Pat­terns to Struc­tures

by Christoph

I first saw the im­age be­low a while ago on Twit­ter (now 𝕏) and soon re­al­ized that it's found all over so­cial me­dia, not re­ally "gone vi­ral" but with 70 hits upon googling. I will put down a few thoughts about it here be­cause it il­lus­trates so well how mi­cro­scopists, es­pe­cially cryo-elec­tron to­mo­g­ra­phers, work. And what dis­tin­guishes their way of work­ing from that of artists, and where both meet.

Fig­ure 1. Shadow sculp­ture by Rashad Alak­barov. Fron­tispiece: de­tail of this im­age. Source

How do you "see" the im­age and what's on/in it? I don't want to be in­tru­sive, so I'll say how and what I see, you may fol­low it for your­self. Re­cep­tors in the retina of my eyes reg­is­ter in­ci­dent light by hue and in­ten­sity, and the op­tic nerves of both eyes then add po­si­tional in­for­ma­tion as they trans­mit these sig­nals to the brain. This is, in a nut­shell, how I see. The neu­rons in the vi­sual cor­tex of my brain then as­sem­ble this stream of sig­nals into a three-di­men­sional im­age by match­ing it with pre­vi­ously stored im­ages. This is what I con­sciously see. I don't need to dwell on the term "con­scious­ness," but in prin­ci­ple vi­sual per­cep­tion works sim­i­larly in all ver­te­brates, hu­mans in­cluded. A hall­mark is the un­con­scious abil­ity to rec­og­nize pat­terns in new im­ages based on stored im­ages. Thus, I "see" what I have seen be­fore. This does not work er­ror-free, just think of op­ti­cal il­lu­sions. And when the per­cep­tion of fa­mil­iar pat­terns fails, a feed­back loop can be con­sciously set in mo­tion: "fo­cus on this part" or "look at it from a dif­fer­ent an­gle." Un­til a pre­vi­ously un­known pat­tern be­comes a fa­mil­iar one, and prac­tic­ing that pat­tern, for ex­am­ple, by danc­ing, singing, talk­ing or draw­ing, helps me with fine-tun­ing the vi­sual mem­o­riza­tion (neu­ro­sci­en­tists would agree).

In this im­age, in the shadow on the wall, I in­stantly saw a re­peat­ing, reg­u­lar pat­tern I'm fa­mil­iar with, the cen­tral 12-pointed star, an is­lamic or­na­ment. This be­cause dur­ing my stud­ies of bio­chem­istry and mi­cro­bi­ol­ogy I dab­bled in such pat­terns that I had first seen in floor and wall mo­saics in the cathe­dral of Ce­falù, Sicily. All I needed to recre­ate such pat­terns was a sheet of pa­per, pen­cil, ruler, com­pass, and get­ting into Pythago­ras' the­o­rem (see here an ex­am­ple from Anna Maksimova's tu­to­r­ial).

Fig­ure 2. A  Top view of a chemore­cep­tor ar­ray in a Sal­monella en­ter­ica mini­cell. OM, outer mem­brane; IM, in­ner mem­brane. Scale bar: 100 nm. Source. B  Side view of a mem­brane-bound chemo­receptor ar­ray (MA) and a cy­to­plas­mic chemore­cep­tor ar­ray (CA) of Vib­rio chole­rae. The cy­to­plas­mic ar­ray is com­posed of two par­al­lel CheA/W base­plates ap­prox­i­mately 35 nm apart. The che­moreceptors are sand­wiched be­tween the two ba­se­plates and are hexag­o­nally packed with a 12 nm cen­ter-to-cen­ter spac­ing. OM, outer mem­brane; IM, in­ner mem­brane. Scale bar: 50 nm. Source

Know­ing this pat­tern so well, I im­me­di­ately no­ticed that the shadow is slightly dis­torted. Do you see that too? It be­comes even clearer when you see other im­ages of this art­work by the Azer­bai­jani artist Rashad Alak­barov, for ex­am­ple here and here. This dis­tor­tion made me re­al­ize that the shadow came from a light source that the artist had placed be­hind a rather chaotic look­ing pile of mikado sticks (square-sec­tion metal tubes of vary­ing lengths for this im­age, wooden sticks for the other two linked im­ages, ap­par­ently). Could I re­con­struct the three-di­men­sional shape of this stack by just look­ing at it from one an­gle? No way.

I have no idea if the artist in­tended an al­lu­sion to Plato's Al­le­gory of the Cave with this work (my take on Plato: you can't per­ceive re­al­ity but only see its shadow). From the many com­ments on this im­age on "so­cial" me­dia, how­ever, I have learned that es­o­ter­i­cally in­clined peo­ple like to un­der­stand it − and are com­fort­able with it − in the sense of the psy­cha­trist and psy­cho­an­a­lyst C.G. Jung: "In all chaos there is a cos­mos, in all dis­or­der a se­cret or­der" (cit.). Mi­cro­scopists can do lit­tle with such a de­fin­i­tive sound­ing blurb.

The ac­tual work of the elec­tron mi­cro­scopists starts when they get to "see" such strik­ing pat­terns as in Fig­ure 2, an ar­ray of fairly reg­u­lar hexa­gonal tiles in a ~100 nm wide "patch" in Fig­ure 2A, and ~15 nm thick and ~100 nm long "sand­wiches" that ei­ther lo­cal­ize in the cy­to­plasm or abut the in­ner cell mem­brane in Fig­ure 2B. They know that these im­ages are "shadow im­ages," sig­nals cast onto a sen­si­tive dig­i­tal screen by a fo­cused elec­tron beam di­rected through a su­per-thin three-di­men­sional ob­ject (the sig­nals recorded by the sen­sor are, of course, nu­mer­i­cal val­ues that can only be viewed by con­vert­ing them into pix­els).

Fig­ure 3. Prin­ci­ple of elec­tron to­mog­ra­phy. a Schematic rep­re­sen­ta­tion of data ac­qui­si­tion. A flex­i­ble rope knot rep­re­sents the ob­ject, em­pha­siz­ing that elec­tron tomo­graphy can re­trieve 3D in­for­ma­tion from struc­tures with in­di­vid­ual topolo­gies. A set of pro­jec­tion im­ages is recorded on a charge-cou­pled de­vice cam­era while the ob­ject is tilted in­cre­men­tally around an axis perpen­dicular to the elec­tron beam. Ow­ing to the lim­ited ac­cu­racy (eu­cen­tric­ity) of the tilt­ing de­vice, the speci­men has to be re­cen­tred and re­fo­cused at each tilt an­gle. Au­to­mated pro­ce­dures have been de­vel­oped to per­form this task with neg­li­gi­ble ex­po­sure of the ob­ject to the elec­tron beam. b The back-pro­jec­tion method ex­plains the prin­ci­ple of the 3D reconstruc­tion in an in­tu­itive man­ner. For each pro­jec­tion, a back-pro­jec­tion body is cal­cu­lated, and the sum of all pro­jec­tion bod­ies yields the den­sity dis­tri­b­u­tion of the orig­i­nal ob­ject − the to­mo­gram. To com­pen­sate for the fact that high-res­o­lu­tion fea­tures change more rapidly with tilt an­gle than do low-res­o­lu­tion fea­tures, an ap­pro­pri­ate weight­ing func­tion has to be ap­plied to the data in the 2D im­ages be­fore cal­cu­lat­ing the re­con­struc­tion. Source

In Fig­ure 2, the "patch" is from Sal­mo­nella and the "sand­wiches" are from Vib­rio, but I still dare to show these, at first glance, com­pletely dif­fer­ent struc­tures from dif­fer­ent or­gan­isms in one fig­ure. In about a dozen bac­te­ria of dis­tant phy­lo­ge­netic re­la­tion­ships ex­actly these two struc­tures were found in as­ton­ish­ingly sim­i­lar di­men­sions. The mea­sured di­men­sions of the "pat­ches" and "sand­wiches" sug­gested that they were dif­fer­ent views of the same cell struc­ture but this had to be proven first.

The re­searchers used two com­mon tech­niques to make sure that their own pat­tern recogn­ition ca­pa­bil­ity was not play­ing tricks on them here. First, they ob­tained power spec­tra − tech­nically com­pa­ra­ble to X‑ray crys­tal­log­ra­phy − of Vib­rio "patch" im­ages show­ing that they have pe­ri­od­ic­i­ties like crys­talline struc­tures (see here, top in­sert). Sec­ond, they used subvol­ume av­er­ag­ing to ver­ify that the "patches" show­ing a hexag­o­nal pat­tern in the top view (Fig­ure 2A) are in­deed hexag­o­nal (see here, bot­tom in­sert).

Note that mi­cro­scopists have no con­trol over how their snap-frozen cells are "caught" in a sam­ple, whether they rest on their sides or stand up­right or at an an­gle, they know they find iden­ti­cal three-di­men­sional cel­lu­lar struc­tures in a wide va­ri­ety of ori­en­ta­tions.

An aside. There is no Wikipedia en­try on "chemosen­sory ar­rays" in bac­te­ria and ar­chaea yet, but if you want to see and know more, I rec­om­mend you to take a look at the re­spec­tive chap­ter of the Cell­struc­ture­at­las, which we al­ready called a source of ex­cel­lent teach­ing and learn­ing ma­te­r­ial here and here.

To make a three-di­men­sional "scan" of rel­e­vant cel­lu­lar struc­tures pos­si­ble, cryo-elec­tron mi­cro­scopists use a trick called "back pro­jec­tion." First, a se­ries of im­ages (=data) is taken of an ob­ject at dif­fer­ent an­gles, that is, the ob­ject is tilted in­cre­men­tally around an axis perpen­di­cular to the elec­tron beam (Fig­ure 3). Sec­ond, for each pro­jec­tion, a back-pro­jec­tion body is com­pu­ta­tion­ally cal­cu­lated, and the sum of all pro­jec­tion bod­ies yields the den­sity dis­tri­b­u­tion of the orig­i­nal ob­ject, the to­mo­gram. As Sali et al. (2003) write: "The qual­ity of a to­mo­gram de­pends crit­i­cally on cov­er­ing as wide a tilt range as pos­si­ble (typ­i­cally ±70°), with tilt incre­ments as small as pos­si­ble (1° to 3°). How­ever, each ad­di­tional ex­po­sure to the beam in­crea­ses the amount of ra­di­a­tion dam­age and the cu­mu­la­tive dose must not ex­ceed a tol­er­a­ble limit." That is, a limit of ap­prox­i­mately 100 electrons/Å2.

Fig­ure 4. Ar­chi­tec­ture of Chemosen­sory Ar­rays in E. coli. (A) The side view and the top-down view of a sig­nal­ing core unit is de­picted in car­toon. The in­ter­face 1 within the core unit is marked with a red line. (B) Re­peats of the sig­nal­ing core units as­sem­ble into ar­rays through in­ter­face 2, marked in black lines. Red and black cir­cles in bro­ken lines show CheA-filled and CheA-empty rings in base­plate, re­spec­tively. © To­mo­graphic im­age of the chemosen­sory ar­rays in the side view near a fla­gel­lar mo­tor in a lysed E. coli cell. (D) Cryo-elec­tron tomo­graphy (cryo-ET) im­age of chemosen­sory ar­rays in the top view in a lysed E. coli cell. The in­sert panel shows subto­mo­gram av­er­ag­ing of the hexag­o­nally packed re­cep­tor trimers-of-dimers. Source

Us­ing such to­mo­grams ob­tained from many hun­dreds of com­pu­ta­tional "aligned" in­di­vid­ual im­ages, Yang & Briegel (2019) were able to chart the sub­struc­tures of the che­mosensory ar­rays of E. coli at near atomic res­o­lu­tion and map their spa­tial arrange­ment (Fig­ure 4). How­ever, since elec­tron den­sity maps are hardly "read­able" ex­cept for very trained eyes, the au­thors have rep­re­sented them diagram­mati­cal­ly (Fig­ure 4A+B). They ex­plain: "The chemo­receptor ho­mod­imers read­ily form trimers-of-dimers through the in­ter­ac­tion at their cyto­plasmic tips (Fig­ure 4A). To form the sig­nal­ing core units, re­cep­tor trimers bind both the ki­nase and cou­pling pro­tein fol­low­ing a strict sto­i­chiom­e­try of 6 re­cep­tor trimers-of-dimers: 1 dimeric CheA:2 monomeric CheWs (Fig­ure 4B). Among the five do­mains of the ki­nase CheA, the P5 do­main di­rectly binds to the re­cep­tor trimers. The P5 do­main topo­log­i­cally re­sem­bles two tan­dem SH3 do­mains. Sim­i­larly, CheW is com­posed of two β‑barrels sand­wich­ing a hydro­phobic core. CheW and P5 bind each other in an al­ter­nat­ing or­der and form pseudo six­fold sym­met­ric rings that link core units to­gether. Within the in­di­vid­ual sig­nal­ing core unit, CheW and P5 es­tab­lish in­ter­face 1; among the sig­nal­ing core units, CheW and P5 form in­ter­face 2, which is cru­cial for form­ing the ex­tended ar­ray lat­tice. Through the net­works of CheAs and CheWs, the re­cep­tor trimers-of-dimers are arranged into a rigid hexag­o­nal pat­tern with a spac­ing of ~12 nm. In the side view, ar­rays are dis­tin­guish­able as a con­tin­u­ous layer par­al­lel to the in­ner mem­brane (Fig­ure 4C). The chemore­cep­tors can be seen as pil­lar-like den­si­ties that ex­tend be­tween the base­plate and the mem­brane."

Fig­ure 5. Mol­e­c­u­lar ar­chi­tec­ture of bac­te­r­ial chemo­sensory ar­rays. Red: chemore­cep­tor trimers-of-dimers, green: CheW adap­tor, blue: CheA ki­nase, turqois: bac­te­r­ial mem­brane (IM). Source. See here for the com­plete Fig­ure: Con­struc­tion and re­fine­ment of an ato­mistic model of the bac­te­r­ial chemosen­sory ar­ray. X‑ray struc­tures of the ki­nase CheA, adap­tor pro­tein CheW, and chemore­cep­tor from the ther­mophilic bac­terium T. mar­itima were used to model key ar­ray sub­struc­tures, which were arranged ac­cord­ing to cryo-ET den­si­ties to pro­duce mod­els of the ex­tended ar­ray ar­chi­tec­ture, namely the CheA2-trimer and CheA2-hexa­mer assem­blies. Sub­se­quently, mol­e­c­u­lar dy­nam­ics flex­i­ble fit­ting (MDFF) sim­u­la­tions with sym­me­try re­straints were car­ried out to re­fine the struc­tures of the com­po­nent mod­els to their ar­ray-bound con­for­ma­tions. Source

With the struc­tural data of the base­plate com­po­nents, the hexa­gons, and the in­trigu­ing tri­mers-of-dimers re­cep­tor oligomers in hand, or bet­ter, in files: would it be pos­si­ble to build a com­plete three-di­men­sional model, the "pile of mikado sticks," of a chemosen­sory ar­ray at near atomic res­o­lu­tion? Goh et al. (2016) at­tempted and man­aged to do just that (Fig­ure 5). Their com­pu­ta­tional method­ol­ogy, out­lined in de­tail in their re­view, is too com­plex to re­capitulate here in brief, but see the leg­end to Fig­ure 5. I will limit my­self to show­ing their model for a chemosen­sory ar­ray of Ther­mo­toga.

An­noy­ingly, this is where the im­pos­si­bil­ity of ap­prox­i­mat­ing the three-di­men­sion­al­ity of mo­dels on pa­per or a com­puter screen, that is, two-di­men­sion­ally, comes into play. I wish I could at least see this ar­ray an­i­mated, or, bet­ter yet, in a ver­sion I could ac­tively zoom into and wan­der around like in Google Earth. Is that too much to ask? The data is there and Keith Cas­sidy has se­lected only one pos­si­ble view for the im­age of the Ther­mo­toga chemosen­sory ar­ray. But the ef­fort of pro­gram­ming an an­i­ma­tion of the im­age is eas­ily com­pa­ra­ble to Rashad Alak­ba­rov's work of pil­ing up the "Mikado sticks" by trial-and-er­ror tin­ker­ing such that when illu­mi­nated they pro­duce this dis­tinct shadow of a 12-pointed star in Fig­ure 1.

One last point. If you are miss­ing clues from the Ther­mo­toga model of the chemosen­sory ar­ray as to how it might be em­bed­ded in the cel­lu­lar con­text, you can be helped. Good­sell & Lasker (2023) have mod­eled a mem­brane-bound chemosen­sory ar­ray into a Caulobac­ter cell, of course also based on struc­tural data (see here).

Do you want to com­ment on this post? We would be happy about it! Please com­ment on Mastodon, or on Bluesky.

Other Posts

  • A Vi­ral Pyra­mid Scheme

    by Merry Youle — In or­der to re­lease their newly as­sem­bled viri­ons, most viruses lyse the cells that have fed and housed them. This ly­sis is not a hap­haz­ard af­fair. Some phages, for ex­am­ple, em­ploy a holin-en­dolysin sys­tem to rup­ture their host's cell mem­brane and di­gest the cell wall at a pre­cisely con­trolled…

  • Where in the world is ERV-Fc?

    Trac­ing the spread of a virus over 30 mil­lion years, 28 species, and 5 con­ti­nents
    by Zachary Williams — When doc­tors in the 1980s no­ticed that peo­ple were dy­ing from se­vere im­mune de­fi­ciency, no­body had ever seen any­thing like it be­fore; it was a com­pletely novel and ter­ri­fy­ing dis­ease. Once we fig­ured out it was caused by a virus,...

  • Pa­le­ovi­rol­ogy

    by Welkin John­son — How does one even be­gin to in­ves­ti­gate the nat­ural his­tory of viruses? The di­nosaurs be­queathed a mot­ley as­sort­ment of bones, teeth, foot­prints strid­ing 'cross an­cient riverbeds, fos­silized eggs, the oc­ca­sional co­pro­lite. The tiny trilo­bite left last­ing and ubiq­ui­tous im­pres­sions, find­ing its way into...

  • Lost & Found #2

    by Christoph — On our way out of 𝕏 (for­merly Twit­ter), we look through our soon-to-be-ob­so­lete "book­marks" and show you what shouldn't just end up in the bin. Here are three more sal­vaged book­marks in 'Lost & Found #2'.

  • An Aro­matic Snip­pet

    by Elio — As friends gather at the din­ner ta­ble, you open that old bot­tle of wine you've been sav­ing for a spe­cial oc­ca­sion. You pour out a small amount in a stem glass, swirl it, and bring it to your nose. You de­tect the aroma of var­i­ous fruits and a hint of flow­ery scent with a note of honey...