Good News for Immunostimulants

Epistemic Sta­tus: Moderate

Way back in 2015 I was writ­ing about the con­nec­tion be­tween can­cer re­mis­sions and the im­mune re­sponse to in­fec­tion. To re­cap the facts:

  • A plu­ral­ity of recorded spon­ta­neous can­cer re­mis­sions hap­pened when the pa­tient had a strong im­mune re­sponse (of­ten with fever) to a bac­te­rial in­fec­tion at the tu­mor site.

  • William Coley’s bac­te­rial ther­a­pies for can­cer at the turn of the 20th cen­tury, while not tested to the stan­dards of mod­ern ex­per­i­men­tal meth­ods, did seem to pro­duce re­cov­ery rates com­pa­rable or su­pe­rior to chemother­apy.

  • En­do­toxin, a poi­sonous sub­stance found in the outer mem­brane of Gram-nega­tive bac­te­ria, can cause tu­mor re­gres­sions.

  • TNF-alpha, an in­flam­ma­tory cy­tokine in­volved in the body’s re­sponse to en­do­toxin, is equally effec­tive at caus­ing tu­mor re­gres­sions; it is too dan­ger­ous to give to pa­tients sys­tem­i­cally, but is an effec­tive can­cer treat­ment for ad­vanced melanoma when used in iso­lated limb per­fu­sion.

  • There are quite a few cases, both in an­i­mals and hu­mans, of in­flam­ma­tory cy­tokines caus­ing com­plete tu­mor re­gres­sions in metastatic can­cers, par­tic­u­larly when in­jected di­rectly into the tu­mor.

At the time, I pre­dicted that if only there were a de­liv­ery mechanism that could more effec­tively iso­late in­flam­ma­tory cy­tokines to the tu­mor site, it might work safely for more than just spe­cial cases like iso­lated limb per­fu­sion; and that there might be some de­liv­ery mechanism that made a bac­te­rial ther­apy like Coley’s tox­ins work.

The heuris­tic here was that when I went look­ing for the biggest re­sponses (re­mis­sions, com­plete tu­mor re­gres­sions) in the tough­est cases (metastatic can­cers, sar­co­mas which don’t re­spond to chemother­apy), many of them seemed to in­volve this pic­ture of acute, in­tense ac­ti­va­tion of the in­nate im­mune re­sponse.

It turns out that two new ther­a­pies with very good re­sults pretty much sup­port this per­spec­tive.

CpG oli­godeoxynu­cleotides, a mo­tif found in bac­te­rial DNA, are the ac­tive in­gre­di­ent in Coley’s tox­ins; they are the part of bac­te­rial lysate that trig­gers the im­munos­tim­u­la­tory effects.

To­day, SD-101, a CpG oli­godeoxynu­cleotide drug pro­duced by the biotech com­pany Dy­navax, is about to pre­sent its re­sults from two tri­als.

This Jan­uary, Stan­ford sci­en­tists re­ported that SD-101 com­bined with an­other im­munother­apy — but no tra­di­tional chemother­apy — erad­i­cated both im­planted and spon­ta­neous tu­mors when in­jected into mice, both at the in­jec­tion site and el­se­where.

We’ll have to see the re­sults of the hu­man tri­als, but this looks promis­ing.

Another drug, NKTR-214, is an en­g­ineered ver­sion of the in­flam­ma­tory cy­tokine IL-2, de­signed to lo­cal­ize more effec­tively to tu­mors. The IL-2 core is at­tached to a chain of polyethylene gly­cols, which re­lease slowly in the body, prefer­en­tially ac­ti­vat­ing the tu­mor-kil­ling re­cep­tors for IL-2 and re­sult­ing in 500x higher con­cen­tra­tions in tu­mors than a similar quan­tity of IL-2 alone. This is the tu­mor-lo­cal­iz­ing prop­erty that could make in­flam­ma­tory cy­tokines safe.

In pa­tients with ad­vanced or metastatic solid tu­mors, pre­vi­ously treated with PD-1 in­hibitors, NKTR-214 re­sulted in 23% of pa­tients ex­pe­rienc­ing par­tial tu­mor re­gres­sion.

While this still doesn’t mean much chance of re­cov­ery, it’s still no­table — _any _treat­ment for ad­vanced can­cers with more than a 20% re­sponse rate is re­mark­able. (Che­mother­apy usu­ally pro­duces par­tial re­sponse rates in the 2-20% range for metastatic can­cers, de­pend­ing on can­cer type and drug reg­i­men.)

It’s early days yet, but I con­tinue to think that im­munos­tim­u­lants have a lot of po­ten­tial in can­cer treat­ment.

More­over, I think this is a lit­tle bit of ev­i­dence against the fre­quently heard claim that it’s im­pos­si­ble to “pick win­ners” in biotech.

The con­ven­tional wis­dom is that you can’t know ahead of time which drugs that seem to work in pre­clini­cal stud­ies (in vitro or in mice) will suc­ceed in hu­mans.

Most pre­clini­cal drug can­di­dates _do _fail, it’s true. And there are a lot of rea­sons to ex­pect this: mouse mod­els are not perfect prox­ies for hu­man dis­eases, ex­per­i­men­tal er­ror and out­right fraud of­ten make early re­sults un­repli­ca­ble, and we don’t un­der­stand all the com­plex­ities of bio­chem­istry that might make a pro­posed mechanism fail.

But the prob­a­bil­ity dis­tri­bu­tion over drug can­di­dates can’t be uniform, or it would have been im­pos­si­ble to ever de­velop effec­tive drugs! The search space of pos­si­bly bioac­tive molecules is too large, and the cost of ex­per­i­ments too high, to get suc­cesses if drugs were tested truly at ran­dom. We would never have got­ten chemother­apy that way.

I think it’s likely that us­ing the sim­ple heuris­tic of “big effects in tough cases point to a real mechanism some­where nearby” gets you bet­ter-than-chance pre­dic­tions of what will work in hu­man tri­als.