Good.  All  right. Please  come  in.  Take  your  seats. We're  going  to  get  started. Welcome  back  to  the  Georgia  Tech  Neuro  Seminar  series. First  a  quick  announcement. The  Neuro  Next  initiative  here  is  going  to  have our  kickoff  day  on  August  25  in  the  evening, with  coming  to  give our  guest  lecture  of  October.  Yes.  Of  this  month. This  month  now.  Thank  you. And  then  on  the  26  will  be  our  kickoff  day. And  so  you  should  have  all  gotten  e  mails  about  that. If  you  haven't  gotten  an  e  mail about  that  and  would  like  to  participate, please  come  see  me  or  Chris  or  Sarah  up  here  am. And  we  can  give  you  more  details  on  how  to  sign  up  today. It's  my  pleasure  to  introduce  Brad  Dickerson. Brad  got  his  undergraduate  degree  in biology  from  Swarthmore  before  he  worked  with Ti  Hedrick  at  UNC  and  then  came as  a  graduate  student  to  the  University  of  Washington, where  I  got  to  overlap  with him  in  the  lab  of  Tom  Daniel  for a  while  while  I  was  a  post  op there  and  he  was  a  graduate  student. He  then  won  multiple  fellowships  to  go  work  with Michael  Dickinson  at  Caltech  for  his  post  doc  work. And  before  he  even  got  done  with  those  fellowships, UNC  snapped  him  up  again, and  he  went  back  to  UNC  for  his  faculty  position. But  then  two  years  into  that, he  got  coached  by  Princeton. So  a  record  of  extreme  success  here. But  the  thing  that  I  really  want  to  say about  Brad  is  when  he's an  incredibly  broad  thinker  and I've  always  really  enjoyed  my  discussions  with  Brad, He's  one  of  those  people  I  try  really  hard  to  listen to  encourage  you  all  to  try  and  do  that  today  too, whenever  I  get  drosophila  envy  in  my  work. It's  not  for  folks  who  are necessarily  just  generating  the  tools. It's  people  like  Brad  who  I  think  are  asking really  profound  questions  while leveraging  all  of  those  really  amazing  tools. And  so  we'll  hear  a  little  bit  about that  today  as  we  go  along. So,  Brad,  thanks  for  coming.  Please  take  it  away. Well,  thank  you.  That's  a  very  kind  introduction. Simon.  Um,  and  thank  you  for  the  invitation. It's  wonderful  to  be  here  and  I  look  forward. I've  already  had  some  really,  really nice  conversations  already. I  look  forward  to  talking  more with  a  few  of  you  later  today.  All  right. So  what  I  want  to  do  is  tell  you a  bit  about  the  work  that  we  do  in  the  lab. And  I'm  actually  from  the  Princeton  area, so  I  was  about  to  say  what  we  do  at  home, and  it  would  be  appropriate  in  a  few  different  ways. But  first,  all  I  want  to  do  is  kind  of  frame the  work  that  the  lab  does. The  way  I  like  to  think  about  our  work  is that  animals  are  both  stable  and  maneuverable. And  that's  true  across  different  styles  of  locomotion. So  whether  we're  talking  about  this  lemer  jumping  through the  forest  or  this  top  down  view  of  a  hummingbird. In  both  cases,  we  see  that  these  animals can  both  navigate  really  complex  environments. They  can  also  execute  the  really  subtle  changes  in locomotion  to  maintain  specific  body  postures while  they're  moving  around. Another  thing  that's  really  important in  these  two  cases  is  that  we're  talking  about  very, very  different  time  scales,  right? Because  this  hummingbirds  flapping its  wings  at  like  25  times  a  second. Now,  timing  is  also  really important  for  the  nervous  system. We've  identified  in  a  few  different  cases, circuits  for  detecting  timing  differences  at the  nano  or  microsecond  timescale to  execute  different  kind  complex  behaviors. Whether  we're  talking  about  hunting  in  the  case of  barnls  or  echo  locating  bats, or  detecting  con,  specifics in  the  case  of  weekly  electric  fish. In  the  case  of  detecting  timing  differences, we  find  a  couple  of  things in  those  cases,  these  circuits. First,  we  see  that  there  are  all  sorts of  bio  mechanical  filters that  help  with  collecting information  about  the  sensory  environment. In  the  case  of  barns, the  ears  are  actually  at  different heights  of  the  head  to  aid in  detecting  inter,  oral  timing  differences. So  that  the  animal  can  detect  where  in as  myth  a  mouse  is. The  other  thing  is  that  sensory  information tends  to  be  organized  in, in  this  case  this  is  the  cricket  circle  system. And  they  can  detect  wind both  direction  and  velocity  really,  really  sensitively. And  that  information  gets  made into  a  map  in  the  central  nervous  system. And  focusing  on  maps, those  maps  can  come  in  two  broad  flavors. Topographic  maps,  which  are just  basically  information  laid  out  in space  like  this  recent  work  in the  case  of  fruit  fly  eye  and  detecting  looming. But  also  in  the  case  of  going  back  to  barnows, we  have  this  case  of  computational  maps where  sensory  information  is  re, routed  based  on  some  computational  principle for  the  central  nervous  system. Okay. But  everything  I've  just  told  you  about  is  in  the  case  of sensory  systems  something  that we're  beginning  to  appreciate  more  and  more. And  this  is  work that  both  a  couple  of  people  in  the  audience. Lina,  Ting  and  Simon  have. Have  brought  up  is  the  role  of timing  in  the  motor  system,  right? And  so  we're  beginning  to appreciate  more  and  more  the  role  of sub  millisecond  timing  differences in  the  case  of  bird  song, in  the  case  of  human  motor  control. And  my  favorite  study  system,  insect  flight. Okay?  And  I  used  to  work  on  moths, and  that's  where  Simon  and  I  met. But  what  I  want  to  do  today  is  talk  to  you  a  bit about  this  kind  of  problem  of bridging  the  gap  between timing  differences  in  sensory  and motor  control  and  maps  and  bio mechanical  filters  in  a  specific  case  of  insect  flight, which  is  fly  flight. So  this  is  a  high  speed  video  of  a  hover fly  out  in  Washington  State. And  there  are  a  couple  of  things I  want  you  to  appreciate. First  is  that  this  is  happening  very,  very  quickly. This  animal  is  beating  its  wings about  200  times  a  second. It  has  to  rapidly  collect sensory  information  from  its  environment  and  turn it  into  changes  in  wingstroke on  a  moment  to  moment  basis. The  other  thing  is  that  this  animal can  do  all  sorts  of  crazy  maneuvers,  right? This  animal  is  both  very,  very  stable,  right? Because  it  can  just  hover  in  place, but  it  can  also  execute  these  rapid  changes  in  behavior. I  think  for  me,  fly  flight  is a  really  nice  system  to  explore all  of  these  different  problems  together,  okay? But  as  much  as  I  would  like  to  work  on  hover  flies, there  are  a  number  of ways  that  they're  less  convenient  to  work  on. I  work  on  fruit  flies  drop,  lana  gaster. And  there  are  a  couple  of  reasons  that drosoph  are  a  good  system for  studying  this  problem,  right? They  already  fly,  right?  They're  called  flies. But  we  also  have  the  benefit  of the  genetic  tool  kit  and  roof  for labeling  and  manipulating  different  cell  types. So  this  is  just  a  confocal  micrograph labeling  the  wing  stirring  muscles. I'll  talk  about  in  a  little  bit  with  GFP,  right? So  we  can  label  and  manipulate  different  cells, turn  them  on  and  off  as  we  see  fit. But  the  other  thing  that's  really  nice  about flies  is  that  they  fly  well  in  a  laboratory  setting. So  this  is  a  set  up  that  is  very  common  in  my  lab, where  we  have  a  fly  tethered  to  a  pin  and  then  we place  it  in  between  an  infrared  LD and  a  pair  of  phot  detectors. So  what  we  can  do  is  in  real  time  track  wing  motion, and  then  also  use  that  output  to  control a  visual  display  where the  fly  is  basically  playing  a  video  game. And  it  can  operate  either  closed  loop, where  it  has  control  over  its  sensory  experience, or  we  can  just  play  some  open  loop  sensory  signals and  see  how  the  animal  responds. Let  me  just  show  you  what  this  looks  like. In  this  case,  we  have  the  fly  that's  tethered  right  here. You  can  see  the  change  in the  wing  stroke  envelope  and  in  this  case the  difference  in  left  minus  right  wing  stroke  amplitude is  controlling  the  angular  velocity  of  the  stripe  here. And  this  is  a  really  powerful  method for  really  just  uncovering how  an  animal  turns  sensory  information  into  behavior. What  we  can  further do  with  the  genetic  tools  available  in  drosoph, we  can  start  to  untangle what's  going  on  more  mechanistically. Unlike  other  flying  animals  like birds  and  bats,  flying  insects, all  the  power  and  control  for  flight resides  in  the  animal  floraxy, these  large  power  muscles  for  flapping  the  wings. But  they  have  a  set  small  steering  muscles on  the  lateral  walls  of  the  thorax that  control  the  subtle  maneuvers that  I'm  more  concerned  about. They  attach  cuticular  elements in  the  thorax  called  sclerites. You  could  think  of  them  as  tendon  like elements  that  help  reconfigure  the  wing  hinge, to  change  wing  motion  and  aerodynamic  forces. And  there  are  four  major sclerites  that  we're  concerned  with. The  baslerte,  first  axillary, the  third  axillary,  and  the  fourth  axillary, which  for  historical  reasons  is  called  the  HG. Each  of  these  sclerites  has  a  set of  steering  muscles  associated  with  that,  attached  to  it. There  are  a  couple  of  things  I want  to  bring  to  your  attention. First  is  just  that  there aren't  very  many  muscles  here,  right? I'm  talking  about  12  muscles  total. The  other  thing  is  that  each  of these  steering  muscles  is  interved by  a  single  motor  neuron,  right? Compared  to  vertebrate  motor  control  systems, you  just  have  one  motor  neuron for  each  muscle  in  this  case. One  question  that  motivates  our  work  is, how  is  it  you  see flying  insects  execute  all  these  very  graceful  maneuvers, But  they're  doing  it  with  a  very  limited  set of  basically  control  *****. So  how  are  they  able  to  accomplish  that? Okay,  the  wing  steering  system can  be  broken  down  anatomically, but  we  can  also  further  break  it  down  functionally. And  there  are  two  major  flavors  of  steering  muscles. For  flies,  we  have  muscles that  are  said  to  be  tonically  active. In  this  case,  flies  are  beating  their  wings at  about  200  times  a  second  or  even  more. In  some  cases,  they  don't  have  really  the  ability to  modulate  the  spike  rate  of  those  neurons. What  they  can  do  instead  is  change  when  the  stroke  cycle. Those  muscles  are  active.  These  muscles fired  a  precise  time, or  phase  in  the  stroke  cycle  by introducing  what's  called  a phase  delay,  or  phase  advance. We  know  from  biomechanical  analysis  that this  changes  the  mechanical  properties  of  the  muscle. These  muscles  that  are  tonically  active  are  more involved  in  relatively  slow  stabilization  reflexes. Then  the  other  flavor  are  muscles  that  are  a  little bit  more  intuitive  and  they're  called  physically  active. Where  they're  typically  inactive, but  then  they  come  on  in  bursts and  they're  more  involved  in  active  maneuvers. But  in  both  cases,  when  the  stroke  cycle, these  muscles  are  active  is  really important  because  phasic  muscles, even  though  they  fire  in  a  burst, they  only  fire  once  per  wing  stroke. When  the  stroke  cycle,  these  muscles  are  active, really  determines  how  the  wing  hinge  is  reconfigured  to change  wing  motion  and  aerodynamic  forces  in  the  flies. Trajectory  timing  is  really  important. How  do  flies  control  timing  of  the  wing  series  system? One  other  thing  I  should  point  out that's  really  important  for  stuff  I'll  talk about  later  is  that  if  you  look at  steering  muscle  activity as  a  function  of  the  turn  magnitude, you  see  that  muscles  that  are  tonically  active in  the  function  as  a  ****,  they're  linearly  recruited. Or  as  muscles  that  are  physically active  are  non  linearly  recruited. But  in  both  cases, timing  is  really  important. Okay,  so  apologies  that  flies  a  little  light. But  what  you  can  imagine  is  if  we're recording  a  wing  stering  muscle  motoron', in  this  case  we're  recording  from a  tonically  active  muscle. It's  firing  once  per  wing  stroke at  a  precise  time,  the  stroke  cycle. The  easiest  thing  you  can  imagine  is  that  there's some  descending  neuron  from the  brain  providing  input  to determine,  to  determine  this  output. And  there's  certainly  descending  neurons  that provide  input  to  the  wing  steering  system. But  none  of  them  fire  in  a  phase  lock  fashion. Instead,  they  actually  show graded  changes  in  membrane  potential. They're  more  consistent  with  the  spike  rate  code. The  idea  for  a  long  time  has  been that  what  flies  do  is  actually combine  visual  input  with rapid  wing  beat  synchronous  mechansensoryeedback, that's  arriving  wing  stroke  to  wingroke to  both  structure  and adjust  the  timing  of  the  wing  steering  system. Okay,  one  thing  that's  nice  about  this  is  that  there  are certainly  mechan  sensors  on  the  wings that  provide  feedback  on a  wing  stroke  to  wing  stroke  basis. But  in  terms  of  adjusting  the  wingtroke, there's  something  a  bit  problematic  about  using wing  beat  sequence  feedback  only  from  the  wings. Which  is  that  the  feedback  from the  wings  is  always  telling  you  basically about  what  the  wing  did  in  current  wings, you  can't  change  the  feedback  from  the  wing. This  feedback  is  the  direct output  of  the  wing  steering  system. If  the  fly  wants  to  adjust  mechan, sensory  feedback  rapidly  to  change the  motor  output  relying  on  the  wing, the  can  sensors  alone  isn't  going  to  cut  it. Okay.  This  is  where  flies, being  flies  is  really  helpful.  Because  flies. You  may  recall  that  compared to  most  flying  insects  that  which  have  four  wings, flies  only  have  two  dynamically  functional  wings where  they  would  have  their  hind  wing. They  have  this  structure  here.  I'm  going  to  spend basically  the  rest  of  my  talk  discussing, which  is  called  the  hat  here,  is  essential  to  fly. If  you  cut  it  off,  a  fly  can't  fly. They  can  flap  their  wings,  but  they can't  stabilize  themselves. All  right,  one  of  the  things  that  we  know  about, about  the  structure  of  the  halt  here  is  that  it serves  as  a  gyroscopic  sensor,  detects  body  rotations. But  the  other  thing  about  the  hall  tear  is  that it's  specific  to  flies. So  basically  anywhere  that  any  flies  you  can  think  of, so  whether  you're  thinking  about  mosquito, house  fly  or  sofa  or  horse  fly excuse  me  or  of  they  all  have  hal  tears,  right? There's  another  group  of  insects  strep  sip  where  they, where  their  front  wings  look  like a  hal  tear,  but  they're  parasites. And  there's  only  one  paper where  people  have  studied  them  and  it's  an  end  of  three. So  we're  just  going  to  focus  on,  on  flies. Okay.  Okay.  So  I mentioned  that  if  you  cut  off  the  hal,  tears  flies. Can't  fly.  And  I  said  that  the  evidence suggests  that  Hales  gyroscopic sensors.  What  do  I  mean  by  that? If  we  imagine  that  a  fly  is  beating  its  wings back  and  forth  in  the  halt  here  in  the  wing  anti  phase, this  side  line  just represents  the  tip  path  of  the  halt  here. Let's  just  take,  for  example, a  case  where  the  fly  gets  rotated  about  pitch. The  halt  here  has  a  tendency  to resist  this  change  in  its  plane  of  oscillation. What's  going  to  end  up  happening,  is its  tip  path  rejectoryes  going  to  change. And  it's  going  to  experience something  called  the  Corriols  force. Which  is  equal  to  the  cross  product of  the  angular  velocity of  the  fly's  body  about  pitch  in  this  case, and  the  tip  velocity  of  the  halt  here. This  is  a  wave  for  flies  detect both  the  direction  of  rotation and  the  actual  magnitude  of  rotation. Okay,  what's  the  evidence  for  this? What  we  can  do  is  build  a  simulator,  a  flight  simulator. This  is  some  previous  work  by my  post  hoc  advisor  called  the  Rock  and  Roll  Arena, where  just  as  you  might  imagine, you  can  rock  the  fly  around  and  about  different  axes. In  this  case,  we  think  about pitch  like  I  was  saying  before, what  we  can  do  is  see  with  an  intact  fly, we  get  these  changes  in  wingstroke  amplitude  that are  compensatory  and  if  you  remove  the  whole  tears, then  the  fly  can't  compensate,  okay? All  right,  so  the  whole  tier  is  a  gyroscopic  sensor and  if  we  take  a  closer  look  at  the  whole  tier. This  is  a  con,  local  micrograph  of a  Drosophohl  tear  labeled  with  GFP. We  see  these  rows  and  rows  of  the  sensors. These  are  mechanic  sensors  called  companiform  sencilla. They  detect  strain  on  the  exoskeleton and  really  anywhere  on insects  that  you'll  find  bending  and  twisting, you'll  find  companiform  sencilla  legs, in  some  cases,  antenna  the  wings. And  in  the  case  of  flies,  the  way  these  work is  you  have  this  dome  shaped  structure that's  embedded  in  spongy  tissue. When  the  skeleton  bends  and  twists,  this  cap, that's  attached  to  some  connective  tissues that's  connected  to  the  dendrite. It  rises  and  falls  and  that opens  up  mechanosensitive  Vion  channels. And  then  you  get  the  firing  of  an  action  potential. And  the  wings  also  have  these  sensors. And  that's  consistent  with  the  idea  that the  whole  tears  are  evolved  from  the  hind  wing. But  the  other  thing,  observation  that  underscores the  whole  tears  evolutionary  history  as  the  hind  wing  is that  it  also  has a  power  muscle  in  a  set  of  steering  muscles. We  don't  need  to  get  caught  up  in  the  names, we  can  just  break  them  into  two  major  groups. More  anterior  basilar  muscles and  more  posterior  axillary  muscles. The,  the  gyroscopic  structure, which  helps  fly  remain  stable. But  this  reflex  is  very, very  sensitive  and  presents  a  problem. Which  is  that  how  can  flies  execute any  maneuver  if  they are  constantly  triggering  this  reflex,  right? We  know  from  some  previous  work that  the  Lear  muscles  receive  visual  input and  this  system  may  be  a  way  that flies  can  be  both  maneuverable while  maintaining  stability. Okay,  that  sets  up the  work  that  I want  to  talk  to  you  about  that  we're  doing  in  my  lab. And  we're  going  to  frame  it  under  this  question  of  how animals  maneuver  without  sacrificing  stability. And  we're  going  to  think  about the  halt  here  in  three  different  ways. First  we're  going  to  think  about the  role  of  the  halter  motor  system  in  flight. Because  if  the  halter  motor  system  is  active  in  some  way, that  could  mean  that  the  halt  here  is  not just  a  passive  gyroscopic  sensor. Instead  it's  a  multi  functional  sensory  organ. Then  we're  going  to  think  about  if  that's  the  case, how  is  the  multi  functionality  of  the  L  tear  achieved. And  then  finally  think  about how  feedback  from  the  hall  tear  centrally organized  in  the  central  nervous  system. Or  organized  in  the  central  nervous  system. All  right,  so  let's  start  here. Okay,  so  just  to restate  what  I  said  a  little  bit  more  formally, just  a  couple  of  slides  ago. If  we  imagine  a  fly  is  flying and  it's  receiving  visual  input, that  information  can  be  sent  to  the  hall  tear  muscles which  could  change  the  motion  or mechanics  of  the  hull  tear  in  some  way. Because  the  hall  tear  has  a  bunch  of mechan  sensors  embedded  within  it  that would  change  the  mechan  sensory  feedback arriving  on  a  windthroke  to  windthroke  basis. Because  the  halt  gear  has  a  direct  acatory  connection with  at  least  one  of  the  wing  steering  muscles. That  would  change  the  timing or  activation  of  the  wing  steering  system. Which  would  change  conformation  of  the  wing  hinge, wing  motion,  aerodynamic  forces  in  the  fly  trajectory. This  is  something  that  I'm  going to  call  the  control  hypothesis. This  is  already  published.  I'm  going  to condense  a  lot  of  work  into  just  a  couple  of  slides, but  this  makes  predictions. The  first  is  that  the  tear  muscle should  be  under  visual  control. The  second  is  that  what  we  should  be  able  to  do  is just  present  visual  input  to  the  fly. And  see  that  the  mechansitory  feedback from  the  halt  hear  is  modulated. Importantly,  we're  not  going  to  rotate  the  fly, we're  going  to  have  the  fly  still. And  then  rotate  the  visual  world and  see  how  the  whole  tear  responds. Then  finally,  we  should  be  able  to  activate the  whole  tear  steering  muscles  and  see changes  in  the  activity  of  the  wing  steering  muscles. Okay,  This  is  where using  Soplo  Genetic  can  be  really,  really  helpful. Because  one  of  the  challenges  with  studying the  whole  tear  is  that  it's a  very  small  structure  and  it's beating  up  and  down  200  times  a  second. What  we  can  do,  something  I  did, was  take  advantage  of  the  fact  that when  motor  neurons  excite  muscle, you  get  this  big  burst  of  calcium. In  this  case,  what  I  could  do  is  express a  genetically  encoded  calcium  indicator, the  hall  tear  steering  muscles, and  then  using  an  epifluorescent  microscope, just  tear  muscle  activity  directly  through  the  cuticle. While  the  fly  is  tethered  and  present different  kinds  of  visual  motion and  there's  no  dissection  here. So  the  fly  is  tethered  and  is  doing  what  it  wants  and  I can  image  tear  muscle  activity. Okay,  I  just  want  to  show  you  a  quick  video. In  this  case  here  the  baszler  here, the  axillaryesI'm  imaging  the  left  left hal  tear  muscles  and reporting  the  left  wing  beat  amplitude. When  the  world  rotates  to  the  left, the  fly  is  going  to  follow  the  motion, left  wing  amplitude  decreases  and  you  can  see that  in  both  cases  muscle  activity  increases. All  right,  fantastic  L  muscles are  active  and  under  visual  control. The  next  thing,  next  part  of  this  is  looking  at, well,  if  that's  the  case, we  should  see  changes  in sensory  activity  from  the  holar  nerve. This  is  a  dissected  fly  nervous  system. Here's  the  brain,  the  cervical connective  and  the  ventral  nerve  cord, and  these  two  large  tracts  running  up to  the  brain  are  the  primary  afference  of  the  hole. They're  huge.  Importantly,  at  this  region  of  the  brain, the  subsophogeozone,  you  have  these  large  axon  terminals. Using  two  ph  laser  scanning  microscopy, what  we  can  do  is  again  express a  genetically  encoded  calcium  indicator. And  then  record  calcum  activity while  the  fly  is  tethered  and  flying. But  importantly  the  fly  is  rigidly  stuck. I'm  going  to  show  you  another  video where  in  this  case  this  is the  right  hale  axon  terminals and  looking  at  the  right  winged  ampitude. Again,  the  world's  moving  to  the  left. Now  winged  ampitude  increases  on  this  side. There  are  a  couple  of  things  I  want  to  point  out. First  is,  before  the  stimulus  comes  on, there's  a  baseline  level  of activity  which  is  consistent  with  the  idea  that the  halter  is  beating  up  and  down and  providing  rhythmic  feedback  to  the  halt. But  then  when  the  world appears  to  rotate  and  the  fly  steers, if  the  halt  here  were  merely  a  passive  gyroscopic  sensor, shouldn't  see  any  change  in  fluorescence  activity,  right? But  you  can  see  that  there's  an  increase in  activity  which  is  consistent  with  the  idea  that the  tear  muscles  in  some  way  are  motion  or  mechanics. And  I  also  did  some  experiments  on  doing the  same  experiment  on  the  wing  sensors  and  saw the  same  thing  That  gives  us  some  sense  that  what  is happening  is  Hal  tear  motion  and  hold  that  thought. I'll  get  back  to  that  in  a  little  bit. Then  the  last  thing  is  that  we should  be  able  to  activate  halter  muscles, record  from  the  wing  steering  system  and  see  a  change  in activity  using  what's  called  split  Gf, split  four  driver  lines  which  selectively label  basically  individual  cells  in  this  case. In  this  case  I'm  labeling these  two  hole  tear  steering  muscle  motor  Neons, you  can  optogenetically  activate  these  with  red  light. Flies  can't  see  red  light while  we're  recording  electrophysiologically from  the  wing  steering  muscles  and see  any  changes  in  activity. Let  me  give  you  an  example  of  this. In  this  case,  we're  recording from  a  muscle  that's  tonically  active. And  these  are  muscle  action  potentials overlaid  for  multiple  wing  strokes. And  you  can  see  that's  firing  really precisely  at  a  specific  time  in  the  stroke  cycle. Now  when  I  just  activate these  two  moons,  which  are  really, really  small,  we  get  this a  phase  advance  rights firing  earlier  in  the  wing  stroke  cycle. And  we  can  also  look  at  a  phase  of the  active  muscle  and  see  that  we  get  recruitment  right at  a  different  set  of  steering muscle  motor  neurons  and recording  again  from  that  same tonically  active  muscle  I  showed  you, we  get  a  phase  delay  just  by activating  four  motor  neurons, we  can  recapitulate  all  the  modes  of control  that  flies  have  over  the  wing  steering  system. But  importantly,  we're  not  looking  at  the, we're  not  activating  wing  muscle  motor  neurons, we're  activating  a  different  set  of  motor  neurons. Okay,  The  basic  idea is  that  the  halt  here  is  not  just  a  gyroscopic  sensor, it's  also  a  adjustable  timing  mechanisms where  visual  information  is  coming  through, sent  to  the  Holter  muscles. Which  is  changing  the  motion or  mechanics  of  the  halt  here. Changing  the  mechanic  sensory  feedback on  a  stroke  by  stroke  basis. And  then  changing  the  timing  or activation  of  the  wing  steering  system. Okay,  the  other  thing  that  we  think is  this  idea  of  the  halt  here, detecting  gyroscopic  forces. The  corols  force  may  be an  epiphenomenon  that  takes advantage  of  this  basic  reflex. All  right,  that  establishes for  us  that  the  halt  here is  a  multi  functional  sensory  structure. Now  what  I  want  to  do  is  think about  how  that  could  even  be  achieved. Most  of  what  I'm  going  to  talk  about  is work  by  postdoc,  in  my  lab. On  a  verb.  Okay,  I mentioned  that  tears  have  these  mechanic  sensors  called compatiform  sencilla  and  like  I  said  earlier, you  can  find  them  on  different  parts  of insects,  from  robber  fly. This  is  a  stick  insect  leg  and  this  is  a  moth  wing. One  thing  that's  really  important,  this  is  actually something  that  I  did  with  Simon  during  the  pandemic. Simon  came  to  me  with  this  interesting  idea of  we  have  the  sense  from using  band  Gaussian  noise  experiments  that  pan  forms and  exhibit  what  we call  derivative  pair  feature  detection. If  you  use  a  model  Hodgkin, Huxley  neuron,  you  get  the  same  dynamics  out. Which  gives  us  a  sense  that  fundamentally the  neurons  underlying  these  compan  forms and  silla  are  very  simple. And  really  what  matters  in  terms  of  how they  encode  information  is how  many  you  have  and  where  you  place  them. The  local  Bi  mechanics  matter  the  most. Okay,  now  think  more specifically  about  the  Hal  and  its  role  is  both. The  gyroscopic  sensor  transducing, the  halter  motor  commands. All  right.  I  showed  you  this  before. What  we  know  about  the  halt  here  is  that  you have  regions  of  sensors  that  have  distinct  morphologies. These  are  scanning  electron  micrographs from  different  families  of  flies. This  region  on  the  stalk,  you  can see  that  they  all  have  morphology. That  the  body  of  the  N  accordion  where  they're  fused. Roy  row  this  region  at  the  base  which is  the  business  end  of  a  gyroscopic  sensing, They  have  more  oval  shape. But  again  you  see  this  Rob  by  row  having  so many  of  them  gives  you  additional  sensitivity. But  you  can  also  see  that  the  morphology, you  can  see  how  the  morphology could  definitely  make  a  difference in  terms  of  what  information  is  relevant. To  these  regions.  And  you  can  take  it one  step  further  and  look  at  the  mechanics, transduction  channels  underlying  these  capa  forms. And  see  that  the  ones  along  the  stalk  are aligned  with  the  long  axis  of  the  halt  here. And  the  idea  for a  long  time  has  been  that  they're  detecting  in plane  bending  while  the  fly  is  flapping  the  halt  here. Whereas  these  ones  at  the  base, they're  aligned  off  axis  and maybe  more  prime  to  detect  shear  strains associated  with  either  gyroscopic  forces or  these  muscle  commands. Okay,  so  we  have some  hypotheses  about  how  these  might  work. But  again,  testing  this  is  difficult because  we're  talking  about  a  moving  structure. The  cell  bodies  are  on  the  halt  here,  right? How  are  we  going  to  test  this  out? Importantly,  this  is  a  thing  because  I  work  under  soofa. There  aren't  specific  driver  lines to  label  just  this  region  or  just  that  region. We  need  a  different  approach. Okay.  I  had  this  idea that  maybe  what  we  could  do  is  go  back  to this  method  of  imaging  directly  through  the  cuticle. Because  these  sensors  are  so superficial  and  there's  so  many  of  them  that you  might  be  able  to  use a  calcium  indicator  and  just  see  it  directly. This  is  what  Anna  does. She's  got  a  fly  tethered  here. She  can  track  wing  motion  using  a,  using  a  camera. We  also  have  a  camera  for  imaging  from  the  halt  here. Importantly,  the  fly  is  flying  in  this  image. It  looks  as  though  the  halt here  is  just  stuck  in  downstroke,  right? But  something  that  we're  also  doing  is  we're  tracking wing  motion  using  infrared  LD  and  phot  detector  pair. And  this  gives  us  a  real  time  measurement. And  what  we  can  do  is  prescribe  when  the  stroke  cycle, we  want  the  camera  shutter  to  be  open  so  that  we  only  see the  halt  in  downstroke can  from  the  halt  here  while  the  fly  is  flying. What  we  can  also  do  is  see  these  two  different  regions. We've  got  the  base  and we've  got  the  stalk  of  the  halt  here. We're  just  looking  at  the  dorsal  side. Okay,  so  if  we  do  that  we'll  be  fine  just  by doing  this  simple  experiments  that both  regions  are  slightly  to  my  surprise, but  in  some  ways  it  makes  lot  of  sense. They're  both  continuously  active  and  they  have  a  mean, a  baseline  level  fluorescence, and  they  fluctuate  about  that  level. In  many  ways  it  makes  good  sense because  they're  mechanic  sensors, they're  sensitive  to  nanometer  scale  deflections. And  this  thing  swinging  up  and  down  180  degrees, 200  times  a  second. Okay,  But  now  what  we  can  do  is  say, all  right,  we  can  image  from  the  halt  here, can  we  understand  how  it's  encoding visual  information  to  control  different  flight  behaviors? Okay.  In  this  case  what  I'm  going  to  show  you is  we're  going  to  present the  world  rotating  around  the  animal  in  different  axes. And  I'm  going  to  show  you  left  in start  wind  amplitude  and  the  change  of  fluorescence. Just  as  an  example  for  this  region  at  the  base. In  this  case  what  we've  got  up  top is  the  arrows  indicate  using  the  right  hand  rule, the  direction  of  rotation  around  the  fly. We  can  see  that  just  as  you  would  expect, the  F,  this  is  the  left  us,  right? So  you  can  see  that  the  fly  is  sensitive  to different  changes  in  visual  motion,  right? What  we  can  also  do  is  then these  dotted  lines  represent  when  the  stimulus  is  on. So  we  can  just  sum  those  and  just  get a  nice  tuning  curve  and also  put  it  in  the  polar  plane,  right? That's  exactly  what  you'd  expect. What  we  find  with  the  halter  sensory  fields is  that  they  are  also  tuned. And  they're  tuned  in  the  same  direction as,  as  wing  motion. Which  is  nice  because  the  work  I  showed  you  previously, looking  at  the  halt  here  at,  on terminals,  we  saw  something  similar. Right?  It's  nice  to  see  that  at  the  cell  bodies. Right,  we're  seeing  something  similar. And  looking  at  the  other  region  of  the  halt  here  again, we  see  that  they're  also  tuned. Okay,  Motion  is  directly  tuned. Now  we  know  that  the  halt  has a  direct  sitory  connection  to at  least  one  or  two  of  the  wing  steering  muscles. If  we  go  back  to  this  idea that  the  wing  steering  system is  functionally  stratified,  right, where  we  have  these  ticlyactive  muscles that  are  more  important  for relatively  slow  and  visually mediated  stabilization  maneuvers and  these  physically  active  muscles. I  told  you  that  visual  input  primarily modulates  these  muscles  and that  they're  linearly  recruited. Something  that  we  asked  is, is  there's  similar  functional  stratification in  the  hall  tier  system? What  we  did  is  we  took the  data  from  the  tuning experiments  and  then  we  just  sort  them agnostic  of  direction  based  on the  strength  of  the  response  from the  most  right  word  to  the  most  left  word  turns. When  we  do  that,  we  get  something  that  looks  like  this, where  each  row  is  a  different  trial, and  the  color  encodes  the  strength  of  the  turn. Then  what  we  can  do  is  sort  these  into deciles  and  plot  the  decile  means,  again,  color  coded. And  then  basically  during  the  stimulus being  on  how  strong  of  a  turn  we  get,  right? If  we  do  this  for  our  different  compare  form  fields, we  get  this  nice  and  surprising, more  linear  relationship  between the  strength  of  the  turn  and  fluorescence, which  gives  us  some  sense  that wide  field  visual  motion  is linearly  recruiting  halter  feedback. Okay,  that's  interesting. The  other  thing  to  keep  in  mind  is  that  flies  execute these  active  maneuvers  that  call  scads,  body  scads. The  halt  here  has  been  implicated in  stopping  these  scads  because  again, the  Hal  is  definitely  a  passive  gyroscopic  sensor. But  it's  unclear  how the  halt  could  be  involved  in  their  initiation. Right.  The  one  thing  that's  very  convenient  about scads  is  that  they're  not just  present  during  free  flight, they're  also  present  in  tethered flight  and  they're  ballistic. What  we  can  do  is  during  tethered  flight, look  for  these  kinds  of  behaviors and  then  see  how  the  halt here  is  or  is  not  active  during  these  kinds  of  maneuvers. Okay,  again,  this  is  basically  finding those  codes  and  then  sorting them  based  on  the  strength  of  the  turn. We  can  do  the  same  thing  with  the  fluorescence  record. Now  we  can  see,  again,  we  get this  smooth  control  over  wing  motion, which  is  exactly  what  you'd  expect. A  fly  can  turn  to  the  left,  to  the  right. But  now  what  we  see  is  that  in both  cases  we  get  linear  or  non  linear  re, get  non  linear  recruitment  in  both  cases. Right?  Also  importantly,  you  can  see  that the  halt  here  signal is  active  before  the  fly  actually  turns. Right?  Which  is  pretty surprising  to  me  and  very  exciting  to  me. Okay,  But  I  told  you  at the  beginning  that  fundamentally  the, the  sensors,  the  compan  form  cinsilla  are really  just  generic  Hodgkin  Hucks  neurons. So  how  can  we  get  this  linear, non  linear  relationship,  right? Well,  something  I  thought  about  as  well. This  is  very  reminiscent  of  what  we see  in  the  wing  steering  system. And  the  whole  tear  has  a  set  of  muscles. Maybe  there's  functional  gravification  in the  whole  tear  steering  system that  could  explain  what  we  see  here. Okay,  so  we  can  record from  the  whole  tear  muscles is  just  another  example  of  that. And  then  look  for  these  scads while  recording  from  the  whole  tear  muscles. Now  in  this  case  what  we  see  is  again if  I  look  at  left  right  wing  being  amplitude, we  get  smooth  control  over  wing  motion. But  now  looking  at  one  group  of  muscles, muscles,  we  get  this  linear  recruitment. We  also  get  non  linear  recruitment. To  put  a  point  on  it, we  have  muscles  that  are tonically  active  and  muscles  are  physically  active. Okay,  Now  what  this  allows  us  to  do  is  ask,  okay, we  have  functional  stratification  in  the  steering  system that  may  be  functionally  stratifying  the  actual  sensors. But  what  we  can  also  do  is  ask, is  there  is  there  a  way  that  we can  understand  how  these  muscles are  actually  controlling  halter  motion to  achieve  the  stratification? Okay,  now  I  want  to  think about  how  could  the fly  potentially  control  whole  tear  motion. One  potential  hypothesis  is  that  basically what's  happening  is  that  the  hal, tear  muscles  mimic  the  effect  of  the  corols  force. Okay,  again,  just  to  remind  you, if  we've  got  the  whole  tear  stroke  plane  and now  we  rotate  about  the  yaw  axis, we  have  some  angular  velocity  omega and  we  have  the  ****  velocity. The  cross  product  of  the  angular  velocity  with the  ****  velocity  is  going  to  be the  Coriolis  force  and  you'll  get  a  change  in  trajectory, in  this  case,  more  of  a  figure  eight  pattern. But  this  is  highly,  highly  exaggerated. Because  if  you  model, this  is  the  plane  motion.  Is  the  out  plane  motion. And  you  can  see  that  the  out  of  plane  displacement  is two  orders  of  magnitude  less  than  the  in  plane  motion. Right?  That  the  motion  that  you're  expecting that  we  could  potentially  measure is  on  the  order  of  one  degree. Which  you  can't  measure  that  very  well. Right?  But  another  alternative  is I  showed  you  that  wing  motion and  the  Who  tear  sensors  are  tuned  in  the  same  direction. What  I  didn't  tell  you  is  that  some  work  that  I  did previously  found  that  the  whole  tear  muscles are  also  tuned  to  visual  motion, but  in  the  opposite  direction. And  that  gives  us  the  idea  that,  well, if  the  whole  tear  sensors are  tuned  in  the  same  direction, and  we  know  from  recording  from the  wing  compare  forms  that, that  increase  in  wing  stroke  amplitude lead  to  increases  in  calcium  activity. This  indicates  that  potentially what's  happening  is  you're  getting  increase in  hall  tear  stroke  amplitude. Maybe  what  the  hall  tear  muscles  are  doing is  regulating  stroke  amplitude. Something  else  is  very  convenient about  finding  out  that  the  hall  tear  muscles, at  least  some  of  them  are  physically  active, is  that  they're  non  lineally  recruited, so  they're  more  of  a  switch. And  what  we  can  potentially  do  is  optionetically  activate these  muscles  and  see  how  that  might  change  tear  motion. Okay,  this  is  something  that  started  a  collaboration  with A  colleague  of  mine  at  Case  Western  Reserve and  her  student  Chris  Lea  has  been doing  these  experiments  where  she  optionically activates  the  same  owner  as  I  showed  you  before. That  lead  to  changes  in the  wing  steering  system,  a  phase  advance. But  now  what  we're  doing  is  just  tracking hall  tear  motion  with  a  high  speed  video  camera. If  we  do  this  we  see  a  decrease  in lear  stroke  amplitude.  This  is  a  Z  score. But  just  so  you  know,  this  is  on the  order  of  about  ten  degrees,  right? So  it's  a  pretty  significant  drop in  hall  tear  stroke  amplitude.  All  right. What  we  think  is  going  on  is  that  you've  got visual  information  sent  to the  Hal  Compan  forms  or  Compa  forms  or  companiforms. What  really  makes  a  difference  is  that  you've  got functionally  stratified  muscles  that  are regulating  stroke  amplitude  for stabilization  reflexes  or  active  maneuvers. And  that's  changing  the  encoding  of  the  hall  tear  itself. Okay,  So  now  we've  got  this  multifunctional  sensor. But  this  actually  presents a  problem  that  I  want  to  talk  about  now  which  is how  Hal  tear  feedback  essentially organized  and  this  is  worked  by  Grass  the  lab. Serene  An  okay. What  I  told  you  just a  couple  of  slides  ago  is  that companiform  field  activity  is linearly  and  non  recruited  and  that's  based on  the  functional  ratification of  the  whole  tear  steering  system. But  the  first  thing  I  showed  you  was  that the  companiforms  are,  are  continuously  active. Right?  And  I  also  showed you  this  anatomy  of  the  halt  to  your  nerve,  right? And  you  have  this  big  cable  of  information. How  is  that  information  preserved as  it  enters  a  central  nervous  system? Right?  That's  not  obvious. Okay.  That's  why  I want  to  explore  in  this  last  part  of  the  talk. Okay?  We  have  some  sense  of the  gross  projection  patterns  of  these  different  fields, these  two  fields  that  I've  been  talking  to  you  about. In  the  case  of  the  wing  steering  system  and the  whole  tears  relationship  to  the  wing  steering  system, we  know  that  it  provides information  to  one  of  these  muscles. One  muscle,  this  region  here  at  the  base provides  information  to  one  that helps  increase  wingstroke  amplitude. But  there  are  all  these  other  ways  that  flies. Control  wings,  stroke  motion, and  we  have  all  these  other  caps  fields  of  panis. We  don't  understand  if  there's any  organizational  logic underlying  the  projection  patterns. Okay.  One  of  the  nice  things  about  working  in Jop  is  that  there's  a  large  community  of  tools  available. My  colleague  John  Tthill  Way on  Lee  John's  at University  of  Washington  Ways  at  Harvard  Medical  School. They  got  together  to  collect  this  EM  data  set, The  female  adult  nerve  cord, where  it's  a  serial  EM, and  you  can  get  a  full  reconstruction  of, in  many  cases,  whatever  neuron  you want  of  the  nerve  cord,  not  the  brain. A  lot  of  people  are  doing  that  and I  have  colleagues  at  Princeton  that  have  been  doing  that. But  just  focusing  on  going  to  call  Fan, we're  going  to  reconstruct  the  whole  tear  nerve and  look  at  its  relationship  to  the  wing  steering  system. Okay?  All  right. We  can  reconstruct  the  whole  tear  nerve and  there  are  about  180  neurons  in the  hall  tear  nerve  just  to emphasize  how  important  the  hall  tear  is  to  the  fly. They're  about  14  inputs  per  neuron,  all  right? But  there  are  about  400  output synapses  per  neuron,  right? This  is  helpful,  being  counting. But  what  we  can  also  do  is  since we  have  the  location  of  these  synapses, that  can  actually  be  a  little  bit  more  powerful. In  the  case  what  Serena  I  came  over  with  this  project  of, can  we  think  about  how  whole  tear  information is  organized  and  we don't  have  the  location  of  the  whole  tear  sensors. Right?  But  serene  came  up  with  this  clever  idea. She  was  like,  well,  I  know  the  morphology  of each  neuron  and  I  have  knowledge  of  the  axon  terminals. Maybe  I  can  do  is  I  can  subtract the  backbone  of  the  neuron  and  then just  classify  each  neuron  based on  the  morphology  of  its  axon  terminals. And  then  compare,  paralyze, each  neurons  morphology  with one  another  and  see  if  any  kind  of  clusters  fall  out. And  when  she  does  that,  she  gets five  different  anatomical  clusters. Right?  Initially  this  was exciting  test  because  there  are  four  major  compan, form  fields  that  Cortonal  organ  that detect  stretch  in  the  halt  here  that's  five, get  five  different  regions  here. Maybe  there's  something  to  that.  But  I think  the  story  gets  a little  bit  more  interesting  actually. Okay,  One  thing  is  we  just  see that  these  bundle  together, this  is  all  done  computationally. Seeing  that  bundle  or  faticulate  together is  a  good  confirmation  that  these  are  real  clusters. But  the  other  thing  that  serene  does  now I  have  those  same  clusters  color  coded. We  have  knowledge  of the  direct  connections  with  the  wing  steering  muscles. Now  what  we  can  do  is  look  muscle  by  muscle,  How? If  there  are  any  patterns  that emerge  from  these  different  clusters. When  she  does  that,  it's  exactly  what  she  sees. Where  in  this  case  this  cluster  seems  to target  these  two  muscles that  both  control  wind  show  amplitude, you  get  this  other  clusters. In  this  case  what  I'm  showing  you is  each  column  represents a  different  Lear  neuron  and each  row  is  a  wing  string  muscle  motor  neuron. And  the  color  indicates  basically the  intensity  of  the  synapse  or  the  number  of  synapses. A  couple  of  things  emerge  here. First  is  that  we  don't  just  target one  anatomical  group  of  muscles, we  are  all  four  sclerit  groups. But  also  each  group is  targeting  different  kinds  of  muscles, functionally  distinct  muscles,  muscles  that  are either  linearly  or  non  linearly  recruited. Okay?  It  seems  like the  hall  is  controlling basically  all  aspects  of  wing  motion, but  also  controlling  different, both  linearly  and  non  linearly  recruited  muscles. Okay?  The  other  thing  we  can  do  is  focus  in on  these  three  morphological  subtypes  and  then  look  at, look  at  the  words  are  escaping  me. Looking  at  a  special  inter, neuron  that  gets  information  from these  primary  reference  and then  directs  it  across  the  midline. Then  I'm  going  to  call win  conflateral  halter,  interneurons. Okay,  sorry  about  that.  And  actually in  this  case  it's  unique. The  reason  I  would  like  to  point  these out  is  like  in  this  example, again  going  back  to  these  two  muscles that  help  increase  wingtroke  amplitude. They  provide  input  that  crosses  the  midline to  control  a  steering  muscle, that  helps  decrease  wingroke  amplitude, the  antagonist,  right? And  this  one  also  helps decrease  wing  stroke  amplitude.  Okay,  great. But  the  thing  that's missing  is  that  this  is  ongoing  work, is  that  we  don't  have  the  antomical  origins of  these  clusters,  right? Because  we're  not  certain  if  the morphological  groups  I'm  showing  you originate  from  individual  regions  or  if  they're  a  mix. Right?  Because  that  would  give us  a  sense  of  if  there's  really a  computational  map  we're  using. Another  method  called  x ray  nano  hoolatomography  gives  us  EM  level  resolution. And  we  prepare  a  dissected  VNC with  the  halt  here  attached  like  you  would  for  EM, but  instead  of  slicing  it  now  we  case it  in  resin  and  ship  it  to  some  collaborators, a  pack  U  at the  European  Synchrotron  facility  in  Grenoble. Now  what  we  can  do  is  get  a  nice reconstruction  of  the  Hal  here, but  also  get  reconstruction of  the  inside  of  the  hal  here. And  then  scan  along  the  hall  to  your  nerve. We're  working  on  reconstructing  this  right  now, but  something  else  that  serine  can  do  is,  again, go  back  to  drosophogenetics  and  say,  well, can  I  use  genetic  tools to  basically  recreate  an  anatomical  cluster? This  is  an  example  where  she's  done  this for  this  cluster  that  labels, that  targets  these  two  neurons. Where  she  can  come  up  with  a  driver  that seems  to  look  to  mimic  the  morphology. But  the  thing  that's  also  powerful  about  this is  that  we  can  also  dissect  away  the  hall  tire, get  a  sense  of  locations  on  the  hall  tie. How  that  pair,  these  different  methods will  allow  us  to  get  a  sense  of  how the  information  is  organized  from  the  periphery  into the  central  nervous  system  we  think  is  going  on. It  seems  like  we  think that  what's  happening  is that  hall  tear  information  is  being  re, routed  into  a  complicated  map for  the  wing  steering  system, but  we're  still  working  on  that. But  going  back  to  the  major  question, some  of  the  major  takeaways, I  think  first  we  see  that  with  the  Halter  motor  system, you've  got  sub  millisecond  control  of the  steering  system  that  enables  high  dynamic  range via  functional  segregation  of  the  motor  system  that functionally  segregates  the  Compat  forms. Then  we  think  it's  happening  that  you've  got these  local  computational  maps  at  motor  targets. Finally,  the  thing  I  just  want  to  leave  you  with, I  brought  this  up  early  on  that the  halt  here  is  multifunctional  sensory  structure. But  another  thing  to  keep  in  mind  is  that because  the  halt  here  is  evolved  from  the  hind  wing, this  control  loop  idea  is  likely  based  on a  conserved  hind  wing  circuit  that  you find  across  flying  insects. That  I  want  to  thank the  lab  funding  collaborators like  Simon  and  take  any  questions  that  you've  got. Thank  you,  Brad.  We're  going  to  ask  that the  first  question  come  from  a  postdoc  or  a  student. We're  quite  patient,  so  we'll  sit  here  in awkward  silence  until  someone  asks  a  question. Great. But  you  may  be  in  the  beginning. You  need  both. This  is  an  interesting  thing  to  detect  body  rotations. You  may  only  need  one, especially  given  the  halter  in  neuron  is  showed you  in  terms  of  do  you  need  both  to  fly  though? I  don't  know  if  anyone's  actually  tried  that  experiment. It's  a  very  easy  thing  to  do. But  this  also  brings  up  another  point, which  is  thinking  about the  you  have  the  ipsilateral  projections and  you  have  the  contra  projections. Is  very  similar  to  the  organization of  the  vertebrate  vesibulocular  reflex. Ipsilateral  control  projections that  help  control  posture. It  just  came  to  mind,  we're  not  certain  but  yeah.  Sorry. Yeah,  sorry.  Excuse  me. The  question  I  had  about  the mechanic  sensors  on  the  hall  theories, you  showed  a  diagram  where  it  was  connected  to  dead  dry, the  force  that  activates  that  sensory  transduction. Is  that  centripal  force or  what's  the  force  that's  like  opening? Oh  yeah.  So  in  the  case  of  the  mechanic  trans, the  mechanic  transduction  channels, it's  what  you  have  is  in  a  hero  to  go  back  to  that. So  many,  in  many  cases  here. Yeah,  so  let's  go  here. What's  happening  on  the  leg or  on  an  antenna,  or  on  the  wing? They're  bending  and  then  this  is  rising  or  falling. In  the  case  of  the  halt  here,  what  ends  up  happening  is the  tip  path  trajectory  changes  which  changes, like  you  get  these  shear  strains  that  develop and  that's  going  to  squish  and  stretch  the  exoskeleton, which  is  going  to  have  the  same  effect. In  this  case  you've  got  the  Correos  force  which  is changing  the  apparent  motion  of  the  halt  here. That's  going  to  cause  that  dome  to rise  and  fall  and  then  open  the  ion  channels. Yes.  Are  there  any  mechano  receptors  in the  muscles  in  insects? No.  But  they  have  Cortonal  organs  that  detect  stretch. And  my  colleague  John  Tuttle  works  on  these  in  legs. Hal  here  also  has  one, but  he's  been  doing  a  lot  of  work  and there's  a  lot  of  work  on  locusts  in  those. What  he's  finding  is  that  that's actually  a  case  also  where  bio  mechanics  makes a  big  difference  because  what  he's  finding  is  that there  are  different  functional  groups, some  that  respond  to  the  full  leg  position  and  changes  in leg  position  as  someone who  knows  him  and  has  read  the  work. It  seemed  like  what  they  thought  was  happening  was that  because  they  were  genetically  identifiable  groups, that  there  are  differences in  the  underlying  like  physiology  of  the  neurons. But  really  what  they  found  is that  each  of  those  subgroups attaches  by  a  tendon  to this  structure  in  the  leg  called  the  archulum. They  have  different  attachment  points  as  the  leg  bend, because  they  have  different  attachment  points  that pulls  on  them  in  different  ways. The  biomechanics  in  that  case also  determines  how  they're  encoding  information because  they  also  did single  cell  RNA  sequencing and  they  found  there  are  basically no  real  differences  among  those  neurons underlies  this  point  that  at  least  for  insects, in  these  cases,  the  underlying  sensory  neurons are  just  very  replaceable, which  is  very  convenient because  then  you  can  have  many  more  of them  and  get  additional  sensitivity  and then  you're  just  changing  how  you  hook  them  up. That's  a  lot  easier  than  making new  neurons  with  different  sensitivities. Yeah,  thank  you.  It's  fascinating. You  talked  about  how the  whole  activity  is  correlated  with future  steering  events  and we  talked  about  it  like  cycle  feedback. But  in  the  movie  it looks  like  the  whole  tears  are  going  out. Yes.  That  I  Could  it  be a  within  stroke  sensory  prediction? Because  they're  leading  the  lighting. Yeah.  It's  Yes.  It's  definitely  possible. Because  what  the  Hal  tear  muscles  are  doing is  trimming  Hal  tear  motion  very  subtly. Wing  stroke  to  wing  stroke. And  one  thing  that's  difficult  for  us  is  that calcium  imaging  is  powerful  because  we  have access  to  systems  that  we  haven't  had  access  to. But  what  you'd  really like  to  do  is  be  able  to  record  from these  muscles  electrophysiologically  and get  a  sense  of  what's  happening. Each  wingtroke's  something  that  we  just  can't  do and  maybe  voltage  sensus will  get  good  enough  that  we  can  do  it  in  flies  one  day. I  mean,  what's  the  current  explanation  for  them  going, oh,  oh,  that's  mechanical. The  idea  there  in  terms  of the  phase  relationship  is  that  that's basically  the  mechanics of  the  flora  because not  all  flies  have  this  180  degree  relationship. The  phase  relationship  for a  given  fly  is  important  for  it. But  it's  not  always  180  degrees  because  there  are some  flies  that  where  it's  like 90  degrees,  some  where  zero, and  that's  determined  by  the  mechanics  of  the  florax, whatever  the  phase  relationship  is  for  that  fly. It  is  really  important.  Yeah,  sorry. Well,  yeah. Noodle  compensated  for  the  lack of  the  mechanical  connection  between the  wing  and  then  the  noodle  control from  the  heart  compensated for  the  damage  of  the  mechanical. Oh,  you're  saying  like  if  you  were  to  disrupt  that? No,  I  don't  think  it  can  because  someone  did these  experiments  a  few  years  ago  that disrupts  how  the  fly  flap  saying. Yeah.  Yes. Can  I  take  away  a  simple  message  that  the  maneuverability comes  from  the  phasic  muscles  and the  stability  comes  from  the  muscles. Yeah,  that  I'm  going to  follow  up  on  that  because  I'm  curious  if the  segregation  as  you're suggesting  that  it  comes  somewhat  from a  connectivity  mapping  of the  hall  tear  fields  to  the  muscles  and the  different  combinatorial  combinations. Right,  computationally, the  difference  between  a  tonic  muscle  and a  phasic  muscle  could  also  be how  much  persistent  excitatory  current  like. If  you  think  of  them  as  simple  integrated  fire  muscles, neurons  or  move  under  neurons  or  muscles. Then  the  tonic  ones  just  need a  small  shift  to  kick  the  phase. But  the  phasic  muscles  need  to  be  like  elevated, above  threshold  and  burst. Could  you  do  it  dynamically  that  way, or  is  that  not  sufficient  to  do  the  segregation? Yeah,  that's  an  interesting  question. Another  way  you  could  think  about this  is  that  you  could  imagine, since  there  are  so  many  descending  in neurons  that  get  visual  information, I  know  that  people  are  working  on  this  and they're  finding  that  there's also  functional  segregation  there. Where  there  are  inter  neurons  that  are  relatively slow  for  stability  and  some  that  come  on  and  bursts. But  someone  who  was  a  grad  student  in  the  lab  as a  postdoc  and  did  some  experiments  where  she optionatically  was  activating  those  in  neurons. You  could  kick  the  system,  you  could get  phase  shift,  you  get  activation. But  particularly  with  those  phase  shifts, they  were  nowhere  near  as strong  as  activating  the  hale  steering  muscles. There's  a  way  for  visual  information  to  aid  in  that, adjusting  the  system,  that  has  to  be  the  case. But  this  gets  back  to  the  idea  that  the, the  control  loop  is  a  conserve  hind  wing  reflex, which  is  that  in  moths  and  in  locust, we  know  from  previous  work  that  the  hind  wing provides  really  strong  information  to  the  fore  wing. What  the  hal  may  be doing  is  really  you  have  the  separation  of labor  between  aerodynamic  force production  and  then  just  timing. But  this  timing  thing  has a  really  strong  influence on  the  wing  steering  system  that  you just  can't  eliminate  because  no reason  for  flies  to  need  a  whole  tear. Other  than  these  constraints  of  evolution. Yeah.  Has  there  been  any  experiments to  genetically  knock  out  the  ano  sensors  on the  hall  tears  to  know  if  it's  like  something  other  than that  someone's  actually  doing  this  as  we  speak. Not  in  my  lab,  but  they're finding  that  just  basically  as  what you'd  expect  as  you  remove more  and  more  of  these  sensors  and  you  rotate  the  fly, they  become  less  and  less  sensitive  to the  Correles  force  question. You  mentioned  that  the  12  wing  muscle, they're  all  controlled  by  the  two  motor  neuron. Do  you  mean  that  all  the  12  are  controlled  by the  same  neuron  or  that  each  muscle is  controlled  by  its  own  two? They're  all  controlled  by,  they  each have  a  single  motor  neuron  there. The  names  of  them  are  each  muscle has  a  different  name  and  there's associated  motor  neuron  for  that  muscle. Yes.  Sorry.  We  know if  we  can  do  like  knockout of  some  of  the  learning  mechanisms. Would  that  tell  us  if  this is  the  Hal  tears  like  compensatory  action is  like  a  learned  behavior  or  do  we  know  if it's  like  a  inbuilt,  like  within  the  gene? I  think  the  idea  is  that  the  halt, one  thing  that's  important  about  the  holes, even  though  the  steering  system  gets  visual  input, it's  pretty  self  contained  reflex, it's  pretty  isolated  from the  learning  mechanisms  that  flies  have. Yeah.  Yeah.  Sure.  Okay.  And  we're at  about  12:15  so  we're  going  to wrap  up  and  thank  Brad  again for  an  excellent  topic.  We  just