2/ In the example image above, a binary signal is encoded with either a rate or synchrony code. The rate code uses high rate = 1, low rate = 0. In contrast, in the synchrony code the cells have a constant rate-of-fire, but high synch = 1, low synch = 0.
-
-
Prikaži ovu nit
-
3/ We hypothesized that PV+ and SST+ interneurons would be differentially sensitive to these two codes, due to their intrinsic and synaptic properties. For example, the short-term depressing synapses of PV+ cells may make them more sensitive to synchrony.
Prikaži ovu nit -
4/ How did we test this? We infected layer 2/3 barrel cortex excitatory cells with ChR2-mCherry, in GFP mouse lines for identifying PV+ and SST+ cells (GAD67 and GinGFP lines, respectively).pic.twitter.com/x8avpO5Jwm
Prikaži ovu nit -
5/ We prepped ex vivo slices and recorded from GFP+ cells (and some GFP-) with WC patch clamp. We then used a digital multimirror device with some custom software (courtesy of
@mightexsystems) in order to selectively activate ten specific presynaptic inputs to the patched cell.pic.twitter.com/A3mJcS66kD
Prikaži ovu nit -
6/ With the patterned optical illumination, we were able to optically excite specific patterns of the presynaptic inputs. We used this to encode (literally) a random 1-bit signal using either the rate or synchrony of the presynaptic inputs.pic.twitter.com/HuE8BSgOnI
Prikaži ovu nit -
7/ We then measured the mutual information (MI) between the recorded neurons sub-threshold and supra-threshold responses and the 1-bit signal, in order to determine their sensitivity to the two different codes.
Prikaži ovu nit -
8/ Small tangent: we cannot say that the recorded neurons were necessarily "encoding" the 1-bit signal, as we don't know their downstream impact. But, we can say *we* encoded the signal, because we know the recorded neurons were receiving the info (i.e. the causation is known).
Prikaži ovu nit -
9/ So what did we find? Broadly, all cell types had some similar responses to the synchrony code: high synchrony led to low variance supra-threshold responses, while low synchrony led to a variable mixture of sub-threshold responses and spiking.pic.twitter.com/wLIRxbmq5F
Prikaži ovu nit -
10/ Likewise, there were some broad similarities in the responses to the rate codes, with all cell types showing more variable supra-threshold and sub-threshold responses, corresponding to the high vs. low conditions, respectively.pic.twitter.com/SyP0iEHD3f
Prikaži ovu nit -
11/ But, the MI analyses showed interesting cell type differences! PV+ cells carried more info in their spiking responses to the synchrony code in the first 5 ms after optical activation than GFP- or SST+ cells. Both interneurons carried more info in their later responses.pic.twitter.com/wtYNrna3Vg
Prikaži ovu nit -
12/ In contrast, for the rate code, the PV+ cells carried less info in their membrane potential than the other cell types. As well, for spiking responses, the SST+ cells carried more info than either GFP- or PV+ cells in their delayed responses.pic.twitter.com/SqM1vwkD4W
Prikaži ovu nit -
13/ So what does this all mean? It suggests that PV+ cells could rapidly communicate information carried by the synchrony of spikes in excitatory cells, while SST+ cells would more slowly communicate info carried by the rate-of-fire of excitatory cells.
Prikaži ovu nit -
14/ Of course, it's important to note the limitations here, two major: 1) It's not in vivo, so grain of salt. 2) We did not inhibit other cells, so our optical activation drove polysynaptic responses. This makes the late (> 5 ms) data potentially a mix of rate and synchrony.
Prikaži ovu nit -
15/ Nonetheless, the data hints at very interesting potential divisions of labour in coding in layer 2/3. This is particularly interesting for me from a learning perspective: perhaps synchrony communicates sensory signals and rates communicate learning signals?
Prikaži ovu nit -
16/ Anyway, we hope that this study will inspire more work. It would be great if some of these issues could be explored in vivo (which is beyond our capabilities). Also, we hope to see it motivate computational models, which should incorporate more cell type data IMO!
Prikaži ovu nit -
17/ Finally, I just want to say that this paper and the beautiful work in it was done by
@tranarama, with support from@lukeyuriprince and others. It was the result of a collaboration between my lab,@KohlLab and the Kwag Lab, courtesy of an@HFSP Young Investigator Grant.Prikaži ovu nit -
Fin/ Also, this
#tweeprint took longer than intended, but it's also special: This paper is one of only two pieces of experimental work that my lab will publish, as I decided to shut down my wet lab to focus on computational work from now on. So this is some limited edition sh*t.Prikaži ovu nit -
Oops, and here is the actual paper!
https://www.biorxiv.org/content/10.1101/671248v3 …Prikaži ovu nit
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.