Dendritic spikes

The term “dendritic spike” describes strong, often stereotypical, voltage transient in the den-drite; they have some kind of threshold (voltage or other input variables, e.g., concentration of neurotransmitter in the synaptic cleft) and the response to suprathreshold input qualitatively differs from subthreshold responses, and finally the event is regenerative (see also definition in Major et al., 2013). So far three different types of dendritic spikes have been reported and they can be classified depending on the main underlying class of conductances/ion channels, i.e., there are NDMA- (based on NMDA gated ion channels; cf. also Figure 2.6), Ca²⁺-spikes (based on voltage gated calcium channels) and Na⁺-spikes (based on voltage gated sodium channels).

Na⁺-spikes

In this thesis, we consider the influence of fast (compared to other types of dendritic spikes, see below) dendritic sodium spikes on the dynamics of recurrent networks. These spikes have been found prominently in hippocampal regions CA1 in basal (Ariav et al., 2003; Losonczy et al., 2008; Remy et al., 2009; M¨uller et al., 2012) and apical dendrites (Golding and Spruston, 1998;

Gasparini et al., 2004; Jarsky et al., 2005; Losonczy and Magee, 2006; Gasparini and Magee, 2006; Makara et al., 2009), recently in CA3 (Kim et al., 2012; Makara and Magee, 2013) and also in the neocortex (Stuart et al., 1997; Larkum et al., 2001; Nevian et al., 2007).

The dendritic Na⁺-spike is initiated by voltage gated sodium channels, causing a sharp rise of the voltage transient, and shaped by concurrent activation of NMDA receptors, voltage gated Ca²⁺- and A-type K⁺-currents (cf. Figure 2.14; Ariav et al., 2003; Losonczy and Magee, 2006;

Remy et al., 2009; Kim et al., 2012). The generation of dendritic spikes requires sufficiently strong, or — if elicited by multiple presynaptic inputs — highly synchronized (in time and space) inputs (Ariav et al., 2003; Gasparini et al., 2004; Gasparini and Magee, 2006). This synchrony detection is remarkably sensitive, probably due to the small membrane time constant in thin dendrites; only sufficiently strong inputs within a very short time interval of up to roughly 3ms may generate dendritic sodium spikes (cf. also Figure 2.14).

Dendritic spikes may cause depolarizations at the soma, which substantially exceed the depo-larizations expected from summation of the effect of single inputs. Interestingly, it has been shown that this increased depolarization may even trigger somatic spikes (e.g., Ariav et al., 2003; Losonczy et al., 2008; M¨uller et al., 2012; Makara and Magee, 2013): The triggering of somatic action potentials by dendritic spikes generated in the apical dendrite (in particular in the apical tuft) has been shown to be highly variable; the dendritic spike is attenuated during transmission down the dendrite, but can be modulated (in particular reinforced) by additional inputs to the apical dendrite, which establishes a kind of gating mechanism for the remotely

40 m µ

Figure 2.14: Nonlinear amplification of synchronous inputs. (a) Fluorescent image of a CA1 pyramidal neuron. Two stimulating electrodes were placed in close proximity to a basal dendrite. (b) Traces showing the individual excitatory postsynaptic potentials (EPSPs) evoked by each of the synaptic stimulating electrodes, the summed synaptic potential dur-ing coincident activation of the two synaptic stimulatdur-ing electrodes, and the arithmetic sum of the two individual responses (bold line). Note the large supralinear amplification and sharpening of the summed synaptic potential, as compared with the expected arith-metic sum response. The fast component is attributed to voltage gated sodium channels, followed by a longer lasting component mediated by subsequent opening of NMDA recep-tors. (c) Voltage traces obtained in response to coincident activation of two closely spaced electrodes at various time delays (0−20ms). Black traces represent voltage responses to activation of the electrodes at time delays of 0 and 2ms. Gray traces show the responses for activation at time delays of{3,5,10,15,20}ms. Note that, in this experiment, the time window for coincident detection was<3ms. Figure and caption modified from Ariav et al.

(2003) with permission.

generated signal (Golding and Spruston, 1998; Gasparini et al., 2004; Jarsky et al., 2005; Makara and Magee, 2013). Dendritic sodium spikes generated in basal (or radial oblique) dendrites are much more reliable in triggering somatic spikes (Ariav et al., 2003; Losonczy et al., 2008; M¨uller et al., 2012; Makara and Magee, 2013). The timing of the evoked somatic action potential can be surprisingly precise; the temporal jitter is in the sub-millisecond range (Ariav et al., 2003;

Losonczy et al., 2008).

Figure 2.15: Characteristics of dendritic spikes evoked by optically stimulated gluta-mate uncaging in strong (red) and weak (blue) dendritic branches(modified from Losonczy et al., 2008, with permission).

The strength of sodium spikes (both the am-plitude ∆V and rate of risedV /dt) is relatively invariant between different trials, but there is remarkable variance between the branches (Losonczy et al., 2008; Makara et al., 2009;

M¨uller et al., 2012). They can be classi-fied in strong and weak dendritic branches (cf. Figure 2.15). The propagation of weak dendritic spikes is strongly attenuated until they reach the soma, and the timing of sub-sequent action potentials — probably elicited by the slow, longer lasting spike component following the initial sodium spike, cf. Fig-ure 2.14 and 2.15 — is relatively unreliable;

thus the temporal jitter between trials is high (Gasparini et al., 2004; Losonczy et al., 2008;

M¨uller et al., 2012). In contrast, dendritic spikes generated in strong branches require less presynaptic stimulation, are transmitted essentially without attenuation, are very reli-able in triggering of somatic action potentials

— probably due to the fast sodium spike — with sub-millisecond precision (Ariav et al., 2003; Losonczy and Magee, 2006; Losonczy et al., 2008; M¨uller et al., 2012; Makara and Magee, 2013).

Interestingly, it has been shown that weak branches can be transformed into strong branches by suitable stimulation protocols — e.g., by mimicking the presynaptic stimulation on hippocampal pyramidal cells during exploration of space, cf. also Section 2.5.3 — and/or by the application of neuromodulators (Losonczy et al., 2008; M¨uller et al., 2012). This suggests that single branches may act as sub-processing units which can be turned on and off in an experience-based manner, e.g., during phases of spatial exploration.

Finally, it is noteworthy that strong dendritic sodium spikes cannot be suppressed by (both recurrent or locally applied) inhibition, whereas weak dendritic spikes are completely attenuated (M¨uller et al., 2012; though the inhibition decreases the probability that the dendritic spike can elicit a somatic spike). Interestingly, this mechanism further increases the output precision of

somatic spikes: All somatic action potentials, but the highly precise ones caused by strong dendritic spikes, are suppressed.

Taken together, (strong) dendritic sodium spikes act as powerful coincidence detectors with remarkable precision, regarding both the detection of input synchrony and the generation of output spikes. They are prominently found in the basal dendrites of hippcampal pyramidal neurons. In this dendritical region most of the recurrent excitatory inputs (from other pyramidal of the same region) in hippocampal area CA1/CA3 arrive (Andersen et al., 2007; Cutsuridis et al., 2010). Thus, the available data suggest that dendritic sodium spikes are crucially involved in processing synchronous inputs in the hippocampus (e.g., during Sharp-Wave-Ripple events;

cf. Sections 2.5.3, 2.5.4 and Chapter 7).

NMDA- and Ca²⁺-spikes

As mentioned above, additionally to dendritic sodium spikes, dendritic NMDA-spikes and calcium-spikes have been reported in the hippocampus and neocortical areas (cf., e.g., Major et al., 2013, for a review). These spike also nonlinearly amplify presynaptic inputs, but are far less sensitive to input synchrony. For example, Polsky et al. (2004) reported that two stimuli on the same dendritic branch cause somatic responses exceeding the level expected from linear summation of inputs up to time intervals of more than 40 ms between the stimuli. This insensibility (com-pared to sodium spikes) is attributed to the fact that, in particular NMDA receptors (cf. also Figure 2.6 and describing text), typically deactivate only slowly (Paoletti, 2011, and references therein). As a consequence, the caused somatic depolarization is long lasting or “plateau-like”

(Major et al., 2013, and references therein), and NMDA-spikes and calcium spikes are often associated with the generation of bursts of action potentials (Traub and Wong, 1982; Williams and Stuart, 1999; Larkum and Zhu, 2002; Milojkovic et al., 2004; Polsky et al., 2009; Long et al., 2010).