Roberto Fernández Galán

Computational Physics, Neuroscience

PhD
30.18

Publications

  • Pavel Anatolyevich Puzerey · Nathan Xing Kodama · Roberto Fernandez Galan
    [Show abstract] [Hide abstract]
    ABSTRACT: Neurons originating from the raphe nuclei of the brainstem are the exclusive source of serotonin (5-HT) to the cortex. Their serotonergic phenotype is specified by the transcriptional regulator Pet-1, which is also necessary for maintaining their neurotransmitter identity across development. Transgenic mice in which Pet-1 is genetically ablated (Pet-1(-/-)) show a dramatic reduction (~80%) in forebrain 5-HT levels, yet no investigations have been carried out to assess the impact of such severe 5-HT depletion on the function of target cortical neurons. Using whole-cell patch clamp methods, 2-D multielectrode arrays (MEA), 3-D morphological neuronal reconstructions, and animal behavior, we investigated the impact of 5-HT depletion on cortical cell-intrinsic and network excitability. We found significant changes in several parameters of cell-intrinsic excitability in cortical pyramidal cells (PC), as well as an increase in spontaneous synaptic excitation through 5-HT3 receptors. These changes are associated with increased local network excitability and oscillatory activity in a 5-HT2 receptor-dependent manner, consistent with previously reported hypersensitivity of cortical 5-HT2 receptors. PC morphology was also altered with a significant reduction in dendritic complexity which may possibly act as a compensatory mechanism for increased excitability. Consistent with this interpretation, when we carried out experiments with convulsant-induced seizures to asses cortical excitability in vivo, we observed no significant differences in seizure parameters between wild-type and Pet-1(-/-) mice. Moreover, MEA recordings of propagating field potentials showed diminished propagation of activity across the cortical sheath. Altogether these findings reveal novel functional changes in neuronal and cortical excitability in mice lacking Pet-1.
    No preview · Article · Nov 2015 · Journal of Neurophysiology
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Random ion channel gating is an important source of noise at the single neuron level. The fundamental description of ion channel fluctuations based on molecular kinetics is well established. A population of ion channels in a single compartment, isopotential cell, comprises a continuous time, discrete state Markov process. For example, the classic Hodgkin-Huxley sodium channel model has eight states connected by twenty transitions. The per capita transition rates between ion channel states are typically voltage dependent. Since the ion channel state in turn determines the voltage evolution, the combined (voltage, channel state) system is a hybrid continuous/discrete Markov process. When the number of copies of each channel type is sufficiently large, a system size expansion leads to practical approximate representation of the fluctuations through a Langevin or Fokker-Planck approach. Several variants are in use, and the large-N limit has been studied rigorously. Nevertheless, for mesoscale population sizes, fast, accurate representation of channel noise fluctuations remains an important challenge. The presentation will highlight recent progress in three areas.
    Full-text · Article · Jul 2015 · BMC Neuroscience
  • Source
    Roberto Fernández Galán · G Bard Ermentrout · Nathaniel N Urban

    Full-text · Dataset · Nov 2014
  • Pavel Anatolyevich Puzerey · Michael J Decker · Roberto Fernandez Galan
    [Show abstract] [Hide abstract]
    ABSTRACT: Serotonin fibers densely innervate the cortical sheath to regulate neuronal excitability, but its role in shaping network dynamics remains undetermined. We show that serotonin provides an excitatory tone to cortical neurons in the form of spontaneous synaptic noise through 5-HT3 receptors, which is persistent and can be augmented using fluoxetine, a selective serotonin reuptake inhibitor. Augmented serotonin signaling also increases cortical network activity by enhancing synaptic excitation through activation of 5-HT2 receptors. This in turn facilitates the emergence of epileptiform network oscillations (10-16 Hz) known as fast runs. A computational model of cortical dynamics demonstrates that these two combined mechanisms, increased background synaptic noise and enhanced synaptic excitation, are sufficient to replicate the emergence fast runs and their statistics. Consistent with these findings, we show that blocking 5-HT2 receptors in vivo significantly raises the threshold for convulsant-induced seizures.
    No preview · Article · Aug 2014 · Journal of Neurophysiology
  • Source
    Pavel A Puzerey · Roberto F Galán
    [Show abstract] [Hide abstract]
    ABSTRACT: Cortical neurons receive barrages of excitatory and inhibitory inputs which are not independent, as network structure and synaptic kinetics impose statistical correlations. Experiments in vitro and in vivo have demonstrated correlations between inhibitory and excitatory synaptic inputs in which inhibition lags behind excitation in cortical neurons. This delay arises in feed-forward inhibition (FFI) circuits and ensures that coincident excitation and inhibition do not preclude neuronal firing. Conversely, inhibition that is too delayed broadens neuronal integration times, thereby diminishing spike-time precision and increasing the firing frequency. This led us to hypothesize that the correlation between excitatory and inhibitory synaptic inputs modulates the encoding of information of neural spike trains. We tested this hypothesis by investigating the effect of such correlations on the information rate (IR) of spike trains using the Hodgkin-Huxley model in which both synaptic and membrane conductances are stochastic. We investigated two different synaptic input regimes: balanced synaptic conductances and balanced currents. Our results show that correlations arising from the synaptic kinetics, τ, and millisecond lags, δ, of inhibition relative to excitation strongly affect the IR of spike trains. In the regime of balanced synaptic currents, for short time lags (δ ~ 1 ms) there is an optimal τ that maximizes the IR of the postsynaptic spike train. Given the short time scales for monosynaptic inhibitory lags and synaptic decay kinetics reported in cortical neurons under physiological contexts, we propose that FFI in cortical circuits is poised to maximize the rate of information transfer between cortical neurons. Our results also provide a possible explanation for how certain drugs and genetic mutations affecting the synaptic kinetics can deteriorate information processing in the brain.
    Full-text · Article · Jun 2014 · Frontiers in Computational Neuroscience
  • [Show abstract] [Hide abstract]
    ABSTRACT: Cardiorespiratory coupling is an encompassing term describing more than the well-recognized influences of respiration on heart rate and blood pressure. Our data indicate that cardiorespiratory coupling reflects a reciprocal interaction between autonomic and respiratory control systems, and the cardiovascular system modulates the ventilatory pattern as well. For example, cardioventilatory coupling refers to the influence of heart beats and arterial pulse pressure on respiration and is the tendency for the next inspiration to start at a preferred latency after the last heart beat in expiration. Multiple complementary, well-described mechanisms mediate respiration's influence on cardiovascular function, whereas mechanisms mediating the cardiovascular system's influence on respiration may only be through the baroreceptors but are just being identified. Our review will describe a differential effect of conditioning rats with either chronic intermittent or sustained hypoxia on sympathetic nerve activity but also on ventilatory pattern variability. Both intermittent and sustained hypoxia increase sympathetic nerve activity after 2 weeks but affect sympatho-respiratory coupling differentially. Intermittent hypoxia enhances sympatho-respiratory coupling, which is associated with low variability in the ventilatory pattern. In contrast, after constant hypobaric hypoxia, 1-to-1 coupling between bursts of sympathetic and phrenic nerve activity is replaced by 2-to-3 coupling. This change in coupling pattern is associated with increased variability of the ventilatory pattern. After baro-denervating hypobaric hypoxic-conditioned rats, splanchnic sympathetic nerve activity becomes tonic (distinct bursts are absent) with decreases during phrenic nerve bursts and ventilatory pattern becomes regular. Thus, conditioning rats to either intermittent or sustained hypoxia accentuates the reciprocal nature of cardiorespiratory coupling. Finally, identifying a compelling physiologic purpose for cardiorespiratory coupling is the biggest barrier for recognizing its significance. Cardiorespiratory coupling has only a small effect on the efficiency of gas exchange; rather, we propose that cardiorespiratory control system may act as weakly coupled oscillator to maintain rhythms within a bounded variability.
    No preview · Article · Apr 2014 · Progress in brain research
  • Source
    Jose L. Perez Velazquez · Roberto F Galán
    [Show abstract] [Hide abstract]
    ABSTRACT: Along with the study of brain activity evoked by external stimuli, an increased interest in the research of background, "noisy" brain activity is fast developing in current neuroscience. It is becoming apparent that this "resting-state" activity is a major factor determining other, more particular, responses to stimuli and hence it can be argued that background activity carries important information used by the nervous systems for adaptive behaviors. In this context, we investigated the generation of information in ongoing brain activity recorded with magnetoencephalography (MEG) in children with autism spectrum disorder (ASD) and non-autistic children. Using a stochastic dynamical model of brain dynamics, we were able to resolve not only the deterministic interactions between brain regions, i.e., the brain's functional connectivity, but also the stochastic inputs to the brain in the resting state; an important component of large-scale neural dynamics that no other method can resolve to date. We then computed the Kullback-Leibler (KLD) divergence, also known as information gain or relative entropy, between the stochastic inputs and the brain activity at different locations (outputs) in children with ASD compared to controls. The divergence between the input noise and the brain's ongoing activity extracted from our stochastic model was significantly higher in autistic relative to non-autistic children. This suggests that brains of subjects with autism create more information at rest. We propose that the excessive production of information in the absence of relevant sensory stimuli or attention to external cues underlies the cognitive differences between individuals with and without autism. We conclude that the information gain in the brain's resting state provides quantitative evidence for perhaps the most typical characteristic in autism: withdrawal into one's inner world.
    Full-text · Article · Dec 2013 · Frontiers in Neuroinformatics

  • No preview · Article · Dec 2013 · Journal of Critical Care
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: This project aimed to determine if a correlation-based measure of functional connectivity can identify epileptogenic zones from intracranial EEG signals, as well as to investigate the prognostic significance of such a measure on seizure outcome following temporal lobe lobectomy. To this end, we retrospectively analyzed 23 adult patients with intractable temporal lobe epilepsy (TLE) who underwent an invasive stereo-EEG (SEEG) evaluation between January 2009 year and January 2012. A follow-up of at least one year was required. The primary outcome measure was complete seizure-freedom at last follow-up. Functional connectivity between two areas in the temporal lobe that were sampled by two SEEG electrode contacts was defined as Pearson's correlation coefficient of interictal activity between those areas. SEEG signals were filtered between 5 and 50 Hz prior to computing this correlation. The mean and standard deviation of the off diagonal elements in the connectivity matrix were also calculated. Analysis of the mean and standard deviation of the functional connections for each patient reveals that 90% of the patients who had weak and homogenous connections were seizure free one year after temporal lobectomy, whereas 85% of the patients who had stronger and more heterogeneous connections within the temporal lobe had recurrence of seizures. This suggests that temporal lobectomy is ineffective in preventing seizure recurrence for patients in whom the temporal lobe is characterized by weakly connected, homogenous networks. This pilot study shows promising potential of a simple measure of functional brain connectivity to identify epileptogenicity and predict the outcome of epilepsy surgery.
    Full-text · Article · Oct 2013 · PLoS ONE
  • Source
    Luis García Domínguez · José Luis Pérez Velázquez · Roberto Fernández Galán
    [Show abstract] [Hide abstract]
    ABSTRACT: We present an efficient approach to discriminate between typical and atypical brains from macroscopic neural dynamics recorded as magnetoencephalograms (MEG). Our approach is based on the fact that spontaneous brain activity can be accurately described with stochastic dynamics, as a multivariate Ornstein-Uhlenbeck process (mOUP). By fitting the data to a mOUP we obtain: 1) the functional connectivity matrix, corresponding to the drift operator, and 2) the traces of background stochastic activity (noise) driving the brain. We applied this method to investigate functional connectivity and background noise in juvenile patients (n = 9) with Asperger's syndrome, a form of autism spectrum disorder (ASD), and compared them to age-matched juvenile control subjects (n = 10). Our analysis reveals significant alterations in both functional brain connectivity and background noise in ASD patients. The dominant connectivity change in ASD relative to control shows enhanced functional excitation from occipital to frontal areas along a parasagittal axis. Background noise in ASD patients is spatially correlated over wide areas, as opposed to control, where areas driven by correlated noise form smaller patches. An analysis of the spatial complexity reveals that it is significantly lower in ASD subjects. Although the detailed physiological mechanisms underlying these alterations cannot be determined from macroscopic brain recordings, we speculate that enhanced occipital-frontal excitation may result from changes in white matter density in ASD, as suggested in previous studies. We also venture that long-range spatial correlations in the background noise may result from less specificity (or more promiscuity) of thalamo-cortical projections. All the calculations involved in our analysis are highly efficient and outperform other algorithms to discriminate typical and atypical brains with a comparable level of accuracy. Altogether our results demonstrate a promising potential of our approach as an efficient biomarker for altered brain dynamics associated with a cognitive phenotype.
    Full-text · Article · Apr 2013 · PLoS ONE
  • Source
    Luis García Domínguez · José Luis Pérez Velázquez · Roberto Fernández Galán
    [Show abstract] [Hide abstract]
    ABSTRACT: We present an efficient approach to discriminate between typical and atypical brains from macroscopic neural dynamics recorded as magnetoencephalograms (MEG). Our approach is based on the fact that spontaneous brain activity can be accurately described with stochastic dynamics, as a multivariate Ornstein-Uhlenbeck process (mOUP). By fitting the data to a mOUP we obtain: 1) the functional connectivity matrix, corresponding to the drift operator, and 2) the traces of background stochastic activity (noise) driving the brain. We applied this method to investigate functional connectivity and background noise in juvenile patients (n = 9) with Asperger's syndrome, a form of autism spectrum disorder (ASD), and compared them to age-matched juvenile control subjects (n = 10). Our analysis reveals significant alterations in both functional brain connectivity and background noise in ASD patients. The dominant connectivity change in ASD relative to control shows enhanced functional excitation from occipital to frontal areas along a parasagittal axis. Background noise in ASD patients is spatially correlated over wide areas, as opposed to control, where areas driven by correlated noise form smaller patches. An analysis of the spatial complexity reveals that it is significantly lower in ASD subjects. Although the detailed physiological mechanisms underlying these alterations cannot be determined from macroscopic brain recordings, we speculate that enhanced occipital-frontal excitation may result from changes in white matter density in ASD, as suggested in previous studies. We also venture that long-range spatial correlations in the background noise may result from less specificity (or more promiscuity) of thalamo-cortical projections. All the calculations involved in our analysis are highly efficient and outperform other algorithms to discriminate typical and atypical brains with a comparable level of accuracy. Altogether our results demonstrate a promising potential of our approach as an efficient biomarker for altered brain dynamics associated with a cognitive phenotype.
    Full-text · Dataset · Apr 2013
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Rett syndrome, a severe X-linked neurodevelopmental disorder caused by mutations in the gene encoding methyl-CpG-binding protein 2 (Mecp2), is associated with a highly irregular respiratory pattern including severe upper-airway dysfunction. Recent work suggests that hyperexcitability of the Hering-Breuer reflex (HBR) pathway contributes to respiratory dysrhythmia in Mecp2 mutant mice. To assess how enhanced HBR input impacts respiratory entrainment by sensory afferents in closed-loop in vivo-like conditions, we investigated the input (vagal stimulus trains) - output (phrenic bursting) entrainment via the HBR in wild-type and MeCP2-deficient mice. Using the in situ perfused brainstem preparation, which maintains an intact pontomedullary axis capable of generating an in vivo-like respiratory rhythm in the absence of the HBR, we mimicked the HBR feedback input by stimulating the vagus nerve (at threshold current, 0.5 ms pulse duration, 75 Hz pulse frequency, 100 ms train duration) at an inter-burst frequency matching that of the intrinsic oscillation of the inspiratory motor output of each preparation. Using this approach, we observed significant input-output entrainment in wild-type mice as measured by the maximum of the cross-correlation function, the peak of the instantaneous relative phase distribution, and the mutual information of the instantaneous phases. This entrainment was associated with a reduction in inspiratory duration during feedback stimulation. In contrast, the strength of input-output entrainment was significantly weaker in Mecp2 (-/+) mice. However, Mecp2 (-/+) mice also had a reduced inspiratory duration during stimulation, indicating that reflex behavior in the HBR pathway was intact. Together, these observations suggest that the respiratory network compensates for enhanced sensitivity of HBR inputs by reducing HBR input-output entrainment.
    Full-text · Article · Apr 2013 · Frontiers in Neural Circuits
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Interactions between oscillators can be investigated with standard tools of time series analysis. However, these methods are insensitive to the directionality of the coupling, i.e., the asymmetry of the interactions. An elegant alternative was proposed by Rosenblum and collaborators [M.] which consists in fitting the empirical phases to a generic model of two weakly coupled phase oscillators. This allows one to obtain the interaction functions defining the coupling and its directionality. A limitation of this approach is that a solution always exists in the least-squares sense, even in the absence of coupling. To preclude spurious results, we propose a three-step protocol: (1) Determine if a statistical dependency exists in the data by evaluating the mutual information of the phases; (2) if so, compute the interaction functions of the oscillators; and (3) validate the empirical oscillator model by comparing the joint probability of the phases obtained from simulating the model with that of the empirical phases. We apply this protocol to a model of two coupled Stuart-Landau oscillators and show that it reliably detects genuine coupling. We also apply this protocol to investigate cardiorespiratory coupling in anesthetized rats. We observe reciprocal coupling between respiration and heartbeat and that the influence of respiration on the heartbeat is generally much stronger than vice versa. In addition, we find that the vagus nerve mediates coupling in both directions.
    Full-text · Dataset · Feb 2013
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Interactions between oscillators can be investigated with standard tools of time series analysis. However, these methods are insensitive to the directionality of the coupling, i.e., the asymmetry of the interactions. An elegant alternative was proposed by Rosenblum and collaborators [M. G. Rosenblum, L. Cimponeriu, A. Bezerianos, A. Patzak, and R. Mrowka, Phys. Rev. E 65, 041909 (2002); M. G. Rosenblum and A. S. Pikovsky, Phys. Rev. E 64, 045202 (2001)] which consists in fitting the empirical phases to a generic model of two weakly coupled phase oscillators. This allows one to obtain the interaction functions defining the coupling and its directionality. A limitation of this approach is that a solution always exists in the least-squares sense, even in the absence of coupling. To preclude spurious results, we propose a three-step protocol: (1) Determine if a statistical dependency exists in the data by evaluating the mutual information of the phases; (2) if so, compute the interaction functions of the oscillators; and (3) validate the empirical oscillator model by comparing the joint probability of the phases obtained from simulating the model with that of the empirical phases. We apply this protocol to a model of two coupled Stuart-Landau oscillators and show that it reliably detects genuine coupling. We also apply this protocol to investigate cardiorespiratory coupling in anesthetized rats. We observe reciprocal coupling between respiration and heartbeat and that the influence of respiration on the heartbeat is generally much stronger than vice versa. In addition, we find that the vagus nerve mediates coupling in both directions.
    Full-text · Article · Feb 2013 · Physical Review E
  • Source
    Nicolaus T Schmandt · Roberto F Galán
    [Show abstract] [Hide abstract]
    ABSTRACT: Markov chains provide realistic models of numerous stochastic processes in nature. We demonstrate that in any Markov chain, the change in occupation number in state $A$ is correlated to the change in occupation number in state $B$ if and only if $A$ and $B$ are directly connected. This implies that if we are only interested in state $A$, fluctuations in $B$ may be replaced with their mean if state $B$ is not directly connected to $A$, which shortens computing time considerably. We show the accuracy and efficacy of our approximation theoretically and in simulations of stochastic ion-channel gating in neurons.
    Full-text · Article · Sep 2012 · Physical Review Letters
  • [Show abstract] [Hide abstract]
    ABSTRACT: Genetic disorders arising from copy number variations in the ERK (extracellular signal-regulated kinase) MAP (mitogen-activated protein) kinases or mutations in their upstream regulators that result in neuro-cardio-facial-cutaneous syndromes are associated with developmental abnormalities, cognitive deficits, and autism. We developed murine models of these disorders by deleting the ERKs at the beginning of neurogenesis and report disrupted cortical progenitor generation and proliferation, which leads to altered cytoarchitecture of the postnatal brain in a gene-dose-dependent manner. We show that these changes are due to ERK-dependent dysregulation of cyclin D1 and p27(Kip1), resulting in cell cycle elongation, favoring neurogenic over self-renewing divisions. The precocious neurogenesis causes premature progenitor pool depletion, altering the number and distribution of pyramidal neurons. Importantly, loss of ERK2 alters the intrinsic excitability of cortical neurons and contributes to perturbations in global network activity. These changes are associated with elevated anxiety and impaired working and hippocampal-dependent memory in these mice. This study provides a novel mechanistic insight into the basis of cortical malformation which may provide a potential link to cognitive deficits in individuals with altered ERK activity.
    No preview · Article · Jun 2012 · The Journal of Neuroscience : The Official Journal of the Society for Neuroscience
  • Roberto Fernández Galán
    [Show abstract] [Hide abstract]
    ABSTRACT: Neurons can respond with highly consistent spike patterns to repetitions of the same stimulus. Analogously, similar neurons receiving a common stimulus can fire highly consistent spike patterns. The former phenomenon is referred to as spike-time reliability, whereas the latter is an example of stochastic synchronization. Both phenomena are quite general and in fact, they also manifest in simplified models of single neuron dynamics, like phase oscillator models, in which the activity of the neuron is determined by its phase-response curve, which in turn is determined by the membrane conductances. Here, we use two measures of spike-time reliability and stochastic synchronization for real neurons and conductance-based models that have been recently introduced in the theory of phase oscillators: the Lyapunov exponent of the oscillator dynamics and the variance of the phase difference between two identical oscillators. Analyzing data from simulations and experiments, we show that, in response to manipulations of membrane conductances, a change of the phase-response curve leading to lower variance of the relative phase is a good predictor of increased spike-time reliability and stochastic synchronization in real and simulated neurons.We also explain why the Lyapunov exponent is not sufficient by itself. Our approach is then exemplified by investigating the effect of certain potassium currents, A and A-like currents, on spike-time precision. Finally, we discuss the biological relevance of our results.
    No preview · Chapter · Jan 2012
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The transient oscillatory model of odor identity encoding seeks to explain how odorants with spatially overlapped patterns of input into primary olfactory networks can be discriminated. This model provides several testable predictions about the distributed nature of subthreshold oscillations and how they control spike timing. To test these predictions, 16 channel electrode arrays were placed within the antennal lobe of the moth Manduca sexta. Unitary spiking and multi site local field potential (LFP) recordings were made during spontaneous activity and in response to repeated presentations of an odor panel. We quantified oscillatory frequency, cross correlations between LFP recording sites, and spike-LFP phase relationships. We show that odor-driven AL oscillations in Manduca are frequency modulating (FM) from ~100–30Hz; this was odorant and stimulus duration dependent. FM oscillatory responses were localized to one or two recording sites suggesting a localized (perhaps glomerular) not distributed source. LFP cross correlations further demonstrated that only a small (r<0.05) distributed and oscillatory component was present. Cross spectral density analysis demonstrated the frequency of these weakly distributed oscillations was state dependent (spontaneous activity= 25-55Hz; odor-driven= 55-85Hz). Surprisingly, vector strength analysis indicated that unitary phase locking of spikes to the LFP was strongest during spontaneous activity and dropped significantly during responses. Application of bicuculline, a GABAA antagonist, significantly lowered the frequency content of odor-driven distributed oscillatory activity. Bicuculline significantly reduced spike phase locking generally, but the ubiquitous pattern of increased phase locking during spontaneous activity persisted. Collectively, these results indicate that oscillations perform poorly as a stimulus-mediated spike synchronizing mechanism for Manduca and hence are incongruent with the transient oscillatory model.
    Full-text · Article · Oct 2011 · Frontiers in Neuroengineering
  • Source
    Dataset: Figure S6
    G. Karl Steinke · Roberto F. Galán
    [Show abstract] [Hide abstract]
    ABSTRACT: Eigenvalue distributions and background power spectrum for networks driven with white and low-pass filtered (colored) noise. The low-pass filter time constant was τ = 10 ms. (A) Uniform distribution of real eigenvalues. (B) Eigenvalue distribution skewed to low negative eigenvalues. (C) Eigenvalue distribution skewed to high negative values. (EPS)
    Preview · Dataset · Oct 2011
  • Source
    Dataset: Figure S2
    G. Karl Steinke · Roberto F. Galán
    [Show abstract] [Hide abstract]
    ABSTRACT: Coherence between pairs of nodes shown in Fig. S1. (A) Low rank node versus low, medium and high rank nodes. (B) Medium rank node versus low, medium and high rank nodes. (C) High rank node versus low, medium and high rank nodes. (EPS)
    Preview · Dataset · Oct 2011

39 Following View all

108 Followers View all