32-37.

[3]Geschwind, Jean-Francois H., et al. “Novel Therapy for Liver Cancer Direct Intraarterial Injection of a Potent Inhibitor of ATP Production.” Cancer research62.14 (2002): 3909-3913.

[4]Zhang, Xiaodong, et al. “Novel therapy for malignant pleural mesothelioma based on anti-energetic effect: an experimental study using 3-Bromopyruvate on nude mice.” Anticancer research 29.4 (2009): 1443-1448.

[5]Schaefer, Niklaus G., et al. “Systemic administration of 3-bromopyruvate in treating disseminated aggressive lymphoma.” Translational Research 159.1 (2012): 51-57.

[6]Chapiro, Julius, et al. “Systemic Delivery of Microencapsulated 3-Bromopyruvate for the Therapy of Pancreatic Cancer.” Clinical Cancer Research (2014): clincanres-1271.

[7]El Sayed, S. M., et al. “D-amino acid oxidase gene therapy sensitizes glioma cells to the antiglycolytic effect of 3-bromopyruvate.” Cancer gene therapy 19.1 (2011): 1-18.

[8]Ko, Y. H., et al. “A translational study “case report” on the small molecule “energy blocker” 3-bromopyruvate (3BP) as a potent anticancer agent: from bench side to bedside.” Journal of bioenergetics and biomembranes 44.1 (2012): 163-170.

[9]Shin, S. W., et al. “Hepatic intra-arterial injection of 3-bromopyruvate in rabbit VX2 tumor.” Acta Radiologica 47.10 (2006): 1036-1041.

[10]Chang, Jung Min, et al. “Local toxicity of hepatic arterial infusion of hexokinase II inhibitor, 3-bromopyruvate: in vivo investigation in normal rabbit model.”Academic radiology 14.1 (2007): 85-92.

[11]Zhang, Qi, et al. “Aerosolized 3-bromopyruvate inhibits lung tumorigenesis without causing liver toxicity.” Cancer Prevention Research 5.5 (2012): 717-725.

[12]Michelakis, E. D., et al. “Metabolic modulation of glioblastoma with dichloroacetate.” Science translational medicine 2.31 (2010): 31ra34-31ra34.

[13]Bonuccelli, Gloria, et al. “The reverse Warburg effect: glycolysis inhibitors prevent the tumor promoting effects of caveolin-1 deficient cancer associated fibroblasts.” Cell cycle 9.10 (2010): 1960-1971.

[14]Diaz-Ruiz, Rodrigo, Michel Rigoulet, and Anne Devin. “The Warburg and Crabtree effects: On the origin of cancer cell energy metabolism and of yeast glucose repression.” Biochimica et Biophysica Acta (BBA)-Bioenergetics 1807.6 (2011): 568-576.

Metformin, Fasting, and Exercise

Epistemic status: fairly confident.

TW: discussion of diet and exercise

In the post on anti-IgG, I talked about the strategy of routing around the complexity of cancer by finding a way to detect and kill a wide range of cancer cells at once. Another set of strategies for routing around complexity focuses on starving a wide range of cancer cells of the resources they need to grow.

Some strategies of this kind, like VEGF inhibitors to prevent angiogenesis, don’t work as well as one might hope: bevacizumab does not prolong survival more than a few months in any cancer.

However, one might hope that it’s possible to do better by looking at a very fundamental aspect of cancer cells: their extremely high energy requirements.  Obviously cells growing and proliferating rapidly need more energy than typical cells; this is especially true given the Warburg effect in which most cancer cells use glycolysis, which is less efficient than cellular respiration and requires more glucose.  Insulin insensitivity, in which tissues are slow to absorb glucose from the blood, leaves more resources available for cancer cells.  (This is a vastly oversimplified model and I do not fully understand how insulin relates to cancer growth.  There are multiple possible mechanisms whereby high insulin levels and insulin resistance contribute to cancer.)

Some simple methods that improve insulin sensitivity — exercise, fasting, and the type 2 diabetes drug metformin — seem to have anti-cancer effects.


The effect of exercise on cancer is extremely confounded by other factors, so I won’t discuss it in much detail, except to say that observational studies have found that regular exercise has a preventative effect on some types of cancer.

At least 170 observational studies have been conducted on the association between exercise and cancer risk. The evidence is strongest for colon, breast, and prostate cancer.  Colon cancer had 43 of 51 studies demonstrate a reduced risk of cancer with physical activity, with a 40-50% reduction in risk; breast cancer had 32 of 44 studies demonstrate a reduced risk of cancer with physical activity, with a 30-40% reduction in risk; prostate cancer had 17 of 30 studies demonstrate a reduced risk of cancer with physical activity, with a 10-30% reduction in risk.[0]

In colon cancer, the effect of exercise was found in both recreational and occupational activity, and was observed even after controlling for BMI and dietary intake.

One possible mechanism for exercise’s preventative effects is that it reduces the level of sex hormones, which play a role in promoting the growth of breast and prostate cancer.  Exercise also improves insulin sensitivity and decreases the circulating levels of insulin-like growth factor (IGF); IGF and circulating insulin play a role in cancer growth.


Metformin is a drug for Type 2 diabetes. It targets the enzyme AMPK, which induces muscles to take up glucose from the blood. This inhibits the production of glucose by the liver, which is why it reduces hyperglycemia in diabetics. It also increases insulin sensitivity and decreases insulin-induced suppression of fatty acid oxidation.

Insulin promotes cancer growth; insulin affects tumor cells either directly or indirectly through sex hormones, insulin-like growth factors, or adipokines.  Cancer cells have high energy requirements because of their rapid growth rate and their dependence on glycolysis.

Epidemiologic Evidence

A retrospective study of 11,876 diabetes patients in a Scottish hospital found that taking metformin slightly reduced the risk of cancer. The odds ratio was 0.86.[1]

A subsequent cohort study at the same hospital, of 4085 metformin users vs. 4085 matched diabetics who didn’t take metformin, found 7.3% of the metformin users vs. 11.6% of the comparators got cancer within 10 years, an odds ratio of 0.63 after adjusting for sex, age, BMI, SES, A1C, smoking, and drug use.  Also, 3.0% of metformin users died of cancer, compared to 6.1% of comparators.[2]

Complete response rates in breast cancer (defined as no sign of invasive carcinoma at the time of surgery, after a course of chemotherapy) were 24% for the metformin group, 8% for the non-metformin diabetic group, and 16% for the nondiabetic group.  This is statistically significant for metformin vs. non-metformin (p = 0.007) but not for metformin vs. nondiabetic.

A study of 1353 patients with diabetes found a 0.43 hazard ratio for cancer mortality among those taking metformin.[8]

A meta-study of 11 epidemiologic studies of metformin and cancer found a pooled relative risk for cancer incidence of 0.55.  The relative risk varied by year of use: 0.77 for 1 year, 0.6 for 2 years, and 0.28 for 5 years.  These studies compared diabetics using metformin to diabetics using other treatments, and included all types of cancers.[6]

A meta-study of 4 cohort studies and 2 RCTs found a pooled risk ratio of 0.66 for all-cancer mortality, 0.67 for all-cancer incidence.[9]

In a cohort of 480,984 Taiwanese participants, cancer incidence was twice as high for diabetics not on metformin as for nondiabetics; diabetics on metformin had similar cancer risk to nondiabetics.  Metformin users vs. diabetic metformin non-users had a hazard ratio of 0.47.[7]

Experimental Human Evidence

In 55 breast cancer patients randomized to metformin or no drug before surgery,  Ki67 levels (a measure of cellular proliferation) in the tumors dropped in all but 2 metformin patients but remained stable in control patients.[12]  However, a different randomized trial found no effect of preoperative metformin on Ki67 levels.[13]

A meta-analysis of randomized controlled studies where diabetics were given either metformin or a comparator found that cancer incidence was no lower in patients given metformin.[14]  This is a nontrivial concern, given that non-metformin drugs tend to be given to patients with more severe diabetes, meaning that observational studies comparing metformin-treated patients to other diabetics may be biased.

In-Vitro Evidence

Metformin inhibits growth of breast cancer cells and upregulates AMP-kinase activity in those sells; siRNA specific to AMP-kinase (blocking its expression) makes the anti-cancer effect of metformin stop.[3]

Metformin preferentially kills breast cancer stem cells, and prevents their transformation into tumor cells.[4]

Metformin inhibits p53-/-  colon cancer cell lines.[5]  These are usually the hardest type of cancer to treat; cancers with the p53 gene mutated are usually advanced and not responsive to most chemotherapies.

Animal Evidence

Metformin + doxorubicin kills all tumors in mice injected with breast cancer stem cells, while doxorubicin alone cause only a 2-fold decrease in tumor volume and metformin alone has little effect.  Mice remain in remission for 60 days after doxorubicin + metformin, vs. 20 days for doxorubicin alone.[4]

Hamsters fed a high fat diet and a carcinogen got pancreatic cancer 50% of the time; hamsters fed the high fat diet, metformin, and the carcinogen didn’t get cancer.[10]

Mice given ovarian cancer xenographs had about half the total tumor mass when treated with metformin; it also inhibits proliferation, metastasis, and angiogenesis. Metformin + cisplatin  resulted in significantly less proliferation, tumor area, mitotic counts, and vasculature than cisplatin or metformin alone.[11]

Metformin Conclusions

Metformin is definitely simple in its mechanism, and quite cheap. It is also probably upstream, though I don’t know yet how early in the process of cancer development its unusually high energy requirements arise.  It’s not especially decisive, given that it only has a moderate preventative effect and only inhibits growth, rather than killing cancer cells.

We don’t yet have direct evidence of how it works in humans as an adjuvant to chemotherapy or what its long-term effects on non-diabetics are; these seem like obvious experiments to run. Moreover, given its good side-effect profile, it’s plausible that healthy people could take metformin as a cancer preventative.


Short-term fasts (<2 days) while taking chemotherapy may make the side effects milder and the effects stronger.  There is very little clinical evidence about this because doctors are understandably concerned about causing excessive weight loss in cancer patients; however, from the information I have available, it appears that short-term fasts followed by eating freely are safe.

In a case study[16] of 10 humans who had fasted voluntarily before and after chemotherapy, patients reported fewer side effects from courses of chemotherapy during which they fasted compared to courses of chemotherapy during which they ate.  They had faster recovery of blood cell and platelet counts during the chemotherapy regimens when they fasted. 

Cancer cells are quicker to die during a period of fasting than healthy cells are; this phenomenon is known as differential stress resistance and has been observed repeatedly in animal and in-vitro studies.

An in-vitro study[15] of tumor cells and RAS-mutated yeast cells found that 48-hour fasts sensitized the cells to chemotherapy; in mice, tumors were less than half the size in fasted mice than fed mice after 34 days and 5 fasting cycles, and mice were able to regain normal weight.

Mice injected with glioma were more responsive to chemotherapy and radiotherapy when they were subjected to 48-hour fasts; by day 28, 85% of the fasting + chemo mice were alive, compared to 40% of the fasting alone and chemo alone mice.  89% of the fasting + radiotherapy mice were alive by day 32, compared to 40% each of the fasting alone and radiotherapy alone mice.[17]

The differential stress resistance response appears to be associated with the effect of fasting on reducing IGF-1, a growth factor which promotes cancer.  Mice injected with metastatic melanoma and treated with doxyrubicin died 40% of the time from the doxorubicin and by 90 days all were dead from either the chemotherapy or metastases; the mice that lacked the ability to produce IGF-1 (mimicking the effect of fasting) survived 60% of the time.[18]

A cohort of Ecuadorian people with IGF deficiency (who are of unusually short stature, due to lack of growth hormone sensitivity) had unusually low rates of cancer; 20% of their unaffected relatives died of cancer but none of the IGF-deficient subjects did, a statistically significant difference.[19]


There’s some kind of rough emerging picture around insulin, glucose levels, and the metabolic syndrome, whereby insulin and growth factors and high blood glucose are associated with cancer growth, and insulin-sensitivity-promoting things like metformin, exercise, and short-term fasts have anti-cancer effects. There are multiple possible causal pathways that seem to point in the same direction. There’s a sort of “antifragile” heuristic here — cancer cells are especially vulnerable in multiple ways to metabolic stress, and messing with their energy supply might be a robust way to attack cancer even though there are surely many undiscovered biochemical pathways. These strategies seem to be more targeted at cancer prevention and growth inhibition than cancer eradication, except for metformin’s surprising effects on advanced cancer cell lines.  As usual, I think the conclusions to draw are “more research here”, not “cancer is cured”, though in this case there are obvious lifestyle choices that people can experiment with. And it seems clear that clinical trials of metformin and fasting during chemotherapy would be useful.


[0]Friedenreich, Christine M., and Marla R. Orenstein. “Physical activity and cancer prevention: etiologic evidence and biological mechanisms.” The Journal of nutrition 132.11 (2002): 3456S-3464S.

[1]Evans, Josie MM, et al. “Metformin and reduced risk of cancer in diabetic patients.” Bmj 330.7503 (2005): 1304-1305.

[2]Libby, Gillian, et al. “New Users of Metformin Are at Low Risk of Incident Cancer A cohort study among people with type 2 diabetes.” Diabetes care 32.9 (2009): 1620-1625.

[3]Zakikhani, Mahvash, et al. “Metformin is an AMP kinase–dependent growth inhibitor for breast cancer cells.” Cancer research 66.21 (2006): 10269-10273.

[4]Hirsch, Heather A., et al. “Metformin selectively targets cancer stem cells, and acts together with chemotherapy to block tumor growth and prolong remission.”Cancer research 69.19 (2009): 7507-7511.

[5]Buzzai, Monica, et al. “Systemic treatment with the antidiabetic drug metformin selectively impairs p53-deficient tumor cell growth.” Cancer research 67.14 (2007): 6745-6752.

[6]DeCensi, Andrea, et al. “Metformin and cancer risk in diabetic patients: a systematic review and meta-analysis.” Cancer prevention research 3.11 (2010): 1451-1461.

[7]Lee, Meei-Shyuan, et al. “Type 2 diabetes increases and metformin reduces total, colorectal, liver and pancreatic cancer incidences in Taiwanese: a representative population prospective cohort study of 800,000 individuals.”BMC cancer 11.1 (2011): 20.

[8]Landman, Gijs WD, et al. “Metformin associated with lower cancer mortality in type 2 diabetes ZODIAC-16.” Diabetes care 33.2 (2010): 322-326.

[9]Noto, Hiroshi, et al. “Cancer risk in diabetic patients treated with metformin: a systematic review and meta-analysis.” PloS one 7.3 (2012): e33411.

[10]Schneider, Matthias B., et al. “Prevention of pancreatic cancer induction in hamsters by metformin.” Gastroenterology 120.5 (2001): 1263-1270.

[11]Rattan, Ramandeep, et al. “Metformin suppresses ovarian cancer growth and metastasis with enhancement of cisplatin cytotoxicity in vivo.” Neoplasia 13.5 (2011): 483-IN28.

[12]Hadad, Sirwan, et al. “Evidence for biological effects of metformin in operable breast cancer: a pre-operative, window-of-opportunity, randomized trial.” Breast cancer research and treatment 128.3 (2011): 783-794.

[13]Bonanni, Bernardo, et al. “Dual effect of metformin on breast cancer proliferation in a randomized presurgical trial.” Journal of Clinical Oncology30.21 (2012): 2593-2600.

[14]Stevens, R. J., et al. “Cancer outcomes and all-cause mortality in adults allocated to metformin: systematic review and collaborative meta-analysis of randomised clinical trials.” Diabetologia 55.10 (2012): 2593-2603.

[15] Lee, Changhan, et al. “Fasting cycles retard growth of tumors and sensitize a range of cancer cell types to chemotherapy.” Science translational medicine4.124 (2012): 124ra27-124ra27.

[16]Safdie, Fernando M., et al. “Fasting and cancer treatment in humans: A case series report.” Aging (Albany NY) 1.12 (2009): 988.

[17]Safdie, Fernando, et al. “Fasting enhances the response of glioma to chemo-and radiotherapy.” (2012): e44603.

[18]Lee, Changhan, et al. “Reduced levels of IGF-I mediate differential protection of normal and cancer cells in response to fasting and improve chemotherapeutic index.” Cancer research 70.4 (2010): 1564-1572.

[19]Guevara-Aguirre, Jaime, et al. “Growth hormone receptor deficiency is associated with a major reduction in pro-aging signaling, cancer, and diabetes in humans.” Science translational medicine 3.70 (2011): 70ra13-70ra13.

CAR-T Immunotherapy

Epistemic status: moderate confidence.

CAR-T immunotherapy has a reputation for being the new exciting thing in cancer, and as far as I can tell, this reputation is mostly deserved.

It’s not a single therapy, but a general technique. You graft the monoclonal antibody of your choice onto a T-cell with retroviral vectors; then it can selectively attack any type of cell that has a distinctive antigen.

An advantage of CAR-T cells is that the chimeric antigen receptor need not depend on HLA, so it can overcome the tumor’s ability to downregulate HLA molecules.  CAR-T cells bind directly to the antigen, usually on the surface of the cell.

Early clinical trials have shown really strong responses to CAR-T therapy in leukemia and lymphoma.

CAR-T caused complete remission in two children with refractory acute lymphoblastic leukemia; one relapsed after two months and one remission was ongoing.  (ALL is curable in 80% of children, but prognosis is poor for those with chemotherapy-resistant disease.) The cells were modified with specificity for CD19, a hallmark of B-cells.

LeY-specific CAR-T cells in relapsed acute myeloid leukemia (AML) caused a (temporary) remission in 1 of 4 patients.

CD-19 specific CAR-T cells produced an 88% complete response rate out of 16 patients with relapsed acute lymphoblastic leukemia; there were 10 complete remissions.

5 out of 5 adult patients with relapsed ALL treated with CD19-specific CAR-T cells had complete remissions.

3 out of 11 patients with high-risk neuroblastoma had complete remissions upon treatment with EBV-specific and GD2-specific CAR-T cells.

Of 3 patients with relapsed indolent B-cell lymphoma treated with CD20-specific CAR-T cells, 2 survived progression-free for 12 and 24 months, and 1 had a partial remission and relapsed after 12 months.

Leukemia and lymphoma are “easy mode” for a few reasons. First of all, several types of such cancers have very specific biomarkers. The Philadelphia chromosome  is pretty much unique to the leukemias on which it exists. B-cell lymphoma is restricted to B-cells, and CD19 and CD20 are markers that occur on almost all B-cells and no other cells.  The common solid-tumor cancers (lung, breast, colon, prostate, etc) have no equivalently specific markers. Second of all, blood cancers are easier for drugs to physically access than solid tumors, which produce a hostile “microenvironment” that protects the tumor against attack, especially immune attack.

Challenges facing CAR-T therapy include the cost of engineering T-cells, and poor in-vivo persistence of transferred T-cells.  Second- and third-generation CAR-T therapies involve editing multiple stimulatory domains, since using only one often fails.  Exogenous cytokine administration, especially IL-2, is used to enhance persistence of T-cells.  A challenge especially for solid tumors is transferring T-cells to the site of disease, and resisting the immune-suppressive effects of the tumor microenvironment.  Fever enhances the adherence of T-cells to the tumor microvasculature, so heat therapy may be helpful here.  (Note that fever is also a factor in spontaneous remissions and bacterial cancer therapies.)

Attempts so far in solid tumors include CAR-T cells specific for HER2, VEGF, EGFR, and GD2.  So far, it doesn’t work most of the time, though the GD2 studies in neuroblastoma got some complete remissions. The HER2-sensitive trial caused one death due to the T-cells attacking healthy tissue and causing respiratory failure. Lack of specificity and lack of persistence have been problems, but as of 2015 we’re still waiting for the results of many solid-tumor clinical studies.

CAR-T is “decisive” in some types of leukemia and lymphoma; complete remissions in a majority of patients is a nearly-unheard-of result in the world of cancer. It’s not particularly “simple”, because it can be customized arbitrarily with different antigens, different activating domains, etc.  It’s also only somewhat “upstream” — T-cells appear fairly late in the process of immune response to cancer.

Ultimately, though, CAR-T is only as good as the specificity of the cell surface antigens it uses.  It’s subject to the universal challenge of targeted cancer drugs: different cancers have different biomarkers, no one biomarker appears in a majority of cancers, and cancers change their biomarkers as they advance.

If anti-IgG pans out as a much broader cancer-detection mechanism, it seems that it could in principle be applicable to CAR-T.  Also, combination CAR-T with multiple antigens pattern-matches to something that could improve effectiveness.

CAR-T is not an underappreciated area of research — it’s very actively funded and a vigorous research field.  To the extent that it pans out, which will become clearer when the current round of clinical trials publish results, it’s evidence for the “cancer establishment” doing something right, very much as imatinib was a triumph of “big science” and targeted chemotherapy.

The cynical viewpoint is that, like imatinib and targeted chemotherapies, a single clear-cut triumph can lend credibility to a wide class of drugs that mostly don’t work on more difficult cancers, and that since CAR-T is so mutable, it is especially vulnerable to such “exploitation.”  One way or another, we’ll find out when the solid tumor results finish coming in.

Routing Around the Cancer Biomarker Problem with Antibodies

Epistemic status: fairly low confidence

One of the heuristics I’ve been using in this series is that drugs that rely on tumor-specific biomarkers are not very effective. We’ve seen that most targeted chemotherapies have little or no effect on survival; we’ve seen that immunotherapies that depend on tumor-specific antigens are limited by the tendency of advanced cancers to “escape” immunosurveillance by shedding their antigens.

The “cancer is hard and complex” model accurately predicts this. Cancer is diverse, both between tumor types and across the course of the disease.  Most tumor-specific biomarkers also occur on some healthy tissues, and don’t occur on all tumors.  Most targeted drugs have problems with both sensitivity and specificity — they miss too many cancer cells, and attack too many healthy cells.

But cancer might not be as hard if you can route around the complexity by focusing on vulnerabilities that all or most cancers have in common.

This paper from a team of researchers led by Sanford Simon provides one possible strategy: detecting endogenous antibodies.

Specific autoantibodies to particular tumor-specific antigens have fairly low sensitivity for cancer. For example, “In patients with hepatocellular carcinoma (HCC), probing for a single autoantibody in the serum gives a positive result in 10–20% of patients; the detection increases to 66% with a panel of ten autoantibodies20. While the sensitivity of tumor detection can be increased by using a panel of antibodies over a single antibody21, the results are still insufficient for diagnosis in many tumor types.”

Antibodies are part of the adaptive immune response. Also known as immunoglobulins, they are Y-shaped molecules produced by B cells that tag microbes, infected cells, or tumor cells for further attack by other parts of the immune system.

What if, instead of trying to detect a tumor-specific antigen with a targeted drug, you tried to detect all the tumor-specific antigens at once by detecting all the antibodies that the body is already using to identify tumor cells?

That’s what Simon and his team did. They attached a fluorescent chemical to an anti-IgG antibody, which binds to all IgG antibodies (regardless of which antibody they’re selecting for.) The fluorescence was 64x greater in mouse tumor tissue than in normal tissue.

The effect worked across a wide variety of mouse  cancer types: breast cancer, prostate cancer, liver cancer, leukemia, etc.

IgG autoantibodies are abundant in all human, rat, and swine sera, increase with age, and are statistically significantly lower in cancer, Alzheimer’s, and Parkinson’s patients compared to age- and gender-matched controls; they are hypothesized to play a role in clearing the body of debris and damaged or mutated cells.

Gold nanoparticles  mixed with blood serum pick up a “corona” of protein which includes IgG molecules of all types, and the total amount of IgG is higher in cancer than non-cancer patients.  “The test has a 90–95% specificity and 50% sensitivity in detecting early stage prostate cancer, representing a significant improvement over the current PSA test.”

I’ve spoken to Dr. Simon and seen some of his unpublished results, which involved distinguishing tumor from healthy tissue down to the cellular level with fluorescent anti-IgG, in both mice and human tumors of many types.

In a sense, this is an “obvious” idea. If the body produces many kinds of tumor-specific antibodies, each of which selects for some cancersthen measuring the overall level of all antibodies would be a much more robust test that selects more specifically for a wider range of cancers.  It’s essentially just bagging applied to cancer detection.

This is usually presented as a method for detecting cancer, but it could also be a method for treating cancer.  Attaching cytotoxic chemotherapy to an anti-IgG antibody (in place of the fluorescent protein) would concentrate the chemo in the tumor as opposed to the rest of the body, allowing higher doses to be administered safely.

It could also be a guide for better cancer surgery; washing the body cavity with IgG-binding fluorescent protein would make the tumor area light up, helping the surgeon cut precisely where needed and making sure to get adequate margins.

Obviously much more development is needed to identify exactly how precise anti-IgG is on humans.  Since cancers eventually lose antigens as they advance, this can be expected to be less effective on advanced cancers.  However, since cancers lose antigens one at a time as they mutate, an anti-IgG-based therapy that was sensitive to all antigens at once would in principle lose effectiveness later than an immunotherapy based only on a single antigen.

As far as cancer heuristics go, IgG detection is definitely “simple”, it is fairly “upstream”, but it is only moderately “decisive” (we don’t yet have extended survival times or extremely high published precision numbers).  I think it’s potentially important and under-appreciated, however, especially given the very high level of current investment in antibody-based immunotherapies.

Cancer Immunosurveillance Interlude

Epistemic status: summary of standard research + rough thoughts of my own

How exactly does the immune system fight cancer?  And what does that imply for immunotherapy opportunities?

In 1909, Paul Ehrlich proposed the hypothesis that the body protects itself from cancer via immune mechanisms.  Fifty years later, Lewis Thomas and F. Macfarlane Burnet revisited the hypothesis.  Thomas’ argument was evolutionary — he said that complex organisms must develop mechanisms to protect against neoplastic disease as they develop mechanisms mediating homograft rejection.  By contrast, Burnet’s argument was functional; he believed tumor-specific antigens could provoke an immune response. In 1964 the first tumor-specific antigens were observed in mice.  

However, it wasn’t until the 1990s that the immune surveillance hypothesis (that the immune system fights cancer) was widely accepted.  This is because nude mice (mice without thymuses, which have greatly reduced numbers of T cells) don’t have higher rates of cancer than normal mice, either spontaneous or induced by carcinogens. If immune-deficient mice aren’t getting more cancer, went the theory, then the immune system can’t be targeting tumors.  What wasn’t known at the time was that nude mice do have some T-cells, a working innate immune system, and natural killer (NK) cells, which weren’t discovered until 1975.  What really confirmed the immune surveillance hypothesis were other kinds of immunodeficient mice which do have elevated rates of cancer.

The cytokine IFN-gamma has been shown to protect the host against spontaneous, chemically-induced, and transplanted tumors.  Mice lacking the IFN-gamma receptor were 10-20x as sensitive to tumor induction. (STAT1, the famous oncogene, is the transcription factor that mediates most of IFN-gamma’s effects on cells.  Mice with mutations in STAT1 are unable to resist injected tumors.)  IFN-gamma is a signaling molecule involved in both adaptive and innate immunity.  It is the key cytokine that prompts T cells (CD4+) to develop into Th1 cells.  It also promotes NK cell activity, NO synthesis, and much more.

NK cells were discovered in 1975 as a distinct cell type that could kill tumor cells w/o prior sensitization.  NK cells are cytotoxic lymphocytes. They exocytose granules containing perforin, which kill cells by inducing apoptosis.  Perforin, true to its name, punches holes in cell membranes; perforin-deficient mice are 2-3x as likely to produce tumors in response to carcinogens.  Also, NK cell ligands binding to TNF receptor superfamily members on cancer cells induce cytotoxicity. 

NK-deficient mice get more tumors.  Infiltration of tumors with NK cells in humans is a positive prognostic indicator, and lower NK-like cytotoxicity of peripheral blood lymphocytes is predictive of cancer risk.

NKG2DL is an activating receptor on NK cells and some other cells. Its ligands do not occur on healthy tissue. NKG2DL expression is activated by stresses such as heat shock, viral infection, DNA damage, or UV radiation.  NKG2DL ligands have been observed on lots of tumor types.  Cancer cells “shed” these ligands a lot, though, which allows them to evade NK attacks.

MICA and MICB are examples of cellular-stress-related NKG2D ligands.  Tumor cells expressing MICA/B are more vulnerable to NK cells.   MICA/B are found in a high percentage of carcinomas; the only healthy tissue that contains these ligands is gastrointestinal epithelium.

In short, we are now aware of various lines of evidence that the immune system does indeed attack cancers, and that certain kinds of immune deficiency make individuals more susceptible to cancer. (The increased rates of cancers in immunosuppressed individuals such as AIDS patients and transplant recipients are further evidence for the immunosurveillance hypothesis.)

The current theory of the immune response to cancer goes as follows:

  1. As the tumor physically expands (causing pressure on surrounding tissues and growing new blood vessels) the innate immune system starts producing inflammation.
  2. NK cells and T cells are recruited to the tumor site.
  3. T cells and NK cells produce tons of IFN-gamma!  This amplifies the immune response!
  4. IFN-gamma activates macrophages
  5. macrophages produce IL-2
  6. IL-2 stimulates NK cells to produce more IFN-gamma! GOTO 3 and repeat, in a positive feedback loop!
  7. IFN-gamma stimulates macrophages and NK cells to do more tumor killing
  8. IFN-gamma causes apoptosis and anti-growth processes in the tumor itself
  9. Now the adaptive immune response starts; dendritic cells are recruited to the tumor site
  10. dendritic cells show tumor antigens to T cells, which home to the tumor
  11. tumor-specific T cells kill tumor cells
  12. T cells produce even more IFN-gamma! GOTO 3, more positive feedback loop!

Clearly, IFN-gamma is a very big deal here. IFN-gamma is involved in the Th1 (or cellular) immune response, which trades off against the Th2 (humoral) immune response.  There’s a simplistic but suggestive theory that a lot of the diseases of modern aging (heart disease, cancer, diabetes, Alzheimer’s) are associated with a shift in the balance of the immune system towards Th2 and away from Th1 responses; this may be relevant here.

Toll-like receptors, which are expressed on tumor cells and release IL-6, a Th2-promoting cytokine, tend to suppress the immune response to tumors, helping the tumor evade immune surveillance.  Blocking the toll-like receptor slows tumor growth and prolongs survival in mice. This is also consistent with the Th1/Th2 balance hypothesis.

There is a whole zoo of tumor-specific antigens which are more common in tumor cells than in healthy cells and which the immune system can recognize.  Some of these (gp100, NY-ESO-1, p53, HER2, etc) have been used as drug targets.  The problem with this strategy is that, as tumors advance, they evade immune surveillance by losing these distinctive antigens, through a process of natural selection.  It’s not uncommon to lose all HLA class I antigens in many carcinomas.  

Spontaneous remissions of cancers often look immunological in origin, but they are rare. Normally, once a cancer is large enough to observe macroscopically, it only grows or stabilizes, never shrinks; if the immune system could reliably destroy macroscopic solid tumors, you’d see tumors shrinking and growing in size without treatment.  The current theory is that “escape” from immune surveillance happens at a cellular level; cells that present antigens die, cells that don’t survive, and so the cancer shifts over time to become resistant to the immune system.  This suggests that immunotherapies which target specific antigens face similar kinds of strategic considerations to antibiotics; you have to be careful to avoid promoting resistance.  (One consequence of this hypothesis is that giving courses of immunotherapy that are too short in duration could be directly counterproductive.)

The immunology of cancer really is complex and I’ve really only scratched the surface here; what I’m trying to do is to generalize from the standard picture given in the research literature and come to some conclusions as to what types of treatment approaches are and are not likely to be fruitful. My suspicion is that antigen-specific immunotherapies, like growth factor-specific targeted therapies, are not particularly likely to work a priori, despite being popular as drug candidates, for similar reasons; most cancers will have multiple “tricks up their sleeve” and attacking one tumor-specific marker won’t necessarily reach a large proportion of tumor types or produce sustained results.

Dunn, Gavin P., Lloyd J. Old, and Robert D. Schreiber. “The immunobiology of cancer immunosurveillance and immunoediting.” Immunity 21.2 (2004): 137-148.

Waldhauer, Inja, and Alexander Steinle. “NK cells and cancer immunosurveillance.” Oncogene 27.45 (2008): 5932-5943.

Dunn, Gavin P., Lloyd J. Old, and Robert D. Schreiber. “The three Es of cancer immunoediting.” Annu. Rev. Immunol. 22 (2004): 329-360.

Khong, Hung T., and Nicholas P. Restifo. “Natural selection of tumor variants in the generation of “tumor escape” phenotypes.” Nature immunology 3.11 (2002): 999-1005.

Huang, Bo, et al. “Toll-like receptors on tumor cells facilitate evasion of immune surveillance.” Cancer research 65.12 (2005): 5009-5014.