Refractory Properties of Cochlear Implant-Induced Spiking in Auditory Nerve Fibers areDependent on Location of Stimulation and Voltage-Gated Channel Type Distribution
Jason Boulet1,2 and Ian Bruce1,2,3
Auditory Engineering Laboratory1, McMaster Integrative Neuroscience Discovery & Study2,Department of Electrical & Computer Engineering3, McMaster University, Hamilton, ON, Canada
A L
AUDITORYENGINEERINGLABORATORY
Abstract
Background Experimental work has demonstrated that auditory nerve fibers(ANFs) of cats cannot fully respond to high rates of electrical stimulation, thusreducing the information transfer to the brain. Miller et al. (2001) have shownthat a limiting factor of the reduced spike information transfer can be attributedto the neuron’s refractory period. A computational model of a node of Ranvierof the ANF (Negm and Bruce, 2008) suggested that low-threshold potassium(KLT) and hyperpolarization-activated cyclic nucleotide-gated cation (HCN)channels (Yi et al., 2010) might be responsible for a larger refractory period(Negm and Bruce, in prep.).Methods We extend that work with a simulation study taking into accountANF morphology (Woo et al., 2010) to consider the differential spiking activityas a function of 1) the location of electrical stimulation and 2) nodal channelcomposition at important locations along the ANF. Specifically, we test threeANF models variants: A) only fast Nav and delayed-rectifier Kv at all nodes,B) with the addition of KLT & HCN channels (Yi et al., 2010) at the firstperipheral node and on the nodes of Ranvier neighboring the soma and C) byexpanding the distribution of KLT channels to all nodes (Bortone et al., 2006).Results In general, we observed the absolute refractory period of model Cto be the greatest followed by model B, then by model A. Models B and Ccontrasted with model A by having a greater probability of spike initiation atthe location of stimulation. Model A did not show a strong relative refractoryperiod at its peripheral nodes. We argue that the washout of the relativerefractory period in this region was dependent on the low correlation betweenthe location of the stimulating electrode and the location of spike initiation.Conclusion Preliminary results indicate that model C is most consistent withthe published physiological data. In addition to the KLT & HCN channels ofmodel C, other ion channel types may be necessary to explain all aspects ofrefractory behavior observed in vivo.This research was supported by the Natural Sciences and Engineering ResearchCouncil of Canada (Discovery Grant #261736).
I. INTRODUCTION
I Recent studies have shown that electrically stimulated type I cat ANFsundergo drops in spike rate over the duration of a pulse train for highpulses rates (Zhang et al., 2007). Computational models of the ANF basedon the Hodgkin–Huxley equations containing only Nav and Kv channelsdo not adequately describe these decrements in spike rate.
I Miller et al. (2001) has shown that the duration of refraction can varygreatly across ANFs, which is also difficult to explain with only Nav andKv channels.
I Yi et al. (2010) have experimentally found HCN channels at the firstperipheral node (or terminal) and the nodes neighboring the soma in mousespiral ganglion cells. KLT channels have been localized on ANF axonsentering rat cochlear nucleus (Bortone et al., 2006).
I A computational membrane-node model of the cat ANF incorporatingNav, Kv, KLT and HCN channels has shown that the HCN and KLTchannels can produce increasing spike-rate adaptation and accommoda-tion with increasing stimulation rate (Negm and Bruce, 2008), as well asincreased refractory periods (Negm and Bruce, in prep.).
I We built a compartmental model of the cat ANF to better understandhow refraction depends on the location and populations of voltage-gatedion channel species (Nav, Kv, KLT and HCN) and the location and rateof electrical stimulation from a CI.
I To simulate the activity of various ion channels types, we utilize stochasticion channel models, because the resulting fluctuations in excitability arethought to be behaviorally significant for CI users (Bruce et al., 1999a,b).
II. METHODS: Ion Channel (In-)Activation and Time Constants
(In-)Activation Functions and Time Constants
0
0.25
0.5
0.75
1
Act
ivat
ion
/ Ina
ctiv
atio
n
Particlem∞ Nav Activation
h∞ Nav Inactivationn∞ Kv Activation
w∞ KLT Activationz∞ KLT Inactivation
r∞ s∞ HCN Activation
10−3
10−2
10−1
100
101
102
−150 −100 −50 0 50 100 150Relative Membrane Potential (mV)
Tim
e C
onst
ant (
ms)
Figure 1 : Steady-state activation/inactivation functions and time constants at 37◦C. Weadjusted w∞, z∞, r∞ and s∞ from 22◦C to 37◦C (Cartee, 2000; Rothman and Manis, 2003).
II. METHODS: ANF Models
Compartmental Model
· · ·p1 p2 p3 p4 c1 c2 c3 c4 c23
10 150 150 150 32.6 150 200 250 300 350
1
1.2
2.3
Peripheral axon Central axonSoma unit: µm
Figure 2 : Feline ANF morphology is based on Woo et al. (2010). The soma is myelinated,which contrasts with the mouse and human ANF.
Ion Channel Distribution
0
50
100
0
50
100
0
50
100
AB
C
p1 p2 p3 p4 c1 c2 c3 c4 c5Node of Ranvier
Cha
nnel
Den
sity
(µm
−2)
ChannelNaKKLTHCN
Figure 3 : Hossain et al. (2005) found high densities of Nav1.6 channels located at p1, p4and c1 in the mouse ANF. Yi et al. (2010) have shown HCN channels at the same nodesin mouse spiral ganglion cells. KLT channels have been localized on ANF axons entering ratcochlear nucleus (Bortone et al., 2006).
Circuit Model
· · ·
Ra,(k−2,k−1)
Cm,k−
1
Rm,k−
1
Elea
k,k−
1
Ve,k−1
Ra,(k−1,k)
Cm,k
Rm,k
Elea
k,k
gNa,k
ENa
gK,k
EK
gKLT
,k
EKLT
gh,k
Eh
Ve,k
Ra,(k,k+1)Cm,k+1 Rm,k+1
Elea
k,k+1
Ve,k+1
Ra,(k+1,k+2)
· · ·
Ra,(k+8,k+9)
Cm,k+9 Rm,k+9
Elea
k,k+9
Ve,k+9
Ra,(k+9,k+10)
· · ·
Node of Ranvier1 of 9 Myelin 9 of 9 Myelin
Figure 4 : The circuit model is solved as a discretized (∆t = 2.5 µs) partial differentialequation (Mino et al., 2004) via time-centered Crank–Nicholson, also used in the neuralsimulation environment NEURON (Carnevale and Hines, 2006).
II. METHODS: Ion Channel Simulation
Channel kinetics obey continuous-time discrete-state Markov processes. Thestate transition diagrams for Nav, Kv (Mino et al., 2002), KLT (Negm andBruce, 2008) and HCN1,4 (Liu and Davis, 2012) are shown in Table 1. Redstates indicate fully open states that contribute to conducting ionic current.We simulated the 4 voltage-gated ion channel types with a channel numbertracking algorithm (Chow and White, 1996; Mino et al., 2002).
Table 1 : State transition diagrams
Nav Kv
m0h0
3αm
βm
m1h0
2αm
2βm
m2h0
αm
3βm
m3h0
αh��βh αh��βh αh��βh αh��βh
m0h1
3αm
βm
m1h1
2αm
2βm
m2h1
αm
3βm
m3h1
n0
4αn
βn
n1
3αn
2βn
n2
2αn
3βn
n3
αn
4βn
n4
KLT HCN
w0z0
4αw
βw
w1z0
3αw
2βw
w2z0
2αw
3βw
w3z0
αw
4βw
w4z0
αz��βz αz��βz αz��βz αz��βz αz��βz
w0z1
4αw
βw
w1z1
3αw
2βw
w2z1
2αw
3βw
w3z1
αw
4βw
w4z1
r0
2αr
βr
r1
αr
2βr
r2
s0
αs
βs
s1
II. METHODS: Electrical Stimulation
Monophasic
Biphasic
unit: µs 25 25IPI
I Three distances from the ANF: intracellular, 0.5 and 1 mmI Nine sites of stimulation, over nodes p1 to c5I Monophasic (cathodic) and biphasic (cathodic, then anodic)I Spherical monopolar extracellular electrode radius: 150 µmI Inter Pulse Intervals (IPIs) ranging from 0.3 to 10 msI 1st pulse amplitude: I1 = (1 + 3RS) θ, 2nd pulse: variableI Successfully propagated spike is voltage discriminated at c17
III. RESULTS: Single-Pulse Response
Firing Efficiency (FE)
σ0
0.5
1
340 360 380 400Injected Current (µA)
Firi
ng E
ffici
ency
, FE
Figure 5 : FE is the probability of aspike given a single pulse of current in-put. Data points are the mean valuesfrom 1000 simulation trials and arefit to an integrated Gaussian (Bruceet al., 1999a). The threshold currentθ is the current at FE = 0.5 and therelative spread is RS = σ/θ. Thisexample is the response from an elec-trode placed 0.5 mm above node c3.
Spatial Spiking ResponseMono Bi
c5c4c3c2c1p4p3p2p1
c5c4c3c2c1p4p3p2p1
c5c4c3c2c1p4p3p2p1
AB
C
p1 p2 p3 p4 c1 c2 c3 c4 c5 p1 p2 p3 p4 c1 c2 c3 c4 c5
Site of Stimulation
Spi
ke In
itiat
ion
Nod
e
0.1
0.2
0.3
0.4
0.5
Probabilityof spikeinitiation
0.900
0.925
0.950
0.975
0.7
0.8
0.9
1.0
0.25
0.50
0.75
Intracellular0.5 m
m1 m
m
0 0.25 0.5 0.75 1Firing Efficiency, FE
Spe
arm
an's
ρ StimMonoBi
ModelABC
Figure 6 : (left) Probability of spike initiation from single pulse stimulation. Results weregathered from 1000 simulation trials. Probability is mapped onto a color as function of thestimulated site (x-axis) and the spike initiation node (y -axis). Gray values indicate no spike.This figure shows the response for an electrode 0.5 mm away from the ANF at FE = 0.5.(right) Spearman’s correlation (ρ). This is a summary of the plot on the left, but over allFE’s and distances.
Threshold Current and Relative Spread
7.18 × 10−5
1.34 × 10−4
1.96 × 10−4
2.58 × 10−4
4.04 × 102
5.84 × 102
7.65 × 102
9.45 × 102
2.22 × 103
2.80 × 103
3.37 × 103
3.95 × 103
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
Thr
esho
ld C
urre
nt, θ
(µA
)
0.03
0.04
0.05
0.06
0.07
0.03
0.04
0.05
0.06
0.025
0.030
0.035
0.040
0.045
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
Rel
ativ
e S
prea
d, R
S
StimMonoBi
ModelABC
Figure 7 : (left) Threshold current (θ) and (right) Relative Spread (RS) across all nodesover 1000 trials.
Mean Latency at FE = 0.5
40
60
80
100
120
60
80
100
120
80
100
120
140
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
Initi
atio
n La
tenc
y (µ
s)
400
500
600
400
500
600
400
450
500
550
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
Pro
paga
tion
Del
ay (
µs)
StimMonoBi
ModelABC
Figure 8 : Mean Latency at FE = 0.5 over 1000 trials. (left) Initiation latency and (right)propagation delay. Both components sum to give the time the spike takes to arrive at nodec17.
Jitter at FE = 0.5
10
20
30
10
20
30
40
20
30
40
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
Initi
atio
n Ji
tter
(µs)
10
20
30
10
20
30
40
1020304050
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
Pro
paga
tion
Del
ay J
itter
(µs
)
StimMonoBi
ModelABC
Figure 9 : Jitter at FE = 0.5 for 1000 trials. (left) Initiation latency and (right) propagationdelay jitter.
III. RESULTS: Spike Initiation and Propagation
Relative Membrane PotentialModel A Model B Model C
c23c22c21c20c19c18c17c16c15c14c13c12c11c10c9c8c7c6c5c4c3c2c1p4p3p2p1
c23c22c21c20c19c18c17c16c15c14c13c12c11c10c9c8c7c6c5c4c3c2c1p4p3p2p1
Mono
Bi
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0.5 1 1.5 2Time (ms)
Nod
es o
f Ran
vier
140
120
100
80
60
40
20
0
−20
−40
V (mV)
Figure 10 : One example of a simulation of the relative membrane potential as a function oftime and node of Ranvier along the ANF. For these particular trials, we present the ANF witha stimulus 0.5 mm over node p4, with an IPI of 750 µs and a second-pulse magnitude of 1.5θ.
III. RESULTS: Threshold Ratio (2nd-pulse/1st-pulse)
We fit the threshold data to
θ2/1 =a1 + a2
a1
(1− exp
[−(
IPI−τabsτ1
)])+ a2
(1− exp
[−(
IPI−τabsτ2
)]) (1)
which has been found to fit the threshold ratio data better with two time scales(Negm and Bruce, in prep.) than with one (Miller et al., 2001).
Threshold RatioMono Bi
1.0
1.5
2.0
2.5
0.0 2.5 5.0 7.5 10.0 0.0 2.5 5.0 7.5 10.0Inter Pulse Interval (ms)
Thr
esho
ld C
urre
nt R
atio
, θ2
1
ModelABC
Figure 11 : Second-to-first pulse threshold ratio (θ2/1). Second pulse thresholds are derivedfrom 100 simulation trials fit to an integrated Gaussian function. In these cases above, weused a stimulus 0.5 mm over node c3.
III. RESULTS: Time Scales of Refraction
Relative Refractory Periods
10−510−410−310−210−1100
10−4
10−3
10−2
10−1
10−5
10−4
10−3
10−2
10−1
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
τ 1 (
ms)
10−0.2
10−0.1
100
100.1
100.2
10−0.2
100
100.2
100.4
10−0.2
10−0.1
100
100.1
100.2
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
τ 2 (
ms)
StimMonoBi
ModelABC
Figure 12 : Relative refractory periods (left) τ1 and (right) τ2.
Absolute Refractory Period
0.3
0.6
0.9
1.2
0.2
0.4
0.6
0.8
0.4
0.6
0.8
Intracellular0.5 m
m1 m
m
p1 p2 p3 p4 c1 c2 c3 c4 c5Site of Stimulation
τ abs
(m
s)
StimMonoBi
ModelABC
0.3
0.6
0.9
1.2
Mono BiStim
τ abs
(m
s) ModelABC
Figure 13 : Absolute refractory period τabs (left) arranged by site of stimulation, distance,model and stimulus, whereas (right) we also show the median values collapsed across site ofstimulation and distance.
III. RESULTS: Spike-Rate Adaptation
Post-Stimulus Time-Histograms at FE = 0.5p1 p2 p3 p4 c1 c2 c3 c4 c5
●●
●
●
● ● ●●●
●
●
● ● ●●●
●●
● ● ●
●● ● ● ● ● ●
●●
●
●●
● ●
●
● ●● ● ● ●
●
●●
●● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●●
●
●
●● ●
●●
●
●
● ● ●
●●
●
●
●● ●
●
● ● ●●
● ●●
● ● ● ●● ●
●
●●
● ● ● ●
●
● ● ● ● ● ●●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ●● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●●
●
●
●● ●
●●
●●
●● ●
●●
●
●
●● ●
●●
●● ● ● ●
●
●
●●
●● ●
●●● ● ● ● ●
●
● ● ● ● ● ●●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
●
●
●
●● ●
●●
●
●
●● ●
●●
●
●
●● ●
●
● ●●
●● ●
●
●●
● ● ● ●
●
●
● ● ● ● ●
●
●●
●● ● ●●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
●● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●●
●
●
●● ●
●●
●
●
● ● ●
●
●●
●
● ● ●
●
●●
●●
● ●●
●●
● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●●
●
●● ● ●
●●
●
●
● ● ●
●●
●
●
● ● ●
●
●●
● ● ● ●
●
●●
●● ● ●
●
● ● ● ● ● ●
●
●● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ●● ● ●●
● ● ● ● ● ●
●
● ● ● ● ● ●
●●
●
●
● ● ●
●●
●
●
● ● ●
●●
●
●● ● ●
●
●●
● ●● ●
●
●●
●●
● ●
●
● ● ● ●● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●●
●
●
●● ●
●●
●
●
●● ●
●●
●●
● ● ●
●
●●
●●
● ●
●
● ● ● ●● ●
●
● ●●
●● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●●
●
●
● ● ●●
●●
●
●● ●●●
●
●
● ● ●
●
●● ●
●● ●
●
● ● ●●
● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
●
● ● ● ● ● ●
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.2 kHz
0.8 kHz
2 kHz
5 kHz
0 50 100
150
200 0 50 100
150
200 0 50 100
150
200 0 50 100
150
200 0 50 100
150
200 0 50 100
150
200 0 50 100
150
200 0 50 100
150
200 0 50 100
150
200
Time (ms)
Firi
ng R
ate
(kH
z)
Model●
●
●
ABC
Figure 14 : PSTHs for our model ANFs. The responses are shown for stimulated nodes p1to c5 at the stimulus rates 200, 800, 2000 and 5000 kHz with monophasic stimulation and ata distance of 0.5 mm. The PSTHs were generated by averaging across 100 simulation trialsof 200 ms for two sets of time-bins (Zhang et al., 2007).
IV. CONCLUSIONS
I Single-pulse threshold currents in models C>B>A.I Models B and C show a stronger relationship between where the stimulus
is delivered and where the spike is initiated than model AI Models B and C show the best fits to the two-time scale refractory func-
tion, as evidenced by the unwieldy relative refractory time constants ofmodel A.
I Model C has a significantly larger absolute refractory period than modelsA and B. Therefore, cell-wide distribution of KLT channels plays a majorrole in increasing the absolute refractory period.
I Model A displays evidence of summation for high stimulus rates of 2000and 5000 pulses/s, similarly to Heffer et al. (2010) in guinea pig ANF.
I Further computational studies must be done to address the relative im-pacts of refraction, spike-rate adaptation, accommodation and facilitation(summation) on changes in spike rate over the duration of a pulse trainfor high stimulation rates.
References
Bortone, D. S., Mitchell, K., and Manis, P. B. (2006). Developmental time course of potassium channel expression
in the rat cochlear nucleus. Hearing Research, 211(1-2):114–125.
Bruce, I. C., White, M. W., Irlicht, L. S., O’Leary, S. J., and Clark, G. M. (1999a). The effects of stochastic
neural activity in a model predicting intensity perception with cochlear implants: low-rate stimulation. IEEE
Transactions on Biomedical Engineering, 46(12):1393–1404.
Bruce, I. C., White, M. W., Irlicht, L. S., O’Leary, S. J., Dynes, S., Javel, E., and Clark, G. M. (1999b). A
stochastic model of the electrically stimulated auditory nerve: single-pulse response. IEEE Transactions on
Biomedical Engineering, 46(6):617–629.
Carnevale, N. T. and Hines, M. L. (2006). The NEURON Book. Cambridge University Press.
Cartee, L. A. (2000). Evaluation of a model of the cochlear neural membrane. II: comparison of model and
physiological measures of membrane properties measured in response to intrameatal electrical stimulation.
Hearing Research, 146(1-2):153–166.
Chow, C. and White, J. A. (1996). Spontaneous action potentials due to channel fluctuations. Biophysical
Journal, 71:3013–3021.
Heffer, L. F., Sly, D. J., Fallon, J. B., White, M. W., Shepherd, R. K., and O’Leary, S. J. (2010). Examining the
auditory nerve fiber response to high rate cochlear implant stimulation: chronic sensorineural hearing loss and
facilitation. Journal of Neurophysiology, 104(6):3124–3135.
Hossain, W. A., Antic, S. D., Yang, Y., Rasband, M. N., and Morest, D. K. (2005). Where is the spike
generator of the cochlear nerve? Voltage-gated sodium channels in the mouse cochlea. Journal of Neuroscience,
25(29):6857–6868.
Liu, Q Manis, P. and Davis, R. (2012). Heterogeneous Distribution of Ih and HCN Channels in Murine Spiral
Ganglion Neurons. MidWinter Meeting: Association for Research in Otolaryngology.
Miller, C. A., Abbas, P. J., and Robinson, B. (2001). Response properties of the refractory auditory nerve fiber.
Journal of the Association for Research in Otolaryngology, 2(3):216–232.
Mino, H., Rubinstein, J. T., Miller, C. A., and Abbas, P. J. (2004). Effects of Electrode-to-Fiber Distance
on Temporal Neural Response With Electrical Stimulation. IEEE Transactions on Biomedical Engineering,
51(1):13–20.
Mino, H., Rubinstein, J. T., and White, J. A. (2002). Comparison of algorithms for the simulation of action
potentials with stochastic sodium channels. Annals of Biomedical Engineering, 30(4):578–587.
Negm, M. H. and Bruce, I. C. (2008). Effects of Ih and IKLT on the response of the auditory nerve to electrical
stimulation in a stochastic Hodgkin-Huxley model. Conference proceedings: Annual International Conference
of the IEEE Engineering in Medicine and Biology Society IEEE Engineering in Medicine and Biology Society
Conference, 2008:5539–5542.
Rothman, J. S. and Manis, P. B. (2003). The Roles Potassium Currents Play in Regulating the Electrical Activity
of Ventral Cochlear Nucleus Neurons. Journal of Neurophysiology, 89(6):3097–3113.
Woo, J., Miller, C. A., and Abbas, P. J. (2010). The dependence of auditory nerve rate adaptation on electric
stimulus parameters, electrode position, and fiber diameter: a computer model study. Journal of the Association
for Research in Otolaryngology, 11(2):283–296.
Yi, E., Roux, I., and Glowatzki, E. (2010). Dendritic HCN Channels Shape Excitatory Postsynaptic Potentials at
the Inner Hair Cell Afferent Synapse in the Mammalian Cochlea. Journal of Neurophysiology, 103(5):2532–2543.
Zhang, F., Miller, C. A., Robinson, B. K., Abbas, P. J., and Hu, N. (2007). Changes across time in spike rate
and spike amplitude of auditory nerve fibers stimulated by electric pulse trains. Journal of the Association for
Research in Otolaryngology, 8(3):356–372.
jasonboulet.files.wordpress.com/2013/02/aro2013poster2.pdf [email protected]