In­vent­ing Med­i­cine by Ac­cess­ing the Hid­den Trea­sure in the Hu­man Mi­cro­bio­me’s Dark Mat­ter

The dark mat­ter of our mi­cro­bio­me

Though rapid progress has been made in un­der­stand­ing the mi­cro­bio­me’s im­pact on hu­man health, a fun­da­men­tal is­sue re­mains large­ly un­solved; and it is pre­vent­ing the field from mak­ing the tran­si­tion from de­vel­op­ing ther­a­peu­tics based on serendip­i­ty to in­vent­ing med­i­cine at scale us­ing en­gi­neer­ing prin­ci­ples. This is­sue, at its heart, is the murky un­der­stand­ing of the so-called “dark mat­ter” in the hu­man mi­cro­bio­me. Col­lec­tive­ly, the dark mat­ter refers to the vast amount of the mi­cro­bial strains that re­main un­cul­tur­able to re­searchers in the lab, and for near­ly any giv­en strain, the ma­jor­i­ty of its genome that can­not be an­no­tat­ed. The in­ac­ces­si­bil­i­ty is­sue of the mi­cro­bial strains can­not be ad­dressed by im­prov­ing our method­ol­o­gy of iso­lat­ing and cul­ti­vat­ing them. It is be­cause, aside from all the lo­gis­ti­cal and bi­o­log­i­cal chal­lenges, the glob­al di­ver­si­ty of the hu­man mi­cro­bio­me, its rate of cre­at­ing new strains, and the ease of se­quenc­ing their genomes, col­lec­tive­ly out­pace such iso­lat­ing and cul­ti­vat­ing ef­forts by or­ders of mag­ni­tude. Along­side the lack of cul­tur­a­bil­i­ty, the an­no­ta­tion is­sue of a strain genome re­lates to our lack of mech­a­nis­tic un­der­stand­ing of its ge­nom­ic parts, which de­mands vig­or­ous mol­e­c­u­lar bi­ol­o­gy work; and tra­di­tion­al­ly, it can on­ly be achieved af­ter the ac­ces­si­bil­i­ty is­sue is ad­dressed.

The im­por­tance of tack­ling these is­sues can­not be over­stat­ed. The hu­man mi­cro­bio­me rep­re­sents the largest and the longest hu­man tri­al ever: there are cur­rent­ly sev­en bil­lion threads run­ning glob­al­ly and every sin­gle thread runs through the en­tire lifes­pan of the host. The mi­crobes, liv­ing on the sur­face and in­side of us, con­stant­ly re­spond to sig­nals from our di­et, med­ical treat­ments, mood, etc through two main in­ter­faces: gen­er­at­ing metabo­lites that ex­ert lo­cal or sys­temic ef­fects, and pro­duc­ing im­muno­genic com­po­nents that elic­it pro­found ef­fects on our im­mune sys­tem. Both in­ter­faces con­tain a large amount of mod­u­la­tors pro­duced by the mi­crobes. For in­stance, a con­ser­v­a­tive es­ti­ma­tion of the num­ber of sec­ondary metabo­lites in the hu­man mi­cro­bio­me pushed the num­ber to be over one mil­lion; and the num­ber of pos­si­ble epi­tope pep­tides en­cod­ed by the hu­man mi­cro­bio­me is even larg­er by at least one or­der of mag­ni­tude. Dif­fer­ent from the oth­er mi­crobes that live in soil or ma­rine en­vi­ron­ments, the hu­man mi­cro­bio­ta are com­prised of mem­bers that have adapt­ed to the hu­man phys­i­ol­o­gy as a re­sult of long term co­hab­i­ta­tion. The adap­ta­tions are in turn re­flect­ed in the com­po­nents that op­er­ate on the above men­tioned two in­ter­faces: they are se­lect­ed by evo­lu­tion­ary forces to be ef­fi­cient at com­mu­ni­cat­ing or mod­u­lat­ing their hu­man host. All these char­ac­ter­is­tics and in­for­ma­tion are per­fect for guid­ing drug de­sign and drug de­vel­op­ment, yet they are con­cealed in the dark mat­ter and are very dif­fi­cult to ac­cess.

Ac­cess­ing the op­por­tu­ni­ty space

Re­cent stud­ies that probed the dark mat­ter gave us a sense of how valu­able they are. A few sec­ondary metabo­lites iden­ti­fied were shown to be im­munomod­u­la­tors; a chem­i­cal screen­ing of fe­cal sam­ples have shown that one third of hu­man GPCRs have lig­and ag­o­nists pro­duced by the mi­cro­bio­me; and a few mi­cro­bio­me-de­rived epi­tope pep­tides elic­it­ing im­mun­odom­i­nance have demon­strat­ed ther­a­peu­tic ben­e­fits in mouse mod­els for ul­cer­a­tive col­i­tis. More­over, on the clin­i­cal side, fe­cal mi­cro­bio­ta trans­plan­ta­tion (FMTs) and live bac­te­ria treat­ments have shown that by al­ter­ing the mi­cro­bio­me, dis­ease states could be changed. Though de­liv­er­ing mi­crobes as a new modal­i­ty of med­i­cine needs to be care­ful­ly reg­u­lat­ed, par­tial­ly for ex­act­ly the fact that they car­ry a sub­stan­tial amount of un­de­fined dark mat­ter, it is with­out doubt that the ev­i­dence demon­strat­ed by those pro­ce­dures val­i­dat­ed the im­por­tance of the mi­cro­bio­ta’s com­po­nents in mod­u­lat­ing the hu­man host. The re­main­ing ques­tion is, how do we ac­cess this hid­den trea­sure in a re­pro­ducible and scal­able man­ner?

We ob­served that there is a big gap be­tween our abil­i­ty to sur­vey the mi­cro­bio­me and our abil­i­ty to en­gi­neer it; and more im­por­tant­ly, this gap is widen­ing. The first prin­ci­ple tells us that the on­ly way to mit­i­gate an is­sue like this is to trans­late the sur­vey­ing pow­er in­to an en­gi­neer­ing pow­er. For­tu­nate­ly, we have a tool that can fa­cil­i­tate this process. With rapid de­vel­op­ment in DNA syn­the­sis, we can now trans­form in sil­i­co ge­nom­ic in­for­ma­tion in­to phys­i­cal form with ease. we ex­pect this “read-write” ax­is to be strength­ened at a fast pace, dri­ven by the strong needs from mul­ti­ple ver­ti­cals such as biosyn­the­sis of in­dus­tri­al chem­i­cals, da­ta stor­age by DNA, pro­tein en­gi­neer­ing, and genome edit­ing by CRISPR sys­tems. With this tech­nol­o­gy, ge­net­ic cir­cuits em­bed­ded in the dark mat­ter that can be con­ve­nient­ly and eco­nom­i­cal­ly re­pro­duced and test­ed with on­ly in sil­i­co da­ta. As a re­sult of con­struct­ing this “read-write-test” ax­is, we can cir­cum­vent the need of cul­ti­vat­ing strains that are very dif­fi­cult to meet, and un­lock the val­ue of the vast amount of se­quenc­ing da­ta of the mi­cro­bio­me. With this pos­si­bil­i­ty at our hori­zon, I want to draw at­ten­tion to the lessons we all learned from the mo­bile phone in­dus­try: it trans­formed it­self from mak­ing sta­t­ic, hard­ware-dri­ven ana­log phones (Nokia was the dom­i­nat­ing play­er) to dy­nam­ic, fea­ture-dri­ven smart­phones (Ap­ple has been a key play­er in this wave) in jaw-drop­ping speed when Steve Jobs in­tro­duced iPhone to the world in 2007. The key fac­tor here is that the in­tro­duc­tion of the iPhone changed the cy­cle of in­tro­duc­ing new fea­tures from years (mak­ing a new phone) to weeks (mak­ing a new app). Could such a trans­for­ma­tion hap­pen in the mi­cro­bio­me field?

Deep­Bio­me Ther­a­peu­tics

In the past years, most mi­cro­bio­me com­pa­nies have been rac­ing each oth­er to ar­rive at mag­ic strain(s) for dis­eases rang­ing from in­fec­tions to cen­tral ner­vous sys­tem dis­or­ders. To dis­cuss the pit­falls of such ef­forts is be­yond this let­ter’s scope. How­ev­er, as the iPhone did not grow out of a vac­u­um, we did not an­tic­i­pate great suc­cess in the mi­cro­bio­me field would come from in­vent­ing drugs in­de­pen­dent­ly from the rest of phar­ma­ceu­ti­cal in­fra­struc­ture by de­liv­er­ing black­box-like live bac­te­ria filled with un­de­fined dark mat­ter. In con­trast, we see this as a per­fect op­por­tu­ni­ty to en­gi­neer a “read-write-test” ax­is to trans­late such mi­cro­bio­me dark mat­ter in­to clear­ly elu­ci­dat­ed nov­el bi­o­log­i­cal in­sight-based ther­a­peu­tics, while work­ing close­ly with test­ed phar­ma­ceu­ti­cal in­fra­struc­ture frame­work. Found­ed by the Broad In­sti­tute and Har­vard Med­ical School alum­ni sci­en­tists Cheng­wei Luo and Alek­san­dar Kos­tic, Deep­Bio­me Ther­a­peu­tics spent the first year at the Pagli­u­ca Har­vard Life Lab build­ing its tech­nol­o­gy plat­form. With a seed round of $7.25M, lead by KTB, Baidu Ven­tures, and the WI Harp­er Group in mid-2019, Deep­Bio­me start­ed de­vel­op­ing its ther­a­peu­tic pipelines in the heart of Kendall Square, Cam­bridge. To date, Deep­Bio­me’s plat­form has an­a­lyzed ter­abytes of mi­cro­bio­me da­ta and gen­er­at­ed drug can­di­dates us­ing in-depth analy­sis and screen­ing meth­ods led by deep learn­ing and oth­er nov­el al­go­rithms. We have shown that a large per­cent­age of these can­di­dates, de­rived from mi­cro­bio­me in­sights, are bioac­tive with pre­ferred phar­ma­ceu­ti­cal char­ac­ter­is­tics. Fur­ther­more, be­cause most of these can­di­dates are di­rect­ly de­rived from the healthy hu­man mi­cro­bio­me, it is not sur­pris­ing that they elic­it very fa­vor­able tox­i­c­i­ty pro­files. More­over, rather than col­lect­ing sam­ples and try­ing to iso­late strains from them, our plat­form was built with scal­a­bil­i­ty in mind since day one to abide by the “read-write-test” mantra. With an ini­tial fo­cus on au­toim­mune dis­eases, we plan to use the same op­er­a­tions to branch in­to near­ly any in­di­ca­tion area in which the mi­cro­bio­me has been demon­strat­ed to play a sub­stan­tial role.

It is note­wor­thy to point out that, de­spite the fact that modal­i­ty and scal­a­bil­i­ty are in our fa­vor, there re­main sev­er­al chal­lenges to be tack­led be­fore we can ful­ly un­leash mi­cro­bio­me’s po­ten­tial. One is the cost of DNA syn­the­sis; though we be­lieve it will even­tu­al­ly de­crease to a point that we can build ge­net­ic cir­cuits em­bed­ded in the dark mat­ter with­out hav­ing to pri­or­i­tize our screen­ing ef­fort, it is cur­rent­ly a speed lim­it­ing step. An­oth­er one is the screen­ing ap­proach it­self, it is of tremen­dous val­ue to en­gi­neer a uni­fied, dis­ease-in­de­pen­dent plat­form; and we ob­serve a few emerg­ing tech­nolo­gies, such as L1000, organoids, and sin­gle cell se­quenc­ing, the could po­ten­tial­ly of­fer a vi­able so­lu­tion. Ul­ti­mate­ly, we firm­ly be­lieve that by un­lock­ing the nov­el bi­o­log­i­cal in­sights hid­den in hu­man mi­cro­bio­me along the “read-write-test” ax­is, our sec­ond genome will of­fer us a pletho­ra of new ther­a­peu­tics.

Author