Neural and Behavioral Mechanisms of Interval Timing in the Striatum
Read:: - [ ] Mello (2016) - Neural and Behavioral Mechanisms of Interval Timing in the Striatum 🛫2023-09-11 !!2 rd citation todoist Print::  ❌ Zotero Link:: Zotero Files:: attachment Reading Note:: Web Rip:: url:: https://run.unl.pt/handle/10362/43665
TABLE without id
file.link as "Related Files",
title as "Title",
type as "type"
FROM "" AND -"ZZ. planning"
WHERE citekey = "melloNeuralBehavioralMechanisms2016"
SORT file.cday DESCAbstract
To guide behavior and learn from its consequences, the brain must represent time over many scales. Yet, the neural signals used to encode time in the seconds to minute range are not known. The striatum is the major input area of the basal ganglia; it plays important roles in learning, motor function and normal timing behavior in the range of seconds to minutes. We investigated how striatal population activity might encode time. To do so, we recorded the electrical activity from striatal neurons in rats performing the serial fixed interval task, a dynamic version of the fixed Interval schedule of reinforcement. The animals performed in conformity with proportional timing, but did not strictly conform to scalar timing predictions, which might reflect a parallel strategy to optimize the adaptation to changes in temporal contingencies and consequently to improve reward rate over the session. Regarding the neural activity, we found that neurons fired at delays spanning tens of seconds and that this pattern of responding reflected the interaction between time and the animals’ ongoing sensorimotor state. Surprisingly, cells rescaled responses in time when intervals changed, indicating that striatal populations encoded relative time. Moreover, time estimates decoded from activity predicted trial-bytrial timing behavior as animals adjusted to new intervals, and disrupting striatal function with local infusion of muscimol led to a decrease in timing performance. Because of practical limitations in testing for sufficiency a biological system, we ran a simple simulation of the task; we have shown that neural responses similar to those we observe are conceptually sufficient to produce temporally adaptive behavior. Furthermore, we attempted to explain temporal processes on the basis of ongoing behavior by decoding temporal estimates from high-speed videos of the animals performing the task; we could not explain the temporal report solely on basis of ongoing behavior. These results suggest that striatal activity forms a scalable population firing rate code for time, providing timing signals that animals use to guide their actions.
Quick Reference
Top Notes
Tasks
Topics
The PRP is the duration of the interval between the acquisition of the reinforcer and the press start time (PST; i.e., the first response) to produce the subsequent reinforcer, and because it is sensitive to the length of the estimated interval, it is the standard metric of timing performance. It has two important features relevant to the interval timing studies: the sensitivity of mean accuracy to the FI and the scalar variance (scalar timing). tp
Serial Fixed Interval Timing (SFI) task is a variation of FI task used in our lab to sample timing performance from a broad range of intervals. In the SFI task the reward became available at t seconds after the previous reward, provided that the animal has responded. Each reward marked the end of one trial and the start of the next one. t varied in a block-wise fashion over intervals from 12 seconds to 1 minute. The performance in this task was consistent with previous results in dynamic versions of FI task [34, 35, 43, 44, 45, 46] in three aspects. Firstly, average PST is a function of the FI. Secondly, the response rate is proportional to the reward rate. Finally, rodents seem to adopt strategies that take into account for the whole distribution of FIs in the session. So that, PST was relatively later in short FI trials than it was within trials with long FIs. Because they adapted quickly to the new interval, taking less than 5 trials, this violation to the scalar variance property could reflect a strategy to facilitate exploration of intervals. The SFI task offers the possibility to analyze steady and changing timing conditions, while providing statistical power to infer parametric relationships between temporal demands, behavior and neural activity tp
the scalar expectancy theory (SET) model (Figure 1.1). SET assumes: an internal poisson-variable pacemaker that generates pulses, an accumulator, a reference memory, a switch and a comparator. When a time marker (cue or reward) is received, the switch allows the pulses to be stored in the accumulator. These pulses were accumulated until a short time after the reinforcement. The rate by which these pulses were generated depended on many other psychological or behavioral variables, such: as arousal, reinforcement magnitude, attention and mood [49]. By the time of the reinforcement, the value stored in the accumulator is transferred to the reference memory and the accumulator is reset to zero. tp
I have a feeling this may also be called a pacemaker accumulator, but this also clarifies why they think these episodes are encoded in the striatum
, sequential-state models characterize orderly transitions between different states which can be used to encode time [51,52,53]. tp
The behavior theory of timing (BeT) formulated by Killeen and Fetterman [54] and the learning-to-time (LeT) formulated by Machado inspired by BeT [30] are the most prominent of the sequential state models. tp
The oscillation-based model uses a library of oscillatory pacemaker neurons, which could be independently entrained in different rhythms, to encode a temporal waveform by forming its Fourier series. Torras [cited in 57] said that this combination could be done either by choosing pacemakers with appropriate oscillation periods or through plastic changes to the period of oscillation of each cell. The beat-frequency model (BF; Figure 1.3A) and its updated and more biologically plausible version, the striatal beat-frequency model (SBF; Figure 1.3B), uses “beats” (i.e., frequency at which cells spike tp
simultaneously) between pairs or groups of oscillatory cells to store time intervals. After resetting the oscillations with a synchronizing event, a specific time can be encoded by selectively weighting the activity of oscillatory cells that are currently active at the time criterion. This process is equivalent to multiplication (e.g., 3 Hz and 5 Hz will first synchronize at 15 Hz), thus providing an efficient process to encode long intervals with neuronal mechanisms which operate in much shorter timescale. tp
Organization of the basal ganglia tp
This whole section is probably something I will need to reread at some point but atm I’m getting nothing out of itc
Extracted Annotations and Comments
Page 18
identifying temporal regularities is extremely important for adaptive behavior. Indeed, to guide their behavior, animals operate with temporal information from different orders of magnitude, from microseconds to years.
Page 18
interval timing (i.e. the ability to perceive, estimate and discriminate intervals between events in the range of seconds to minutes to hours) has been identified in organisms as diverse as insects [7], birds [8], fish [9], rat pups [10] and adult rodents [11], primates [12], human infants [13] and adults [14]. Interval timing is critical for important adaptive behaviors. In foraging, animals use temporal estimates to estimate how much reward per time (i.e., the rate of return) any given behavioral strategy or area can provide [15,16]
Page 19
it has been suggested that interval timing requires rudimentary comparisons and estimations of quantities. These operations could be the basis for high cognitive faculties, such as arithmetics [20]. Moreover, interval timing is not an isolated faculty. It interacts directly with processes such as attention [21,22], memory [23], reward expectation [24] and arousal [25], so that variations in these processes cause temporal delusions. Therefore, Interval timing is not only a primitive ability that is useful to detect relevant patterns from the everchanging environment and generate anticipatory behaviors. But might underlie most, if not all, high cognitive functions of the human brain. By understanding the brain implements interval timing, we can gain insight into the processes supported by it, and perhaps identify unifying principles of how the nervous systems across species organize information about the environment and behavior.
Page 19
In contrast to other sensorial modalities (e.g., visual, tactile, etc), timing has no sensorial organ
Maybe we can call this “Building the Electric Time Organ”
Page 19
Hence, the sense of time either emerges from an internal clock mechanism, which generates a trackable time varying signal, or alternatively, arises from learning the temporal statistics of change in sensory and/or motor signals, which vary naturally with time
Page 19
The Basal Ganglia (BG), and especially the striatum, are necessary for time estimation in the supra-second range [26].
BG
Page 20
the means by which the BG might perform temporal computations remains elusive.
Page 20
Psychophysical studies of interval timing Much of our knowledge about interval timing is derived from psychophysical studies. These studies are based on retrospective and prospective timing methodologies to collect time duration judgments from a subject [33].
Page 20
studies of interval timing
OK so we are getting a bunch of different types of interval timing tasks
Page 20
the prospective study of time durations is based on the estimation, production, and reproduction of time intervals [34]. Estimation protocols require that a subject observes an interval and reports orally how much time has elapsed. Production protocols, on contrary, inform the subject about a temporal constraints (usually an interval between actions) he/she will have in order to perform an action. Usually a symbolic cue is associated with a particular interval, a spoken communication (e.g., “five seconds”). Because estimation and production protocols require some verbal interaction, most of timing research that uses animal models employ the reproduction protocols. In this protocol the subject is presented to one interval with a given duration criterion, then the subject has to reproduce this interval. Animals are usually deprived of either food or water, so that they are motivated to perform an action (e.g., pressing a lever, pushing a button) in a programmed schedule of time in order to receive a reward (i.e., drop of water, food pellet).
Page 20
The most often studied interval timing schedule is perhaps the fixed interval (FI) schedule of reinforcement and its variations. During a FI schedule, the behavior is reinforced for the first response (e.g., press of a lever) made after elapse of a pre-determined interval since the previous reinforcement. When the reinforcer is delivered, the cycle restarts
Page 20
First, the subjects consume the reward (as explained, the reward !1
Page 21
marks the end of the previous and the beginning of the subsequent FI). Secondly, the subjects generally engage in grooming or exploratory behaviors. Thirdly, the animals gradually orient their position and actions towards the response site (e.g., lever). Finally, despite the absence of any external time cues, the animals start to respond after a fixed proportion of the interval has elapsed. The responses under FI schedule generally are manifested in two characteristic patterns. The first one is the scallop performance. This pattern describes an increase in the frequency of responses as the end of the FI approaches. The second pattern, and more frequently observed in over-trained subjects, is the break-and-run. In this pattern, response frequency is kept at a fixed rate from the moment responding starts to the moment of the reinforcement. This fixed rate of response varies together with the reward rate (i.e., magnitude of reward per time in the FI; [36]), and the post reinforcement pause (PRP) of responding is proportional to the length of the interval.
Page 21
The mean accuracy sensitivity describes the observed phenomena that PSTs tend to occur in average around one half [35] to two thirds of the length of FI. The scalar variance feature characterizes that the dispersion of PST is a linear function of the average PST. Consequently, it also predicts that the coefficient of variation (standard deviation of PST divided by the average PST) is constant across all FIs. This latter feature is considered to be a manifestation of Weber-Fechner’s Law in the time domain, also known as Scalar Timing [37]. The Weber-Fechner’s law is obeyed by many sensory modalities [38, 39, 40, 41]. It states that the threshold to detect changes in magnitudes of stimuli (i.e., the just noticeable difference) is proportional to the magnitude. In other words, the just noticeable difference is a constant ratio between the measured magnitudes, for all magnitudes. Scalar timing affects behavior and neuronal activation by making them increasingly less precise as the timed interval lengthens [42].
Page 22
in timing tasks and explain the mechanisms of interval timing [Franois cited in 47, 48]. These models diverge in how well and how generally they can predict temporal performance, and what are the underlying mechanisms supporting interval timing. Broadly, these models can be grouped into at least three categories that vary in assumptions and explanatory power, !2
Page 23
namely: information-processing models, beat-frequency models and sequential-state models
Page 23
The perceived duration was a monotonic function of the total number of pulses transferred into the accumulator. The behavioral response is dependent on the ratio between the values stored in the accumulator and reference memory. For instance, when the difference falls below a threshold (which may also vary) in a FI task, responding at a steady rate begins. This feature explains that steady!22 Pacemaker generate pulses Switch start/stop accumulation process Accumulator accumulate pulses over time Memory store values passed by the accumulator Theshold sets a criteria for action to be taken Comparator (memory - accumulator) / memory Stimulus Response Figure 1.1 | Schematics of the scalar expectancy theory model. Components of the information processing model for time: pacemaker, accumulator, reference memory, switch and comparator. The pacemaker generates pulses through a Poisson-variable process. Environmental stimuli change the switch state, allowing the pulses to go into the accumulator. If the accumulator has a value stored during the moment of the stimulus, this value is passed to the memory and the accumulator is reset to zero. Finally the ratio of the difference between the values stored in the accumulator and the memory is compared with an established threshold to generate a behavioral response
Page 23
Fig 1.1
Page 24
state measures of time discrimination, such as wait time (i.e., PST, break point) on fixed interval schedules or peak-rate time (on peak procedure), are proportional to the to-be-timed interval (i.e., proportional accuracy sensitivity to the FI; [35]). Because SET posits that the error generated during the accumulation of pulses is proportional to the duration criterion, it presents an explanation for scalar variance sensitivity to the interval. More importantly, SET incorporates two features that have been supported by experimental data. Firstly, the current time estimate (encoding) and the memory for times reinforced in the past (decoding) follow independent laws [27]; and secondly, the behavior is driven by some sort of comparison between current and remembered time of reinforcement [29]
Page 25
These models are based on the empirical observation that sequential chains of behaviors emerge in tasks where reward delivery is contingent on passage of time (e.g., FI, SFI; Figure 1.2A). For instance, in a FI task, behavior would transit from consummatory, to post-consummatory, to exploration, to reorientation to the source of reinforcer and finally to the reinforced behavior across the interval.
Page 25
In BeT, each behavior is associated with a distinct underlying state. Transitions between states occur probabilistically driven by a poisson-variable pacemaker. The speed of this pacemaker depends on the rate of reinforcement, so that increases in reinforcement rate lead to an increase of the speed. The successive underlying states take on the role of a clock process.
Page 25
Consequently, to perform at a temporal criteria, subjects would learn to use their temporally organized behavioral states as discriminative stimuli. Thus, instead of reading an internal clock, subjects are assumed to use their current sensorimotor states to tell time.
Page 25
LeT extends BeT by positing that each underlying state is associated with an operant response, and that association strength varies through means of differential reinforcement in the context they were learned (Figure 1.2B).
Page 25
Therefore, the strength of the operant response at a given moment is the result of the combination between the predominantly active state at that moment, and how strong is the association between this state and the response.
Page 25
Two major distinctions between information-processing and sequentialstate models may argue for the broader explanatory model of the latter models. Firstly, in the former, the decision to respond is made only after the target time interval has elapsed, while in BeT it is done in anticipation to that time interval [53].
Page 25
. In Information-processing models the memory stores are independent, and because of that this type models have no mechanism to accommodate
I’m probably going to have to seek out more about BET and LET elsewhere
Page 26
contextual effects. In contrast LeT is sensitive to the errors that occur during the learning of the two discriminations; these errors weaken the connection between the behavioral states and the associated operant responses. Therefore, these errors would bias green keys to be perceived as long and blue as short, regardless the fact that both cues relate to the exact same interval.
I don’t get what they are talking about here with color keys. It probably has something to do with figure 1.2-
Lol nope. no idea what they are referring to
Page 27
The SBF posits that loops involving the cortex (CTX), the basal ganglia and the thalamus implement these mechanisms. More specifically, the striatum would act as a coincidence detector; the DA signals would synchronize cortical and thalamic oscillations at every meaningful event (e.g., reward), hyper-polarizing the striatum membrane, and thereby resetting the integration mechanisms. DA signals would also serve as reinforcement/teaching cues, strengthening the cortico-striatal representation of a particular duration criterion. Once synchronized, neurons oscillate at their inherent periods, allowing the patterns of activity to become meaningful. Striatal spiny neurons fire when a previously reinforced pattern of input is detected, consequently informing that the time criterion was reached. The striatum can entrain itself in the current oscillatory inputs through the striato-thalamic-cortical loop, allowing for alterations of time perception.
Page 27
In our current understanding, interval timing is a complex and primitive function of the brain which engages multiple areas of the brain depending on environmental and behavioral demands. Data from functional magnetic resonance imaging (fMRI) show that multiple areas have time dependent activity which is also affected by task and context, suggesting that interval timing is a distributed and complex process in the brain [62]. But, not all of these areas are required for, or modulate, timing performance equally.
Page 28
As we saw, some models rely on the assumption that the cortex provides the temporal basis for time estimation. But timing experiments done in decorticated animals [64] call this hypothesis into question. These experiments showed that decorticated animals are still able to perform in interval timing tasks.
Page 28
Many studies that attempted to affect interval timing through means of cerebellar lesions have failed [69]. Nonetheless, data from stroke patients [70] with lesion in the middle to superior lateral dentate nuclei, especially in the left hemisphere, suggest that the cerebellum is necessary for proper interval timing in durations lower than 12 seconds
Page 28
Why and how cerebellum contributes to timing in this range is still to be shown. It might coordinate learned actions at a fine timescale [71], playing a mediating role between the sub-second timing [72] and the supra-second timing.
todo follow up on this. See if there’s been any change in the past 6 years
Page 29
On the other hand, a recent growing body of evidence [73-78] highlights the importance of the hippocampus, an area that is usually associated to spatial learning [79] and explicit memory [80], for interval timing. Gradual changes in hippocampal activity are strongly influenced by time and distance [76]. Additionally, lesions in the dorsal or ventral hippocampus produce leftward or rightward shifts in time estimation respectively [78]. Curiously, the effects that hippocampal inactivations have on time estimation seem to be stronger when the time scale estimated is over one minute [81], and the temporal discrimination is difficult (i.e., intervals with similar durations). Yet, the same group of studies could not provide the evidence that manipulations in the hippocampus produce disruptions on timing of intervals above a second and below one minute. Anatomically, the hippocampus is highly connected with other areas relevant to interval timing such as the nucleus accumbens (Nac; [82]) and the PFC [83]. This connectivity pattern strengthens the argument that the hippocampus has a relevant role in interval timing.
Page 29
Arguably, the study of interval timing mechanisms in the BG has proven to be more prolific regarding unveiling the biological mechanisms of interval timing. Evidence from multiples sources implicates the BG, and more specifically the striatum, as a locus for the representation of supra-secondbelow-one-minute timing. Firstly, activity in the striatum is modulated by timing task as shown in studies using ensemble recording techniques in animals [84,85], and regional increase in blood flow captured by fMRI in humans during interval timing tasks [42,86]. Secondly, striatal lesions [26], diseases that affect the BG such as Huntington’s [87] and Attention Deficit Disorder [88], all cause interval timing dysfunctions.
Page 29
Furthermore, patients with disorders that involve meso-striatal dopaminergic pathways, such as schizophrenia [89,90,91] and Parkinson’s disease (PD; [92,93]), display impaired performances during interval timing tasks. PD is characterized by a progressive degeneration of nigrostriatal dopaminergic projections, leading to low levels of dopamine (DA) in the striatum. These low levels of DA cause interval timing deficits which can be alleviated by L-dopa (L-3,4-dihydroxyphenylallanine; a precursor of dopamine) treatment [94]. Malapani et al. [92,93] leveraged this modulatory effect of DA !2
Page 30
over interval timing and the therapeutic effect of L-dopa to segregate storage from retrieval dysfunction in the temporal memory in PD patients. Malapani’s data suggest that DA has the power to increase discrimination between intervals on retrieval, to control the speed and the extension of internal representation of time during encoding
Page 30
How directly DA affects time perception might be a difficult question to answer, because DA is involved in multiple processes other than timing. For instance, although genetic manipulations that affect the DA system in the BG [95] cause interval timing dysfunctions, a different source of evidence [96] suggests that DA dependent timing deficits might be a cofound of manipulations which affect directly animals’ motivation
I might want to focus on DA a bit
Page 30
Altogether, the multiple areas involved in interval timing seem to constitute not one but multiple “internal clocks”, which use diverse sources of information to implement aspects of interval timing. These areas appear to be mutually influenced by each other to generate congruent temporal estimations and subsequent adaptive behavior. A very clear example of such coordination of multiple clocks derives from the interaction among timing mechanisms of different time scales. For instance, the cerebellum exerts an influence in time estimation in the second to sub-second range. It is possible that the cerebellum exerts its influence to the BG either through modulation of thalamic input or through projections to VTA and PFC [97-100].
Ok so at this point I feel like I am just getting a list of possible areas, but nothing giving me an actual layout or network of interaction and it’s getting annoying.Â
Like, I have no idea what else could be done but it’s just one “could be this or these areas” after another.
I gotta rewrite this in some way there is a clear and total “this area, does all these things” the problem of course is that all the areas do multiple things, so you would have to repeat a lot for each part|
Page 30
Conversely, circadian timing mechanisms can affect the interval timing indirectly by regulating DA [101]. Finally, the hippocampus might have a direct effect on the computations done in the striatum, especially in long intervals, in which animals are more likely to move (so distance can be an extra source of information about the rate of change of the environment), and when information about sequence is relevant [74,76].
Page 31
. A variety of processes including reinforcement learning [104-106], motor control [107-111], limbic [112,115] and associative functions [116-119] depend on the BG
Page 34
Functionally, it has been hypothesized that the BG receive inputs from other areas of the brain and act upon these inputs as a filter. According to this hypothesis, the BG would select information derived from cortical and thalamic activity and send the resulting information back to the cortical source of the information. In parallel, it would also send copy of the same information to other systems of the brain to implement behavior. Since DA modulates the gain of cortico-striatal synapses [138-141], and because reinforcement-based plasticity occurs in the BG, it is thought that this plasticity might influence the input selection process based on previous experience [142]. The striatum and DA are considered to be key components to this filtering process.
Page 34
The way in which information travels through the multiple nuclei of the BG might offer clues about their functional role. In the canonical perspective of the BG, information from the striatum can pass through the nuclei of the BG by two different parallel circuits: the direct and the indirect pathways (Figure 1.4). Neurons within the striatum can project directly to the EP/SNr that constitute the output nuclei of the BG (direct pathway). Or instead, striatal neurons can project first to intermediate nuclei, namely GPe and STN, and then to the output nuclei EP/SNr (indirect pathway; [143]). These signaling pathways are regulated by DA in the striatum, and they have been the subject of intense study. It is known that driving activity in the direct pathway increases motor output (i.e., locomotion; [138,141,123]). On the other hand, stimulation of the indirect pathway seems to inhibit behavior [123]. For long it has been hypothesized that the direct pathway encodes the set of behavior plans available to the animals in a given context [19]. Thus, driving !3
Page 35
the direct pathway would make these motor plans stronger, and consequently increase the motor output. Complementarily to this hypothesis, the indirect pathway would map the set of competing behaviors which must be suppressed so that the selected behavioral plan can occur with less interference. In this perspective, the BG would be responsible for filtering motor, associative and limbic information using a center-surround-like receptive field in the pertinent space (e.g., behavioral, cognitive) [19].
Page 35
striatal neurons
Striatal neurons appear to be important especially for SBF theory where they project the information to be stored in the STR
Page 38
Decodings and decoders
pick back up here.
Actually , while this section is more interesting to me, it is a good place to skip in relation to the current assignment
Page 58
CHAPTER 2: A scalable population code for time in the striatum
So I’m going to comment on this from the paper @melloScalablePopulationCode2015