Experiments were carried out in isolated ganglia obtained from leeches Haementeria officinalis with 15–80 mm body lengths. The somatic diameters of Retzius neurons in these ganglia ranged from 21 to 75 μm. Intersegmental ganglia were dissected out as in De-Miguel et al. (2001) and maintained in leech Ringer fluid composed of (mM): NaCl, 115; KCl, 4; CaCl₂, 1.8; Glucose, 11; Trismaleate 10 buffered to pH 7.4 (JT Baker, México D.F., México). The somata of Retzius neurons could be unequivocally identified by their characteristic size and position in the ganglion (see De-Miguel et al., 2001). Experiments were made at room temperature (20–25°C).

The somatic diameters were measured from digital images of the ganglion by use of a Motic Digital Camera Plus 2.0 (China) adapted to a Nikon binocular microscope. Since most somata were elliptical, the minor and major axis of every Retzius cell were measured with Moticanplus 2.0 software. The surface area (A) of the soma was calculated with the relationship A = 4abπ, where “a” and “b” were the minor and major radiuses respectively. The average diameter of each spherical soma obtained as d = (A/π)^1/2 ranged from 21 to 75 μm. The largest diameters obtained here were similar to those reported in adult animals by Garcia-Perez et al. (2004). Based on the same previous study, the length (l) of the coupled dendrites of each neuron was calculated by assuming isometric growth in a proportion of 0.83 times the soma diameter (Garcia-Perez et al., 2004).

Intracellular recordings were made with borosilicate glass microelectrodes (FHC Inc, Bowdoinham, ME, USA) filled with 3 M KCl. Their tip resistances ranged from 16 to 24 MΩ. The microelectrodes were coupled to an Axoclamp pre-Amplifier (Molecular Devices: Foster City, CA, USA). Our recordings were filtered by a custom design Bessel filter with a cutoff frequency of 400 s⁻¹. Data were acquired by an analog to digital board Digidata 1200 (Molecular Devices, Foster City, CA, USA) using Axoclamp 9.0 and stored in a PC for further analysis. The resting potential of neurons was between −55 and −60 mV and was held at −60 mV by injection of DC negative current when necessary. Current injection and voltage recordings from the driving neuron were made through a single electrode under discontinuous current clamp conditions with the switch operating at 0.9 KHz. Pseudorandom noise currents with 10 nA peak to peak amplitude and containing frequencies between 0.1 and 100 s⁻¹ were injected through the microelectrode by a custom designed noise generator. The voltage from the follower neuron was recorded by an independent microelectrode.

Simultaneous recordings from both neurons needed to last for at least 20 min under optimal conditions, during which we repeated three times a sequence of 5 min acquisition of spontaneously-appearing EPSPs. This was followed by injection of a series of 16 square negative current pulses lasting 200 ms with gradually increasing amplitudes up to −3.0 nA to measure the input resistance of the neurons, the stationary coupling ratio and the somatic time constant. In the last part of the sequence we applied a 60 s-lasting pseudorandom noise current to determine the voltage frequency responses of each pair of neurons. The inclusion criteria for the data analysis was the value of the membrane time constant of the soma, the input resistance of the pair of neurons and the maintenance of the rise time and amplitude of spontaneous EPSPs along the whole experiment. The repetition of this whole sequence for at least three times allowed us to obtain enough EPSPs to analyze their amplitude and rise time distribution, to measure the frequency of the action potentials and the biophysical variables contributing to synaptic integration.

The list of the parameters estimated in this study is presented in Table 1. The membrane time constant (τ) of the soma was calculated from a single exponential fitting of the passive membrane discharge of steady-state voltage responses upon the end of the square current pulses. Even though it is well-known that these discharges include information of the membrane properties of the dendritic tree, expressed as a combined series of exponentials (Rall, 1969), the large soma of Retzius neurons, which is isopotentially connected with the axon is nearly two orders of magnitude thicker than the 1 μm diameter dendrites. Therefore, the soma-axon compartment dominates the membrane charge and discharge at the recording site (Garcia-Perez et al., 2004; Vazquez et al., 2009). For our analysis we considered the specific capacitance of the membrane as C_m = 1 μF/cm². The experimentally obtained τ values served then to calculate the somatic membrane resistance value (r_m[MΩ.cm]) by assuming that τ = r_mc_m.

The frequency-dependent coupling ratio was calculated by dividing the average power spectrum of the follower neuron by the average power spectrum of the driving neuron (Di Caprio et al., 1974; Yang and Chapman, 1983). The stationary coupling ratio was obtained from the “y” axis intersection at a 0.1 s⁻¹ frequency.

The values of the dendritic membrane resistance (r_dend) and space constant (λ), and the coupling resistance (r_c) were obtained by fitting the frequency responses of the neurons to a model of electrically-coupled Retzius neurons based on linear cable theory. The model based on the morphology of the Retzius neurons is schematized in Figure 1 (Garcia-Perez et al., 2004; Vazquez et al., 2009). In the model the soma is isopotentially connected to the axon and both are represented as a parallel circuit of membrane resistance and capacitance. The coupled dendrites are represented by cylinders with a length of 0.83 times the soma diameter (Garcia-Perez et al., 2004). One extreme of the dendrites is connected to the soma/axon compartment and the other extreme is connected to the dendrites of the coupled neuron through a resistor representing the electrical synapse. For this, the ratio of the power spectra of the pseudo-random driving and follower voltage responses to pseudorandom noise injection containing frequencies from 0.1 to 100 s⁻¹ gave the frequency-dependent coupling coefficient of the neuron pair. The model was fed with the diameter and membrane resistance of the soma calculated from the steady state voltage responses to current injection. We assumed that at birth Retzius neurons already have their characteristic arborization (Jellies et al., 1987) and that their growth is isometric. Therefore, all the morphological parameters increase linearly. The dendrite length was set as 0.85 times the soma diameter and as free parameters we left the r_m and r_c values. In addition, the r_c value obtained with the model was confirmed with that obtained from the steady state measurements described above.

Examples of intact leeches with the smaller and larger sizes used in this study are shown in Figure 2. Their isolated ganglia are also shown displaying the somata of the Retzius neurons. The overall electrical activity recorded from nine pairs of these Retzius neurons with average soma diameters between 21 and 75 μm was similar to that already described in adult neurons (Hagiwara and Morita, 1962), consisting of a resting potential between −55 and −60 mV, on top of which pairs of EPSPs appeared simultaneously in both somata with amplitudes that varied randomly between neurons and from one pair to the next (Figures 3A–C), thus suggesting presynaptic stepwise variations in transmitter release (De-Miguel et al., 2001). Summation of EPSPs produced in the dendrites of both Retzius neurons produced the action potentials that characterized the basal activity of Retzius neurons (Figure 3D).

In neurons held at −60 mV the EPSPs occurred at average frequencies ranging from one neuron pair to another between 0.5 and 1.7 s⁻¹, again without any correlation with the soma diameter (Figure 4A). The action potentials that complement the baseline electrical activity of these neurons were produced with average frequencies of 0.004–0.0017 s⁻¹ upon summation of the EPSPs originated in both coupled neurons (Figure 4B). However, the action potential frequency did not correlate with the EPSPs frequency (Figure 4C), presumably because summation was multivariable and included the proportion of local and propagated EPSPs in each pair of neurons, the individual EPSPs arriving from the coupled dendrites, the stochastic presynaptic amplitude fluctuations, and as will be seen below, the modulation of the EPSP amplitude upon the leak of synaptic currents across the coupling resistance.

A third parameter contributing to integration in these neurons is their electrical coupling. In adult neurons the coupling ratio calculated from steady state voltage responses upon square current pulses (Figure 5A) ranges between 0.2 and 0.7 (Garcia-Perez et al., 2004). In these experiments, the coupling ratio had a smaller range presumably because the unpredictability of its variations from leech to leech. However, the coupling ratio of individual neuron pairs failed to correlate with their soma diameter (Figure 5B). For example, neurons from the two smaller pairs studied, having 21 and 26 μm soma diameters, had coupling ratios of 0.42 and 0.64, respectively, whereas neurons with a 36 μm diameter also had a high 0.64 coupling ratio. By contrast, the largest neuron studied here had a 75 μm diameter and an intermediate 0.53 coupling ratio.

The amplitude and rise time of EPSPs also provide information about integration in terms of their presynaptic contents, their origin in the dendrites and their spread along the dendrite to the soma and their spread to the coupled neuron across the electrical synapse (De-Miguel et al., 2001; Garcia-Perez et al., 2004). The rise time distribution of EPSPs in every neuron had two peaks, with the value of each peak indicating the EPSPs origin in the coupled dendrites (Figure 6). Those EPSPs with fast rise times (59–81%, n = 9 neurons) ranging from 3.63 ± 0.5 to 5.83 ± 0.9 ms are produced in the coupled dendrites of the driver neuron, whereas the EPSPs (19–41%) with slower rise times between 7.33 ± 0.6 and 9.13 ± 1.3 ms (n = 9) and smaller amplitudes (not shown, De-Miguel et al., 2001) are originated in the dendrites of the coupled neuron, from which they arrived after crossing the electrical synapse (De-Miguel et al., 2001; Garcia-Perez et al., 2004). We have shown that the slower EPSPs are detected upon local failures in transmitter release onto the driving neuron, since otherwise they appear as a hump in the decay phase of the local EPSPs upon summation in the primary axon (Vazquez et al., 2009). The ranges of EPSP rise time values and their proportions were similar to those already described in adult neurons (Garcia-Perez et al., 2004), and had no correlation with the soma diameter (0.067; Figure 7A) thus suggesting that in spite of the dendrite elongation, the EPSP electrotonic spread remains constant as neurons grow.

The amplitude distribution of the local EPSPs obtained from Gaussian fittings to the fast-rising EPSP population recorded from the nine pairs plus three other neurons (Figure 6) had one peak formed by the smaller events (n = 21 neurons); a second peak (n = 19 neurons) had events whose amplitude nearly doubled that of the first peak, and in records from two neurons a third small peak contained even largest EPSPs. These distributions along with the transmission failures onto each neuron suggested a quantal nature in the EPSP amplitudes, with the smaller amplitude peaks being produced by the unitary events.

The unitary EPSP amplitude varied from neuron to neuron between 0.36 ± 0.05 and 0.7 ± 0.8 mV. However these variations were not correlated with the soma diameter (R² = 0.03; Figure 7B). Instead of that and consistent with our previous prediction (Garcia-Perez et al., 2004) the EPSP amplitude was inversely proportional to the coupling ratio of the neuron pair (Figure 7C) with a 0.78 correlation coefficient. The explanation for this effect is that electrical coupling modulates the amplitude of the EPSPs by allowing a synaptic current leak in proportion to the coupling resistance value. Indeed, the EPSP amplitude was similar in neurons with similar coupling ratio. For example, in a small neuron with a 22 μm soma diameter and a 0.61 coupling ratio, the EPSP distribution had a major 0.45 ± 0.05 mV amplitude peak, whereas a larger neuron in another pair with a 33 μm soma diameter and a similar 0.62 coupling ratio, produced unitary EPSPs with a 0.43 ± 0.5 mV amplitude. The variability increases of the data at high coupling ratios shown in Figure 7 could be due to contamination by the 0.18 mV root-mean-square (RMS) noise in the recordings and to a higher susceptibility of smaller EPSPs to input resistance differences and to the multivariable influences on their spread. Consistently with the behavior of EPSPs, the frequency of the action potentials decreased as the coupling ratio increased. The low correlation coefficient of 0.5 of the plot shown in Figure 7B can be due to the multiple variables influencing EPSP summation. These data confirm that at similar coupling coefficients the membrane properties of the neurons compensate their size to produce constant integrative properties.

In addition to the somatic membrane resistance and capacitance of the soma estimated from the voltage changes upon hyperpolarizing square current pulses (Figure 5), the frequency-dependent coupling ratio was calculated by injecting pseudo-random noise current into the driving neuron (Figure 8A). The power spectrum of the follower neuron was then divided by the average power spectrum of the driving neuron (Figure 8B) to obtain the frequency-dependent coupling ratio (Figures 8C,D). Model fitting to the data provided a calculation of the λ value of the dendrites and confirmed the somatic membrane time constant. In addition, the coupling ratios ranging from 0.41 to 0.64 obtained from steady-state responses correlated linearly (R = 0.96) with those obtained from noise analysis extrapolating the model fitting to a 0.1 s⁻¹ frequency.

The time constant values measured from the exponential voltage decay had a scattered distribution between 20 and 30 ms for the whole range of somatic diameters (Figure 9). This range was within the 18–40 ms range obtained in our previous experiments in isolated adult somata (Garcia-Perez et al., 2004), thus suggesting that the correction factor for growth was in the coupled dendrites. It is worth to remark here that the large soma-axon dominates the voltage responses of these neurons upon current injection, thus making the contribution of the dendrites negligible in soma charging experiments. Therefore these voltage responses are expressed as a single exponential change, as compared with the dual exponentials that characterize the somatic voltage changes in the presence of small somata and large dendritic trees (Rall, 1959). As expected, the membrane capacitance (c_m) values of the soma-axon compartment increased linearly (R = 0.99) with the soma diameter, from 6.6 pF when the diameter was 21 μm to 23.4 pF when the soma diameter was 75 μm. By contrast, the soma membrane resistance (r_m) decreased exponentially (R² = 0.9) from 3.4 MΩ.cm when the diameter was 21 μm to 1.4 MΩ.cm when the soma diameter was 74 μm (Figure 9B). Therefore the variability of the somatic time constant values was produced by the combination of the linear capacitance increase and the exponential resistance decrease as the soma increased.

The parameter compensating for the neuron growth was the membrane resistance of the dendrites, which displayed an exponential increase (R² = 0.82) as the somatic diameter was larger (Figure 9C). By assuming a realistic constant value of 85 KΩ/cm for the internal resistance of the dendrites (Garcia-Perez et al., 2004) the exponential increase in the r_dend value also produced an increase in the dendritic λ value (Figure 9C, right axis). Therefore, the effective length L = l/λ of the dendrites increased exponentially (R² = 0.91) as the soma diameter increased (Figure 9D) to reach a plateau when the soma approached a 50 μm diameter.

The persistency of the axo-somatic time constant as neurons grow imposes a similar decay to EPSPs arriving at the primary axon-soma compartment independently on the neuron size. Since the frequency range at which EPSPs appear similar as neurons grow, the temporal summation that determines the basal firing frequency also remains constant provided amplitudes of EPSPs are similar. Therefore, the variable determining the firing frequency at all neuronal sizes is the coupling ratio of the neurons. A second factor contributing to synaptic integration is the small dendro-somatic conductance coefficient, which reduces significantly the amplitude of EPSPs upon arrival at the axon (Vazquez et al., 2009). Although this coefficient changes the amplitude of all the EPSPs but not their shape, the remodeling of the EPSPs entering the soma/axon compartment determines how they sum (Vazquez et al., 2009). Our data indicates a constant dendro-somatic coefficient since neurons of different sizes but similar coupling ratios had similar EPSP amplitudes and firing frequencies. It was surprising that the electrical coupling values of small neurons spans over a similar wide range as adult neurons. This result indicates that from early postnatal life electrical coupling has already its modulatory capacity to regulate the basal firing frequency of the neurons through the regulation of the EPSP amplitude.

Serotonergic neurons of vertebrates and invertebrates are multimodal in the sense that they release transmitter from synapses or from extrasynaptic sites, depending on their electrical activity pattern (De-Miguel et al., 2015). In general, serotonergic neurons display spontaneous firing at low frequencies (Mosko and Jacobs, 1974; Wang and Aghajanian, 1977; Cunningham and Lakoski, 1988; Mason, 1997; Veasey et al., 1997), which do not evoke extrasynaptic transmission. Even the smallest Retzius neurons studied here had established the pattern of EPSPs generation that characterizes adult neurons. Summation of EPSPs produces the low firing frequency of Retzius neurons (Garcia-Perez et al., 2004). The early establishment of this pattern implies that from early life this circuit has acquired the basal firing frequency that will continue during adulthood, in contrast to the gradual increase in the number of synaptic inputs in cricket or rat neurons as they grow (Chiba et al., 1992; Liu et al., 1996). At their basal firing frequency, Retzius neurons may release serotonin from synaptic endings (Henderson, 1983; Liu and Nicholls, 1989). However, serotonergic neurons also release serotonin extrasynaptically from their soma, dendrites and varicosities when the firing frequency increases (Trueta et al., 2003; Leon-Pinzon et al., 2014) and by doing this may modulate whole circuits (De-Miguel et al., 2015). This extrasynaptic exocytosis requires another set of excitatory connections from mechanosensory neurons (Szczupak and Kristan, 1995; Velázquez-Ulloa et al., 2003). Firing of pressure sensory neurons upon skin stimulation produces a proportional firing increase in adult Retzius neurons (Velázquez-Ulloa et al., 2003). The massive serotonin release from the soma increases the levels of circulating serotonin in the ganglion (Willard, 1981) and evokes fictive swimming (Nusbaum and Kristan, 1986) or crawling (De-Miguel et al., 2015). A whole scheme to understand serotonergic function will require a comprehensive analysis of the development of this alternative pathway. Another question concerns the modulation of electrical coupling. In serotonergic neurons this modulation may be produced by serotonin release (Colombaioni and Brunelli, 1988; Rorig and Suttor, 1996; Szabo et al., 2010) acting on the gap junction channels (Phelan et al., 1998; Kandarian et al., 2012). However Retzius neurons from small animals are insensitive to serotonergic modulation (Groome et al., 1995), thus suggesting on one hand that the extrasynaptic communication may not yet be functional and on the other that the modulation of gap junctions is produced by a different and yet unknown mechanism.