Affective and cognitive prefrontal cortex projections to the lateral habenula in humans.

Karin Vadovičová, Roberto Gasparotti

Neuroradiology, P.le Spedali Civili 1, 25123 Brescia, Italy. email: vadovick@tcd.ie

Abstract

The anterior insula (AI) and dorsal anterior cingulate cortex (dACC) are known to process information about pain,
loss, adversities, bad, harmful or suboptimal choices and consequences that threaten survival or well-being. Also
the pregenual ACC (pgACC) is linked to loss and pain, being activated by sad thoughts and regrets. The lateral
habenula (LHb) is stimulated by predicted and received pain, discomfort, aversive outcome, loss. Its chronic
stimulation makes us feel worse/low and gradually stops us choosing and moving for the suboptimal or punished
choices, by direct and indirect (via rostromedial tegmental nucleus RMTg) inhibition of DRN and VTA/SNc. The
response selectivity of LHb neurons suggests their cortical input from affective and cognitive evaluative regions
that make expectations about bad, unpleasant or suboptimal outcomes. Based on these facts we predicted direct
corticohabenular projections from the dACC, pgACC and AI, as part of the adversity processing circuit that learns
to avoid bad outcomes by suppressing dopamine and serotonin signal. To test this connectivity we used a
Diffusion Tensor Imaging (DTI). We found dACC, pgACC, AI, adjacent caudolateral and lateral OFC projections
to LHb. We predicted no corticohabenular projections from the reward processing regions: the ventral ACC
(vACC) and medial OFC (mOFC) because both respond most strongly to high value stimuli, plus they induce
serotonin and dopamine release, respectively. This lack of LHb projections was confirmed for vACC and likely for
mOFC. The surprising findings were the strong corticohabenular projections from the cognitive prefrontal cortex
(PFC) regions, known for flexible reasoning and planning, that combine whatever information are relevant for
reaching current goals. We propose that the prefrontohabenular projections provide a teaching signal for value-
based choice behaviour, to learn to deselect, avoid or inhibit the potentially harmful, low valued or wrong choices,
goals, strategies, predictions, models and ways of doing things, to prevent bad or suboptimal consequences.

1. Introduction

We examined the cortical input from the affective and cognitive prefrontal regions to the lateral habenula in
humans. We predicted that dACC, pgACC and AI activate the LHb via direct and indirect (ventral striatum)
projections, forming the adversity processing circuit. This circuit biases the learning and behaviour towards
gradual inhibition of punished or suboptimal choices by potentiating the D2 loop of ventral striatum (VS) and by
suppressing dopamine signal. This strengthens the inhibitory avoidance and inhibitory self-control, plus
suppresses motivation and drive to move and work for goals and rewards (Vadovičová and Gasparotti, 2013,
Figure 1). In addition, the overstimulation of LHb causes discomfort and aversion leading to further avoidance
learning, learned helplessness and depression. Support for this comes from rats studies where learned
helplessness correlated with increased metabolic activity in habenula and lateral septum, plus with synaptic
potentiation of LHb neurons (Mirrione et al., 2014, Li et al., 2011).

We expected no habenular projections from the reward processing cortical regions, as the good, rewarding,
valuable choices, appraised by mOFC, tend to move and motivate us to act and go for them, via mOFC input to
the D1 loop of VS, and by inducing dopamine release in VTA. Similarly, when we are doing well, reaching good
outcomes, safety and gains, the vACC lighten up mood and increases well-being and fulfilment by inducing
serotonin release in the brain (likely via bed nucleus of stria terminalis). Thus the connectivity of the reward
versus adversity processing circuit is the cause of their competition, their inverse effects on neuromodulators
control, and their inverse effects on decision making, goal-directed behaviour and well-being (Vadovičová and
Gasparotti, 2013). The activation of LHb by globus pallidus interna (GPi) suppresses dopamine and serotonin
signaling in VTA/SNc and DRN (Jhou et al., 2009, Hong and Jhou, 2011, Christoph et al., 1986, Matsumoto and
Hikosaka, 2009, Wang and Aghajanian, 1977) directly and via the inhibitory RMTg. In rats the LHb possibly
suppresses the wake/arousal and locomotion promoting histamine by efferents to supramamillary nucleus (Kiss
et al., 2002, Vanni-Mercier et al., 1984, Onodera et al., 1994), plus projects to noradrenergic locus coeruleus
(Herkenham and Nauta, 1979). The electrophysiology studies in macaques (Matsumoto and Hikosaka, 2009)
showed that LHb neurons respond to punishment cues, unfavourable outcomes and reward omissions, being
most excited by the most negative of the available outcomes, firing inversely to the VTA/SNc neurons, which are
excited by expectation of valuable outcomes. While the reward processing circuit learns about good - valuable
choices which increase survival, prospects or well-being, the adversity processing circuit learns about potentially
bad, wrong, harmful or unpleasant choices and outcomes, which decrease well-being and survival chances.
Good, interesting things are usually linked to approach and motivation to gain them, while bad things such as
pain, harm and loss induce avoidance and aversion.

Figure 1. The competition between reward and adversity processing circuit in choice behaviour. This model of value-based learning
shows how the affective processing causes selection of good/valuable and de-selection of bad/harmful choices, by controlling dopamine
and serotonin signaling in the brain (Vadovičová and Gasparotti, 2013). The implicit bias/inclination for the 'Go for it' versus 'Stop yourself -
avoid it' response is learned in the motivational ventral striatum (VS), through potentiation of cortical glutamatergic synapses by dopamine
increase (at D1 loop) or its decrease (at D2 loop). Reward and adversity processing circuits are marked green and red. D1 and D2 neurons
in VS are intermixed. The D1 neurons of VS disinhibit dopamine neurons by inhibiting the GABA interneurons in VTA. The indirect D2 loop
(in orange) biases the choice selection toward inhibitory avoidance. The prefrontal projections of mediodorsal thalamus (MDT) enforce the
representations of choices and goals in working memory, as part of cortico-striato-thalamo-cortical loop. Dopamine source VTA is marked
green, serotonin source DRN in yellow. The projections with excitatory effects are brown, with inhibitory effect are grey. Dopamine
attenuates the output of the adversity processing circuit (in red) and potentiates that of the reward processing circuit (in green). Serotonin
attenuates the AI, dACC, LHb, GPi, SNr, STN and motivational D1 loop of VS and enforces vACC, SNc (via SNr), GPe.

This study is based on the affective processing model, which describes the competing reward and adversity
processing circuits, which use value information to bias choice behaviour, learning and affective states
(Vadovičová and Gasparotti, 2013). This model specifies the interaction of cortical regions with the VTA, DRN,
LHb, D1 and D2 loop of ventral striatum, during value-based learning, decision making and goal-directed control
of behaviour. It states that dopamine signaling directs and drives us towards valuable, worthy – good, rewarding,
novel, interesting, useful, relevant and meaningful things, choices. It proposes that dopamine guides us to seek,
choose, prefer, want, desire, engage with, get interested, inclined, even addicted (in love) to them, and to hope,
move, go and work for the valuable, survival and well-being promoting things (food, people, safety, affection,
beauty, goals). While serotonin’s role is to keep our consumption and wanting within the limits of homeostasis,
and to signal when we reached the 'comfort zone'. Thus optimal brain serotonin levels promote well-being,
fulfilment, satisfaction, feeling well, alright, at ease, non-deprived. They attenuate drive, motivation, impulsivity,
motion and effort, calm down worries, aggression, pain, deprivation and slow us down to rest.

The affective evaluations and interpretations in the dACC and AI bias the response selection towards inhibitory
avoidance, by their input to the LHb and ventral striatum. The dACC learns about dangers, pain, negative
feedback, bad consequences and suboptimal outcomes. It predicts and warns us when we are not doing well, to
prevent harm, loss (of resources, loved ones, time) or failure. Its warning signal induces worry, precaution and
alarm state, leading to attention, alertness, mobilization (for fight or flight) in risky, speed or accuracy demanding
situations. This warning signal from dACC urges the prefrontal cortex to switch away from the inadequate or
faulty strategies, to think why things go wrong and find solutions how to change/adjust our world or behaviour to
stop losing or getting harmed. The AI detects and reacts with aversion to bad, inferior or noxious quality of
objects, subjects and social conducts, and also to their moral, conceptual, contextual or task related wrongness.
For the pgACC, active in regret, sorrow and sadness (Drevets et al., 1992, George et al., 1995, Brody et al.,
2001), we predicted similar LHb projections as for dACC and AI, leading to passive avoidance.

To test the proposed corticohabenular projections in humans we used the Diffusion Tensor Imaging (DTI)
probabilistic tractography. This method does not discriminate the afferent from efferent axonal fibres. But because
the tracing studies in animals found only the corticohabenular, no habenulocortical projections, we assumed that
the fibre tracts in our study are the LHb afferents. The medial PFC projections to LHb that regulate dopamine
system were shown already in 1982 (Greatrex and Phillipson, 1982) in rats. A tracing study in macaca fuscata
found dACC but no vACC/BA 25 projections to LHb (Chiba et al., 2001). Frontohabenular projections were also
shown by DTI tractography in humans (Shelton et. al, 2012), which in their Figure seems to originate in the
medial BA 10. The retrograde and anterograde tracing study (Kim and Lee, 2012) in rats found corticohabenular
projections from the AI, cingulate, prelimbic and infralimbic cortex. The infralimbic cortex in rats is a homologue of
vACC in humans, while prelimbic cortex is a homologue of dACC. They found that dense descending projections
terminating in the mediodorsal thalamus (MDT) made en passant and terminal projections to the LHb. The PFC is
reciprocally connected with MDT, forming cognitive and affective cortico-thalamo-cortical loops passing via dorsal
and ventral striatum (Alexander et al., 1986). Proposed functional connectivity of the adversity processing circuit
was supported also by the independent component analysis (ICA) of fMRI data. We found co-activation of dACC,
AI and LHb plus pgACC, PFC, PAG and RMTg regions by 3 different ICA group analyses, both in resting state
and during task (unpublished data). A majority of fMRI studies show robust co-activation of dACC and AI in tasks
involving pain, harm, loss, error, failure, bad or suboptimal outcomes, danger or distress. What these events have
in common is that they are bad/harmful to us and decrease survival chances, so we learn to avoid them by de-
selection of choices leading to bad consequences. This de-selection is done both consciously - by changing our
goals and plans in the PFC, and unconsciously - by probabilistic learning of bad or wrong choices by basal
ganglia, and by inhibition of dopamine and serotonin release in the VTA/SNc by LHb output – what then affects all
brain regions with dopaminergic or serotonergic receptors.

2. Materials and Methods

We used 3Tesla DTI datasets of 18 healthy participants (24-30 years old) obtained from the NKI Rockland
Sample as part of the 1000 Functional Connectomes Project (http://fcon_1000.projects.nitrc.org/indi/pro/nki.html).
The DTI data were acquired with 137 gradient directions, 2mm isotropic voxels, 64 slices and FOV 106 x 90. The
T1 weighted anatomical images were acquired with TR/TE/TI = 2500/3.5/1200ms, FOV 256m, flip angle 8
degrees and 1 mm isotropic voxels.

We used the FMRIB Software Library FSL (http://www.fmrib.ox.ac.uk/fsl/) version 4.1.9 for our DTI analysis, with
a Probtrackx tool (Behrens et al., 2007) for probabilistic tractography. This method generated probabilistic
connectivity distributions for the tested axonal projections for each participant. Each DTI dataset has been
analyzed independently, using standard FSL procedures. The pre-processing steps included head motion
(Jenkinson et al., 2001, 2002) and eddy current correction (Behrens et al., 2003). The BEDPOSTX tool was
applied to calculate the diffusion tensor and other diffusion parameters, to model crossing fibres within each voxel
of the brain. The results were coregistered to the anatomical image and then normalized to the FSL MNI template
(MNI152 2mm). The input seeds for the probabilistic tractography analysis were selected manually in the right
hemisphere of each brain using the anatomical image. The main seed regions of interest were in the AI, dACC
and pgACC. The LHb was the target region. The mOFC and vACC seed regions were selected to test the lack of
input from the reward regions to the LHb. Our vACC seed contained mainly the Brodmann area (BA) 25 thus the
most posterior part of vACC and some voxels anterodorsally adjacent to it. Our mOFC seed contained the
posterior half of the gyrus rectus, to avoid the adjacent ventral BA 10. Additional exploratory seeds sampled the
remaining prefrontal cortex regions, to test the input from each prefrontal area to the LHb. Our LHb seed regions
contained also MHb voxels, as their border is not clearly discriminable in anatomical image. The inclusion of MHb
voxels in our LHb seed region should not affect the prefrontohabenular connectivity results, because the MHb has
no known PFC input.

3. Results

Our probabilistic tractography results confirmed the predicted projections from the AI, dACC and pgACC to the
LHb in all 18 participants (Figure 7 to 14, 17 and 19). In all 18 brains we found that the AI forms functional
processing module with the adjacent caudolateral OFC (clOFC), as they were strongly interconnected and both
projected to LHb and other regions. Possible reason for AI and clOFC interconnectivity is that they process
similar kind of information about bad, harmful, aversive, inferior or suboptimal qualities/attributes of objects,
subjects or conduct. So the AI/clOFC selectively learns about things of low or negative value, which are aversive,
unpleasant, disliked or safer to avoid. This category of stimuli includes also contextual (regarding the current task
or situation), conceptual (false, strange, misfit) and moral wrongness of things and behaviour.

For the reward processing regions, we confirmed the lack of vACC fibre tract to LHb in all 18 participants (Figure
15, 16 and 18). In the mOFC we found no LHb tracts in 15 brains and questionable tracts to the LHb in 3 brains.
The few mOFC tracts were linked to the LHb indirectly: via hypothalamic projections and via frontopolar (BA 10)
projections. Because of the adjacency of the corticohabenular tract from the ventral BA10 with the reciprocal
BA10 to mOFC connections, we could not discriminate these tracts in 3 participants, for which the results were
inconclusive. Our prefrontohabenular connectivity findings were repeated also with 1.5 Tesla scanner, using
same probabilistic tractography analysis in 3 additional participants (Figure 2 to 9 and 20 to 24).

The studied prefrontal fiber tracts to the lateral habenula passed via the internal capsule, basal ganglia and
anteroventral thalamus. The same fiber tracts were found after applying the exclusion masks in the mediodorsal
thalamus and superior colliculus, thus above and under the LHb. The dorso-ventral position of the cortical tracts
crossing the striatum via capsula interna, depended on the vertical position of the individual cortical seed regions.
So, the ventral BA 10 projections crossed striatum between the nucleus accumbens and ventral anterior
putamen, forming horizontal fiber tract to the LHb. But projections from more dorsal regions such as dACC or BA
9 crossed striatum between the lateral caudate nucleus and putamen, forming a diagonal tract. In addition, the AI
and temporal pole reached the LHb by posteriorly localized tracts that branched towards hippocampus, not
passing via capsula interna (Figure 25). The temporal poles are known to process conceptual and semantic
information about meanings of things and characteristics of people. This region is interconnected with BA 10,
dACC and vACC, so receives information about bad and good values of the objects, subjects and concepts.

Unexpected were the strong projections to LHb from the cognitive PFC regions: from seeds in the superior,
middle and inferior frontal gyrus and from the medial and lateral frontal pole or BA 10. We found
prefrontohabenular fiber tracts from the lateral OFC (lOFC), BA 47, 46, 45, 44, BA 8, 9 and 10 (Figure 2 to 6 and
20). So we found corticohabenular fibre tracks from both cognitive and affective PFC regions and AI, involved in
decision making, except from the vACC and possibly except mOFC.

We found also a multisynaptic fibre tract between the septum and medial habenula MHb (Figure 21 to 24 and 26),
passing via the hippocampus  fornix  septum MHb. The septum was connected also with the ventral
anterior thalamus and with the supramammillary nucleus, which stimulates a theta rhythm during exploration of
environment and is connected with the hippocampus via fornix, plus receives also the LHb output. The MHb
receives inputs from cholinergic or substance P neurons of triangular septal nucleus and septofimbral nucleus
(Herkenham and Nauta, 1977), GABAergic medial septum and diagonal band nucleus (Qin and Luo, 2009),
dopaminergic VTA, serotonergic raphe (Herkenham and Nauta, 1977) and noradrenergic locus coeruleus
(Gottesfeld, 1983). MHb projects via the interpeduncular nucleus (Herkenham and Nauta, 1979) to median raphe
nucleus and dorsolateral tegmentum, plus to pineal body in rats (Ronnekleiv and Moller, 1979). this pathway
regulates the sleep cycle, possibly by stimulating the slow wave sleep as the MHb, known for dense mu opioid
receptors, MHb neurons produce sleep inducing interleukin IL-18 (Sugama et al., 2002) and in rats is linked with
pineal gland.

4. Discussion

Our DTI tractography results support the functional connectivity of proposed adversity processing circuit, formed
by the dACC, AI and adjacent clOFC input to the LHb. This circuit detects, learns about and predicts potential
adversities and forwards the information about bad, harmful or suboptimal choices and consequences to the LHb,
to suppress dopamine and serotonin release in the VTA and DRN (Vadovičová and Gasparotti, 2013). Further
evidence for the causal role of this corticohabenular circuit in affective processing is the co-activation of dACC, AI
and LHb during negative feedback (Ullsperger and von Cramon. 2003). Besides the AI, clOFC and dACC, we
found also the pgACC and lateral OFC projections to LHb. Wide evidence from functional, behavioural,
pharmacological and mental disorders studies supports this circuit based model of the affective processing.

In accordance with the affective circuit competition model (Vadovičová, 2013), in which the mOFC and vACC
suppress the LHb via VTA and DRN stimulation, we found no direct vACC and probably only indirect mOFC
projections to LHb. The mOFC fibre tracts were either passing to LHb via hypothalamus, which was clearly not a
direct projection, or via the ventral BA 10. The lateral hypothalamus has known input to the LHb (Herkenham and
Nauta, 1977) and the mOFC is reciprocally connected with the hypothalamus and BA 10. The anterior half of
gyrus rectus belongs to the ventral BA 10 and is anteriorly adjacent to mOFC region. The ventral BA 10 has
strong projections to the LHb. So the questionable mOFC to LHb fibre tracts found in 3 of 18 participants are
likely formed by the mOFC to ventral BA 10 projections, adjacent to separate BA 10 to LHb projections. Our study
supports the opposite effects of the adversity versus reward processing circuit on the activation versus inhibition
of LHb. Many studies found the mOFC response to rewarding or pleasant stimuli and wins versus lateral OFC
response to aversive options, punishment or loss (O'Doherty et al. (2001).

The unexpected findings of this study were the robust projections to LHb from the cognitive PFC regions, known
for flexible coding, combining and holding in working memory any kind of goal-related or relevant information. The
PFC projections were strongest from the frontal pole also known as BA 10, which is interconnected with all PFC
regions plus with the associative cortex in temporal poles and superior temporal gyrus. Thus well informed and
suited for flexible learning, reasoning, planning and goal-directed control of behaviour. The medial BA 10 is
interconnected with the hippocampus and linked to temporal context, introspection, intentions/goals coding,
planning and possibly to hierarchical temporal organisation of our thoughts. The lateral BA 10 or rostrolateral PFC
activated by informational novelty and problem solving is most extended in humans. It probably induces
dopamine release in the medial SNc that stimulates novelty seeking via motivational D1 loop of ventral striatum
and goal-pursuit via D1 loop of medial head of caudate. We propose that lateral BA 10 role is to seek and find out
what is going on – the links between causes and consequences, contingencies, patterns, rules, to make cognitive
predictions about our world, to test them and to apply the right guesses/hypotheses and ways of doing things to
reach our goals. The dorsal PFC is linked to spatial context, spatial organization, planning and control of
behaviour, while the ventrolateral PFC is guiding the behaviour using the meanings, ideas and interrelations
between things, actions and events.

The robust cognitive PFC input to the LHb is possibly linked to the inhibitory self-control and context/goal
dependent de-selection of wrong, irrelevant or inappropriate information, ideas, decisions, plans, strategies or
ways of doing things, depending on current task or goal. All prefrontal regions generate predictions about the
world. The affective regions predict the reward value of choices and consequences, to bias decision making and
goals/intentions formation in the medial BA 10. These goals are then used by all PFC regions to plan and guide
the execution of behaviour to reach goals and avoid harm or loss.

Based on found connectivity and wide literature data we propose that affective prefrontohabenular input inhibits
VTA, leading to potentiation of the D2 loop of ventral striatum, causing inhibitory avoidance, self-control, inhibition
and de-selection of harmful or suboptimal choices. Similarly, the cognitive prefrontohabenular input about
negative feedback serves as a teaching signal that potentiates the D2 loop of head of caudate (via SNc inhibition
and suppression of dopamine release) to gather evidence on what went wrong, failed or was incorrect and to bias
decisions via cortico-striato-thalamo-cortical loop. So the right, correct, valid predictions, ideas, models, strategies
and ways of doing things (to reach our goals), the 'know how', IF-THEN rules or algorithms are learned by the
enforcement of D1 loop of the caudate head, by dopamine increase after the evidence that proved them
right/correct. The evidence for wrongness of the same prediction, hypothesis or way of solving the task is inferred
from the strength of the glutamatergic synapses on D2 loop of caudate head, which summate the experienced
negative outcomes (for example when learning grammar rules in humans or in instrumental learning in animals).
So the evidence ‘for’ versus ‘against’ the validity/correctness of the current prediction, guess, strategy, model or
rule is memorized by the D1 versus D2 loop strength, to learn the probabilities of being right or wrong or doing
things in the right or wrong way (in given context, situation).

In addition the affective and cognitive pefrontal input to LHb causes a de-selection of the non-valuable choices
and information from the working memory by suppressing dopamine release in PFC. So the LHb activation
caused by PFC input may decrease the dopamine signal (from SNc) in PFC towards boring or irrelevant,
nonsignificant information, deselecting the representation of uninteresting information in working memory, to
occupy our attention with more useful information - depending on subjective values, priorities, motivations,
intentions and goals. The useful, relevant, meaningful information induce dopamine release from SNc, to direct
and move us toward interesting and informationally valuable stuff. Support for this idea comes from a TMS study
where stimulation of dorsolateral PFC increased dopamine release in the caudate nucleus, where both DLPFC
and SNc project (Strafella et al., 2001).

Finding cognitive PFC projections directly to LHb means there might be some prefrontal neuronal populations that
are preferentially projecting to the LHb and other preferentially projecting to the SNc, depending on the value,
meaning, significance or usefulness of the incoming information, after sorting the incoming information by their
informational value - their usefulness for current goal or task. By their SNc and LHb efferents, the cognitive PFC
regions bias our learning, selection and de-selection of information depending on their meaning and predictive
value for current goal/aim, task or context.

5. Conclusions

Using DTI probabilistic tractography we confirmed the cortical inputs to Lhb in humans from affective regions: AI,
clOFC, lOFC, dACC and pgACC, linked to inhibitory self-control and avoidance learning. As predicted we found
no LHb projection from vACC. Unexpected were the robust PFC projections to the LHb from the cognitive
prefrontal regions: BA 10 medial and lateral, BA 44, 45, 46, 47, 9 and 8.

Abbreviations

AI anterior insula LHb lateral habenula

BA Brodmann area mOFC medial orbitofrontal cortex

clOFC caudo-lateral orbitofrontal cortex PFC prefrontal cortex

dACC dorsal anterior cingulate cortex pgACC pregenual anterior cingulate cortex

DTI diffusion tensor imaging RMTg rostromedial tegmental nucleaus

fMRI functional magnetic resonance imaging SNc substantia nigra

FSL FMRIB Software Library vACC ventral anterior cingulate cortex

VTA ventral tegmental area

References

Alexander GE, DeLong MR, Strick PL (1986) Parallel organization of functionally segregated circuits linking basal
ganglia and cortex.Annu Rev Neurosci 9:357–381.

Behrens TEJ, Berg HJ, Jbabdi S, Rushworth MF, Woolrich MW (2007) Probabilistic diffusion tractography with
multiple fibre orientations: What can we gain? Neuroimage34(1):144-55.

Behrens TEJ, Woolrich MW, Jenkinson M, Johansen-Berg H, NunesRG,Clare S, Matthews PM, Brady JM, Smith
SM (2003) Characterization and propagation of uncertainty in diffusion-weighted MR imaging. MagnReson Med
50(5):1077-1088.

Brody AL, Saxena S, Stoessel P, et al. Regional brain metabolic changes in patients with major depression
treated with either paroxetine or interpersonal therapy: preliminary findings. Arch Gen Psychiatry. 2001;58:631–
40.

Chiba T, Kayahara T, Nakano K. (2001). Efferent projections of infralimbic and prelimbic areas of the medial
prefrontal cortex in the Japanese monkey, Macacafuscata.Brain Res. 888, 83–101.

Christoph G,Leonzio R, Wilcox K. (1986) Stimulation of the lateral habenula inhibits dopamine-containing neurons
in the substantia nigra and ventral tegmental area of the rat. J Neurosci 6: 613–619.

Drevets, W. C., Spitznagel, E. L., MacLeod, A. K. & Raichle, M. E. Discriminatory capability of PET
measurements of regional blood flow in familial pure depressive disease. Abstr. Soc. Neurosci. 18, 1596 (1992).

FMRIB Software Library (FSL, version 4.1.9) http://www.fmrib.ox.ac.uk/fsl/ Smith et al., 2004, Woolrich et al.,
2009, Jenkinson et al., 2012.

George MS, Ketter TA, Parekh PI, Horwitz B, Herscovitch P, Post RM (1995) Brain activity during transient
sadness and happiness in healthy women. Am. J. Psychiatry, 152 (3) :341–351.

Gottesfeld Z (1983) Origin and distribution of noradrenergic innervation in the habenula: a neurochemical study.
Brain Res. 275:299–304.
Greatrex RM., Phillipson OT (1982) Demonstration of synaptic input from prefrontal cortex to the habenula in the
rat. Brain Res238:192–197.

Herkenham M, Nauta WJ (1979) Efferent connections of the habenular nuclei in the rat. J Comp Neurol 187:19-
47.

Hong S, Hikosaka O (2008)The globus pallidus sends reward-related signals to the lateral habenula. Neuron 60:
720-729.

Hong S, Jhou TC, Smith M, Saleem KS, Hikosaka O (2011) Negative reward signals from the lateral habenula to
dopamine neurons are mediated by rostromedial tegmental nucleus in primates. J Neurosci 31:11457-11471.

Jenkinson M, Bannister PR, Brady JM, and Smith SM (2002) Improved optimisation for the robust and accurate
linear registration and motion correction of brain images.Neuroimage17(2):825-841.

Jenkinson M, Smith SM (2001) A global optimisation method for robust affine registration of brain images. Med
Image Anal 5(2):143-156.

Jenkinson M, Beckmann CF, Behrens TE, Woolrich MW, Smith SM (2012) FSL. NeuroImage 62:782-90.

Jhou TC, Fields HL, Baxter MG, Saper CB, Holland PC (2009) The rostromedial tegmental nucleus (RMTg), a
GABAergic afferent to midbrain dopamine neurons, encodes aversive stimuli and inhibits motor responses.
Neuron 61:786–800.

Kim U, Lee T (2012) Topography of descending projections from anterior insular and medial prefrontal regions to
the lateral habenula of the epithalamus in the rat. Eur J Neurosci 35:1253–1269.

Kiss J, Csáki A, Bokor H, Kocsis K, Kocsis B (2002) Possible glutamatergic/aspartatergic projections to the
supramammillary nucleus and their origins in the rat studied by selective [(3)H]D-aspartate labelling and
immunocytochemistry. Neuroscience 111:671–691.

Li B, Piriz J, Mirrione M, Chung Ch, Proulx Ch D, Schulz D, Henn F, et al. (2011). Synaptic potentiation onto
habenula neurons in the learned helplessness model of depression. Nature. 470(7335):535-539.

Matsumoto M, Hikosaka O (2009) Representation of negative motivational value in the primate lateral habenula.
Nat Neurosci12(1):77-84.

Mirrione M, Schulz D, Kyle A, Lapidus B, Zhang S, Goodman W, Henn FA (2014). Increased metabolic activity in
the septum and habenula during stress is linked to subsequent expression of learned helplessness behavior.
Front Hum Neurosci. 8:00029.

Nathan Kline Institute (NKI) Rockland Sample (http://fcon_1000.projects.nitrc.org/indi/pro/nki.html)

O'Doherty J, Kringelbach ML, Rolls ET, Hornak J, Andrews C (2001) Abstract reward and punishment in the
human orbitofrontal cortex. Nat Neurosci4:95-102.

Onodera K, Yamatodani A, Watanabe T, Wada H (1994) Neuropharmacology of the histaminergic neuron system
in the brain and its relationship with behavioral disorders. Prog Neurobiol 42: 685–702.

Qin, C., and Luo, M. (2009). Neurochemical phenotypes of the afferent and efferent projections of the mouse
medial habenula. Neuroscience 161, 827–837.
Ronnekleiv OK, Moller M (1979) Brain-pineal nervous connections in the rat: an ultrastructure study following
habenular lesion. Exp Brain Res 37:551-562.

Ullsperger M, von Cramon DY (2003) Error monitoring using external feedback: specific roles of the habenular
complex, the reward system, and the cingulate motor area revealed by functional magnetic resonance imaging. J
Neurosci 23:4308–4314.

Shelton L, Pendse G, Maleki N, Moulton EA, Lebel A, Becerra L, Borsook D (2012) Mapping pain activation and
connectivity of the human habenula. J Neurophysiol 107: 2633–2648.

Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TEJ, Johansen-Berg H, Bannister PR, De Luca
M, Drobnjak I, Flitney DE, Niazy R, Saunders J, Vickers J, Zhang Y, De Stefano N, Brady JM, Matthews PM
(2004) Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage
23(S1):208-219.

Strafella AP, Paus T, Barrett J, Dagher A (2001) Repetitive transcranial magnetic stimulation of the human
prefrontal cortex induces dopamine release in the caudate nucleus. J Neurosci 21:157-164.

Sugama S, Cho BP, Baker H, Joh TH, Lucero J, Conti B (2002) Neurons of the superior nucleus of the medial
habenula and ependymal cells express IL-18 in rat CNS. Brain Res. 958, 1–9.

Vadovičová K, Gasparotti R (2013) Functional network for learning about bad, wrong or suboptimal
choices,deeds and consequences, directed towards inhibitory avoidance and self-control. arXiv:1304.4201 [q-
bio.NC], http://arxiv.org/abs/1304.4201

Vanni-Mercier G, Sakai K, Jouvet M (1984) ‘’Waking-state specific” neurons in the caudal hypothalamus of the
cat. C R Acad Sci III 298:195–200.

Wang RY, Aghajanian GK (1977) Physiological evidence for habenula as major link between forebrain and
midbrain raphe. Science197: 89–91.

Woolrich MW, Jbabdi S, Patenaude M, Chappell, Makni S, Behrens T, Beckmann C, Jenkinson M, Smith SM
(2009) Bayesian analysis of neuroimaging data in FSL. NeuroImage, 45:173-186.

Supplementary material. A sample of prefrontal DTI fibre tracts to the LHb.

Figure 2. Frontal pole to LHb fibre tract.

Figure 3. Frontal pole to LHb fibre tract.

Figure 4. rostrolateral PFC (red) and inferior frontal gyrus IFG (blue) fibre tracts to LHb.
Figure 5. ventral BA 10 fibre tract to LHb.

Figure 6. rostrolateral PFC fibre tract to LHb.

Figure 7. pgACC to LHb projection. The pgACC fibre tract to LHb is red, pgACC seed region is orange.
Visible is also the pgACC projection to dACC, which projects to LHb itself.

Figure 8. pgACC to LHb projection. Another slice, with branching towards frontal pole.
Figure 9. anterior insula (AI) and adjacent caudolateral OFC fibre tract to LHb.

Figure 10. anterior insula (AI) and adjacent frontal operculum fibre tract to LHb.
Figure 11. dACC to LHb fibre tract.

Figure 12. dACC to LHb fibre tract.

Figure 13. pgACC and frontal pole fibre tract to LHb.

Figure 14. pgACC and frontal pole fibre tract to LHb, another slices.
Figure 15. vACC to LHb, no fibre tract found. vACC is connected with medial BA 10.

Figure 16. vACC connectivity with medial BA 10 fibre tract.

Figure 17. vACC to dACC projection, plus ventral medial BA 10 projection to LHb.

Figure 18. vACC to dACC, septum and hypothalamus projection. Plus a separate medial BA 10
projection to LHb and hypothalamic projection to MHb.
Figure 19. pgACC to LHb fibre tract. Visible is also the LHb projection to PAG.

Figure 20. Frontal pole to LHb projection.

Figure 21. hippocampus fornix  septum  via anterior thalamus  hypothalamus MHb.

Figure 22. Medial habenula tract. Hippocampus projects to fornix, fornix to septum to anterior
thalamus and hypothalamus, hypothalamus to MHb.
Figure 23. medial habenula tract. same description as in Figure 22.

Figure 24. medial habenula tract. same description as in Figure 22.

Figure 25. temporal pole to LHb fibre tract.

Figure 26. AI/clOFC fibre tract to LHb. Branching from AI to temporal pole and from TP to LHb.
Figure 27. AI/clOFC fibre tract to LHb. Reciprocal connectivity between AI and clOFC via capsula
externa. Branching from AI to temporal pole.

Figure 28. AI/clOFC fibre tract to LHb. AI and clOFC projections to LHb via capsula interna.
Figure 29. Septum to MHb fibre tract.