Serotonin and sudden unexpected death in epilepsy

Abstract

Epilepsy is a highly prevalent disease characterized by recurrent, spontaneous seizures. Approximately one-third of epilepsy patients will not achieve seizure freedom with medical management and become refractory to conventional treatments. These patients are at greatest risk for sudden unexpected death in epilepsy (SUDEP). The exact etiology of SUDEP is unknown, but a combination of respiratory, cardiac, neuronal electrographic dysfunction, and arousal impairment is thought to underlie SUDEP. Serotonin (5-HT) is involved in regulation of breathing, sleep/wake states, arousal, and seizure modulation and has been implicated in the pathophysiology of SUDEP. This review explores the current state of the understanding of the relationship between 5-HT, epilepsy, and respiratory and autonomic control processes relevant to SUDEP in epilepsy patients and in animal models.

Introduction

The epilepsies are a heterogeneous group of diseases in which patients have unpredictable, spontaneous seizures (Fisher et al., 2017). Despite the existence of dozens of therapies to treat seizures, approximately 35% of patients will not achieve seizure freedom (Chen et al., 2018). These patients are at greatest risk for sudden unexpected death in epilepsy (SUDEP) (Nashef et al., 2007). SUDEP is the leading cause of death in patients with medically refractory epilepsy. It is second only to stroke in years of potential life lost to neurological disease, and thus poses a major public health burden (Thurman et al., 2014). The exact etiology of SUDEP is unknown. Studies in patients and animal models suggest that cardiorespiratory dysregulation and impairment of arousal contributes to SUDEP (Figure 1). Because of its role in influencing seizures, and in regulating sleep and wakefulness, arousal, circadian rhythms, breathing, and cardiac activity, serotonin (5-HT) has been implicated in the pathophysiology of SUDEP. There is large body of evidence to support this. Studies in epilepsy patients have focused on the effect of 5-HT enhancing drugs on seizure semiology, frequency, and autonomic function, whereas studies in animal models have focused on the effect of stimulating 5-HT networks and focal or systemic administration of serotonergic drugs on seizure development and death. In general, increases in 5-HT may be protective against seizures and SUDEP. Conversely, seizures and epilepsy pathologies may reduce 5-HT tone and increase risk of both seizures and SUDEP (Bagdy et al., 2007; Richerson et al., 2011). This review delves into the interplay between 5-HT, epilepsy, and SUDEP in epilepsy patients and animal models.

Figure 1.

Seizures can provoke respiratory dysfunction and blood gas derangements in epilepsy patients and animal models. (A) Oxygen saturation and ETCO₂ traces from an epilepsy patient during a complex partial left temporal onset seizure without secondary generalization. Arrowheads depict duration of the seizure. Vertical lines indicate the start and end of apnea. Redrawn with permission from Seyal and Bateman 2009. (B) Terminal apnea precedes terminal asystole in monitored SUDEP cases in human epilepsy patients. Used with permission from Ryvlin et al. 2013. (C) Terminal apnea precedes terminal asystole in mice that succumb to MES induced seizures. Five min EEG, plethysmography, and ECG traces from an animal that died following a MES seizure (50 mA, 60 Hz, 200 ms). Magnified recordings taken (i) before, (ii) during, and (iii) after the seizure. Used with permission from Buchanan 2014.

5-HT in epilepsy and seizures

Effects of seizures on 5-HT

Seizure-induced alterations in 5-HT have been observed in epilepsy patients and in animal models. For example, temporal lobe epilepsy (TLE) patients exhibit decreased binding to 5-HT_1A receptors within the raphe, thalamus, amygdala, neocortex, orbitofrontal cortex, fusiform gyrus, anterior cingulate cortex, and hippocampus (Assem-Hilger et al., 2010; Hasler et al., 2007; Henry et al., 2013; Martinez et al., 2013; Merlet et al., 2004; Savic et al., 2004; Toczek et al., 2003). However, increased 5-HT_1A receptor density within the hippocampus is associated with longer disease duration in TLE patients (da Fonseca et al., 2017; Schlicker et al., 1996). A decrease in hippocampal 5-HT levels in TLE patients may account for this increase in receptor density but decrease in receptor binding (da Fonseca et al., 2015).

Studies in epilepsy patients have also observed seizure-induced alterations in expression of the 5-HT transporter (5-HTT) genes. 5-HTT is responsible for regulating 5-HT availability in the synaptic terminal. 5-HTT genes are less transcriptionally efficient in epilepsy patients, potentially contributing to reduced 5-HT reuptake (Esmail et al., 2015; Schenkel et al., 2011). Similarly, a reduction in 5-HTT activity in the insular cortex is observed in depressed TLE patients compared to those without depression (Martinez et al., 2013). 5-HTT binding is also reduced within the neocortex of post-mortem samples from TLE patients (Rocha et al., 2007). Changes in 5-HTT efficiency may represent an endogenous compensatory mechanism to account for seizure-induced decreases in 5-HT or 5-HT receptor binding by prolonging synaptic 5-HT availability (Martinez et al., 2013). Given that polymorphisms in the 5-HTT gene SLC6A4 may be associated with pharmaco-resistance in TLE patients, 5-HTT abnormalities may contribute to TLE pathology (Kauffman et al., 2009). However, it is unclear whether these abnormalities predispose patients to seizures or whether the seizures themselves cause changes in gene expression that lead to these polymorphisms.

Seizures may also influence levels of 5-HT metabolites. 5-hydroxyindoleacetic acid (5-HIAA), a major 5-HT metabolite, is decreased in the cerebrospinal fluid (CSF) of adults with progressive myoclonic epilepsy and children with febrile convulsions (Giroud et al., 1990; Pranzatelli et al., 1995). Levels of 5-HIAA are increased in the CSF of epilepsy patients treated with certain anti-epileptic drugs (AEDs) (Reynolds et al., 1975). Pediatric epilepsy patients also exhibit decreased concentrations of tryptophan, an amino acid essential for 5-HT production, within blood plasma and CSF (Ko et al., 1993; Marion et al., 1985). These findings may extend to adults, as polymorphisms in the tryptophan hydroxylase 2 (TPH2) gene, which encodes the major biosynthetic enzyme for 5-HT in the brain, are associated with psychiatric disorders in TLE patients and may contribute to TLE pathology (Bragatti et al., 2014; Gao et al., 2012).

Changes in 5-HT metabolite levels and receptor expression are also observed in animal models. Many genetic animal epilepsy models exhibit alterations in 5-HT receptor expression. 5-HT concentration, synaptosomal 5-HT uptake, and tryptophan hydroxylase (TpOH) activity is reduced in the genetically epilepsy prone rat (GEPR), a model of myoclonic epilepsy (Faingold, 1988) that exhibits acoustically evoked seizures (Dailey et al., 1989; Jobe et al., 1986; Jobe et al., 1982; Statnick et al., 1996). Likewise, EL (also called El or E1) mice (Imaizumi et al., 1959; Suzuki, 2004), which experience convulsive seizures upon ‘throwing’ stimulations (repeated ~10 cm vertical tosses into the air), have abnormal 5-HT levels and [³H]5-HT binding within the cortex and brainstem (Hiramatsu, 1981, 1983; Suzuki, 2013). This 5-HT deficit occurs as a direct result of seizures, as stimulated (‘thrown’) EL mice exhibit lower interictal 5-HT levels than non-stimulated EL mice.

DBA/2 mice, which are susceptible to audiogenic seizures, seizure-induced respiratory arrest (S-IRA), and death (Tupal et al., 2006), show reduced expression of 5-HT_2C, 5-HT₃, and 5-HT₄ in the medulla (Feng et al., 2017; Uteshev et al., 2010). Similarly, 5-HT_2B, 5-HT_2C, and 5-HT_3B expression is reduced within the brainstem of DBA/1 mice, which are also susceptible death following audiogenic seizures (Faingold et al., 2011a). Consistent with a reduction in 5-HT receptor expression influencing seizure susceptibility, 5-HT_2C knock-out mice have increased spontaneous death from seizures (Brennan et al., 1997). As aforementioned in epilepsy patients, it is unclear in many models whether seizures precipitate these changes in 5-HT or whether their genetics produces changes in 5-HT that contribute to seizures.

Chemical seizure induction can likewise induce changes in 5-HT. Rats subjected to the pilocarpine-status epilepticus epilepsy model exhibit decreased 5-HT and 5-HIAA levels within the hippocampus during the chronic phase (Lin et al., 2013; Luna-Munguia et al., 2019). This may be due to seizure-induced reduction of the number of TpOH-positive cells within the raphe nuclei, where most 5-HT is produced (Lin et al., 2013), and may explain reduced hippocampal response to DRN stimulation in this model (Mazarati et al., 2008). Moreover, there is reduction in 5-HT₇ expression in rats administered pilocarpine (Cavalheiro, 1995; Yang et al., 2012). In these models, it is more likely that the seizures directly produce changes in 5-HT metabolite and receptor expression.

Effects of 5-HT on seizures

Previous findings indicate 5-HT deficits are present in TLE patients. Conversely, co-administration of drugs that increase available 5-HT, such as selective serotonin reuptake inhibitors (SSRIs), may improve seizure control in epilepsy patients (Specchio et al., 2004). In epilepsy patients experiencing focal impaired awareness (formerly complex partial) seizures and focal to bilateral tonic clonic (formerly secondarily generalized) seizures (Fisher et al., 2017), adjunctive fluoxetine or citalopram both reduce seizure frequency (Albano et al., 2006; Favale et al., 2003; Favale et al., 1995). Fluoxetine and fenfluramine are also well-tolerated and reduce seizure frequency in patients with Dravet syndrome, a severe form of epilepsy with a high rate of SUDEP (Figure 2) (Ceulemans et al., 2012; Ceulemans et al., 2016; Meador, 2014; Schoonjans et al., 2017). These observations are consistent with trends seen in animal models.

Figure 2.

Increasing 5-HT tone reduces SUDEP risk in animal models and may protect against high-risk seizures in epilepsy patients. (A) Pretreatment with saline (n =21) or 15 (n = 9), 20 (n = 11), or 40 (n = 6) mg/kg fenfluramine 30 min before induction of audiogenic seizures reduces seizure-induced respiratory arrest in DBA/1 mice. *, p < 0.05; **, p < 0.01; #, p < 0.001. Redrawn with permission from Tupal and Faingold 2019. (B) Acute administration of 15 (n = 10), 20 (n = 12), 25 (n = 15) mg/kg fluoxetine 30 min before induction of audiogenic seizures reduces of S-IRA in DBA/2 mice. *, p < 0.05; #, p < 0.005. Redrawn with permission from Tupal and Faingold 2006. (C) Survival following a MES seizure (50 mA Lmx1b^f/f; 30 mA Lmx1b^f/f/p) in Lmx1b^f/f and Lmx1b^f/f/p is increased by mechanical ventilation or pretreatment with DOI (0.3 mg/kg Lmx1b^f/f; 1 mg/kg Lmx1b^f/f/p), citalopram (20 mg/kg), TCB-2 (10 mg/kg), or MK-212 (10 mg/kg) 30 min prior to seizure induction. n = 7 per drug/genotype. *, p < 0.001. Redrawn with permission from Buchanan et al. 2014. (D) Adjunctive fenfluramine reduces frequency of convulsive seizures in Dravet syndrome patients. Mo., month; Baseline, n = 9; 3 mo., n = 9; 6 mo., n = 8; 9 mo., n = 7; 1 yr., n = 6. *, p < 0.05. Redrawn with permission from Schoojans et al. 2017.

Early studies in inbred and wild type mouse strains demonstrate that strains displaying the highest levels of whole brain biogenic amines experience greater latency to maximal electroshock (MES) induced seizures (Scudder et al., 1966). Elevated intracranial 5-HT concentrations protect against pentylenetetrazol (PTZ) induced seizures in C57BL/6 and DBA/2 mice, and against audiogenic seizures in DBA/2 mice (Schlesinger et al., 1969). Conversely, decreasing 5-HT levels appears to increase seizure susceptibility. Reserpine, which reduces 5-HT levels by interfering with vesicular monoamine transporters (Giarman et al., 1964), reduces whole brain 5-HT and increases susceptibility to minimal electroshock seizures in rats (Wenger et al., 1973). This effect is prolonged with selective inhibition of 5-HT synthesis (Wenger et al., 1973). 5-HT receptor blockade with cyproheptadine, or inhibition of 5-HT synthesis with para-chloroamphetamine (PCPA) similarly decreases electroshock seizure threshold (Buterbaugh, 1978; Kilian et al., 1973).

The pro-convulsant effects of 5-HT depletion in animal models can be prevented with 5-HT enhancement (Buterbaugh, 1978; Kilian et al., 1973; Przegaliński, 1985; Truscott, 1975). Administration of 5-hydroxytryptophan (5-HTP), a 5-HT precursor, elevates whole brain levels of 5-HT and reverses reserpine induced enhancement of audiogenic seizures in DBA/2 mice (Boggan et al., 1973). 5-HTP pretreatment increases MES seizure threshold in rats (Kilian et al., 1973). This trend is recapitulated in more recent studies. For instance, 5-HTP administration reduces S-IRA following acoustic seizures in DBA/1 mice (Zhang et al., 2016). In a subset of mice 5-HTP also reduced the incidence of the tonic seizures associated with S-IRA. Effects on S-IRA by 5-HTP may be due to direct effects on breathing and/or seizure protection (Buchanan et al., 2016).

Furthermore, pharmacologically increasing extracellular 5-HT with SSRIs and other antidepressants has anticonvulsant effects across a variety of animal models (Table 1, Figure 2) (Castano-Monsalve, 2013; Igelstrom, 2012; Jobe et al., 2005; Jobe et al., 1999; Johannessen Landmark et al., 2016; Kanner, 2013, 2016; Kondziella et al., 2009; Li et al., 2018; Maguire et al., 2014; Payandemehr et al., 2012; Santos Junior et al., 2002; Vermoesen et al., 2011).

Table 1.

Review of literature supporting anticonvulsant effect of 5-HT enhancing drugs in electrical, chemical, and genetic animal models of epilepsy.

MODEL	SPECIES	DRUG	DOSE (mg/kg)	STUDY PROTOCOL	SEIZURE OUTCOME MEASURE	REFERENCE
PTZ	Zebrafish	CIT, REB, BUP	10–300 μM	1 min before sz induction	Seizure threshold	Vermoesen et al. (2011)
PTZ	Mouse	REB, BUP	5–10; 10–40	30 min before sz induction	Seizure threshold	Vermoesen et al. (2011)
PTZ	Mouse	FLX	20	60 min before sz induction	Survival	Kecskemeti et al. (2005)
PTZ	Mouse	FLX	1–20	30 min before sz induction	Seizure threshold, latency	Ugale et al. (2004)
PTZ	Mouse	FLX	10, 15	30 min before sz induction	Seizure threshold, susceptibility	Borowicz et al. (2012)
PTZ	Mouse	HBK-14	20	30 min before sz induction	Seizure threshold	Pytka et al. (2017)
PTZ	Mouse	SRT	10	Drug for 14 days before sz induction	Seizure threshold	Heydari and Davoudi (2017)
PTZ	Rat	TNP	20–80	30 min before sz induction	Seizure latency, frequency, severity	Reeta et al. (2016)
PTZ	Mouse	SR 57227	1–10	30 min before sz induction	Seizure latency, severity, mortality	Li et al. (2014)
PTZ	Mouse	TNP	2.5–80	30 min before sz induction	Seizure onset	Uzbay et al. (2007)
PTZ	Rat	VEN	25–50	30 min before sz induction	Seizure severity and latency	Santos et al. (2002)
PTZ	Mouse	mCPBG, CIT	1–10; 0.1–50	30 min before sz induction	Seizure threshold	Payandemehr et al. (2012)
PTZ	Mouse	APZ	0.5–8	30, 60, 120 min before sz induction	Seizure latency	Shafaroodi et al. (2015)
PTZ/DBA/2	Mouse	5-HT	800 μg	Intracranial injection 10 min before sz induction	Seizure severity	Schlesinger et al. (1969)
PTZ	Mouse	REB	4–15	30 min before sz induction	Seizure threshold	Poplawska et al. (2015)
PTZ (kindling)	Mouse	FLV	5–20	30 min before sz induction	Seizure severity, brain dmg	Alhaj et al. (2015)
Pilocarpine	Zebrafish	CIT	10–300 μM	1 min before sz induction	Seizure threshold	Vermoesen et al. (2011)
Pilocarpine	Rat	FLX	20	Drug for 5 days after sz induction	Seizure frequency	Hernandez et al. (2002)
Pilocarpine	Rat	FLX	20	Drug for 10 days once epileptic	Seizure threshold	Mazarati et al. (2008)
Pilocarpine	Rat	PAR	5	Drug for 4 days after sz induction	Seizure frequency, severity, brain dmg	Lin et al. (2017)
KA	Mouse	CIT	10	Drug for 7 days before sz induction	Seizure severity	Jaako et al. (2011)
KA	Rat	BUP	10, 50	30 min before sz induction	Seizure latency, severity, brain dmg	Lin et al. (2013)
KA	Rat	CIT, REB	15, 20–30	Drug for 4 days once epileptic	Seizure frequency, duration	Vermoesen et al. (2012)
MES	Mouse	APZ	1–2	60 min before sz induction	Seizure mortality, tonic phase duration	Shafaroodi et al. (2015)
MES	Mouse	BUP	15–30	30 min before sz induction	Seizure threshold	Tutka et al. (2004)
MES	Mouse	REB, FLX, CIT, DLX	0.1–30; 10; 20; 10	30 min before sz induction	Seizure mortality	Kruse et al. (2019)
MES	Mouse	VEN	12.5, 25	30 min before sz induction	Seizure threshold	Borowicz et al. (2011)
MES	Mouse	VEN	12.5, 25	Drug for 14 days before sz induction	Seizure threshold	Borowicz et al. (2011)
MES	Mouse	FLX	2.5, 5, 10	30 min before sz induction	AED ED50	Leander (1992)
MES	Rat	FLX, FEN	10, 15	60 min before sz induction	Hind limb extension	Buterbaugh (1978)
MES	Mouse	DOI, CIT, TCB-2, MK-212	0.3, 1; 20; 10; 10	30 min before sz induction	Seizure mortality	Buchanan et al. (2014)
MES	Mouse	HBK-14, HBK-15	20	30 min before sz induction	Seizure threshold	Pytka et al. (2017)
MES	Mouse	FLX	10	60 min before sz induction	Seizure threshold	Raju et al. (1999)
MES/PTZ	Mouse/Rat	5-HTP	100	30 min before sz induction	Seizure threshold	Killian and Frey (1973)
MES	Mouse	REB	8–16	30 min before sz induction	Seizure threshold	Borowicz et al. (2014)
6 Hz Corneal Stimulation	Mouse	HBK-14, HBK-15	20	30 min before sz induction	Seizure threshold	Pytka et al. (2017)
Ear shock	Mouse	FLX	15–25	30 min before sz induction	Seizure threshold	Borowicz et al. (2006)
Hippocampal kindling	Rat	FLX	10 nmol	15 min before sz induction	Electrical threshold	Wada et al. (1993)
Hippocampal kindling	Rat	5-HTP	50	60 min before sz induction	Electrical threshold	Cavalheiro et al. (1981)
Hippocampal kindling	Rat	FLX	10	Drug for 21 days before sz induction	Electrical threshold	Wada et al. (1995)
Amygdala kindling	Cat	FLX	2–10	1–49 hr before seizure induction	Electrical threshold	Siegal and Murphy (1979)
Amygdala kindling	Mouse	FLX	10	Osmotic mini pump, 30 day administration	Afterdischarge threshold	Li et al. (2018)
GEPR-3 and GEPR-9	Rat	FLX	30	1–5 hr before seizure induction	Seizure severity	Dailey et al. (1992)
GEPR-3 and GEPR-9	Rat	FLX	7–20	Sz induced every 7 days for 28 days	Seizure threshold	Dailey et al. (1992)
GEPR-9	Rat	FLX	15	60 min before sz induction	Severity and latency	Yan et al. (1994)
El mice	Mouse	CIT	0.01, 0.02, 0.04	Drug for 14 days before sz induction	Seizure threshold	Kabuto et al. (1994)
El mice	Mouse	FLX	10	Drug for 3 or 7 days before sz induction	Seizure threshold	Richman and Heinrichs (2007)
DBA/1 and PTZ	Mouse	5-HTP	100–150	Drug for 2 days before sz induction	S-IRA, seizure incidence	Zhang et al. (2015)
DBA/1	Mouse	FLX	30	30 min before sz induction	S-IRA	Zeng et al. (2015)
DBA/1	Mouse	SRT	40–75	30 min before sz induction	S-IRA, seizure incidence	Faingold and Randall (2013)
DBA/1	Mouse	FEN	10–40	30 min, 12 hr, 24 hr, or 72 hr before sz induction	Seizure severity, ED50, S-IRA	Tupal and Faingold (2019)
DBA/2	Mouse	FLX	15–25	30 min or 72 hr before sz induction	S-IRA, seizure incidence	Tupal and Faingold (2006)
DBA/1	Mouse	SR 57227, FLX	20–40; 40	30 min before sz induction	S-IRA, seizure incidence	Faingold et al. (2016)
DBA/1	Mouse	FLX	15–70 acute, 20 chronic	30 min before or for 5 days before sz induction	S-IRA, seizure incidence	Faingold et al. (2011)
DBA/1	Mouse	FLV, PAR,VEN	50–80; 75–100; 25–100	30, 60, 120 min prior to sz induction	S-IRA, seizure incidence	Faingold et al. (2014)
Audiogenic Sz	Rat	FLX	25	60 min before sz induction	Seizure severity	Ismaylova and Madzhidi (2012)
Audiogenic Sz	Rat	FLX	20	Drug for 14 days before sz induction	Seizure severity	Sarkissova et al. (2016)
WAG/Rij Absence Sz	Rat	FLX, DLX	30; 10–30	Drug administration for 7 weeks	Seizure and immobility duration	Citraro et al. (2015)
Flurothyl	Mouse	REB	20	Drug for 21 days before sz induction	Seizure latency	Ahern et al. (2006)
Focal pilocarpine	Rat	CIT	1 μM	4 hr intracranial infusion before sz induction	Seizure severity	Clinckers et al. (2004)
Focal bicuculline	Rat	FLX	5–20	60 min before sz induction	Seizure frequency, severity	Prendiville and Gale (1993)
Focal bicuculline	Rat	FLX	1.75–7 nM	15 min before sz induction	Seizure threshold	Pasini et al. (1992)
Picrotoxin	Mouse	FLX	20	65 min before sz induction	Seizure latency	Pericic et al. (2005)
Picrotoxin	Mouse	FLX	20	Drug for 5 days before sz induction	Seizure latency	Pericic et al. (2005)
4-AP	Rat	FLX	10	Drug for 7 days before sz induction	Seizure latency, brain dmg	Shiha et al. (2017)
4-AP	Rat	SRT	2.5, 25	4 hr before sz induction	Seizure frequency, EEG changes	Sitges et al. (2012)
4-AP	Rat	SRT	0.75	Drug for 7 days before sz induction	Seizure frequency, EEG changes	Sitges et al. (2012)
4-AP	Rat	SRT	0.75, 2.5	4 hr before sz induction	Seizure suppression	Sitges et al. (2016)

BUP, bupropion; CIT, citalopram; FEN, fenfluramine; FLV, fluvoxamine; FLX, fluoxetine; GEPR, genetically epilepsy prone rats; hr, hour; KA, kainic acid; MAP, maprotiline; min, minute; PAR, paroxetine; REB, reboxetine; SRT, sertraline; sz, seizure; VEN, venlafaxine; CLM, clemizole.

In the PTZ seizure model, venlafaxine and tianeptine increase latency to convulsions and reduce seizure severity (Reeta et al., 2016; Santos Junior et al., 2002; Uzbay et al., 2007). Fluoxetine administration similarly increases seizure latency and the seizure survival rate of PTZ induced seizures (Kecskemeti et al., 2005; Ugale et al., 2004). It has been observed that concomitant fluoxetine and AED (phenytoin, carbamazepine, ameltolide) administration may be synergistic. Fluoxetine may enhance the effect of AEDS and protect against MES and PTZ seizures in mice (Borowicz et al., 2012; Borowicz et al., 2006; Leander, 1992). Fluoxetine also reduces the ED₅₀ of phenytoin, carbamazepine, and ameltolide against MES seizures by half (Leander, 1992). Acutely enhancing 5-HT tone with other drugs such as reboxetine, citalopram, and the antidepressant bupropion also increases PTZ seizure threshold in mice (Payandemehr et al., 2012; Poplawska et al., 2015; Vermoesen et al., 2011). A similar trend is noted in zebrafish, with acute citalopram, reboxetine, and bupropion reducing behavioral PTZ seizure threshold (Vermoesen et al., 2011). Chronic 5-HT enhancement may also be seizure protective, as chronic administration of the SSRI sertraline increases PTZ seizure threshold in mice (Heydari et al., 2017).

Increases in 5-HT are anticonvulsant in several other chemical seizure models as well (Table 1). Chronic fluoxetine or paroxetine reduces seizure frequency and severity in the pilocarpine seizure model whereas 5-HT depletion increases seizure susceptibility and frequency (Hernandez et al., 2002; Lin et al., 2017; Trindade-Filho et al., 2008). Chronically increasing 5-HT tone with citalopram or reboxetine likewise reduces kainic acid seizure frequency and severity (Jaako et al., 2011; Vermoesen et al., 2012). This is supported by the observation that mutant En1^Cre/+;Otx2^flox/flox mice, which express 5-HT hyperinnervation in ventral midbrain, hippocampal CA3 region, and cerebral cortex, are resistant to kainic acid-induced seizures (Tripathi et al., 2015).

Genetic animal models of epilepsy also exhibit reduced seizure propensity following increases in 5-HT (Table 1). Fluoxetine and duloxetine reduce absence seizure and immobility duration in the WAG/Rij rat absence seizure model (Citraro et al., 2015). Chronic fluoxetine and citalopram administration reduces seizure frequency in EL mice (Kabuto et al., 1994; Richman et al., 2007). Similarly, audiogenic seizures are suppressed in seizure-susceptible Wistar rats administered either acute or chronic fluoxetine (Ismailova et al., 2012; Sarkissova et al., 2016). 5-HT enhancement also has anticonvulsant effects on audiogenic seizures in GEPR-3 and −9 rats and this effect is antagonized by PCPA 5-HT depletion (Browning et al., 1997; Dailey et al., 1996; Dailey et al., 1992; Yan et al., 1994).

Observations in epilepsy patients and electrical, chemical, and genetic animal models consistently demonstrate that increases in 5-HT are seizure protective and that blockade of 5-HT binding or decreases in 5-HT are proconvulsant. However, the neural mechanisms underlying this effect are not clear. Future research aimed at receptor mechanisms and underlying circuitry will be critical to discerning how 5-HT exerts its anticonvulsant effects,

Effect of anti-epileptic drugs on 5-HT

5-HT may contribute to the anticonvulsant mechanism of some AEDs (Bagdy et al., 2007; Dailey et al., 1996; Griffin et al., 2017). For instance, carbamazepine increases free blood plasma tryptophan concentrations, whereas diphenylhydantoin decreases tryptophan concentrations in epilepsy patients, although both are anticonvulsants (Pratt et al., 1984).

Bonnycastle et al. (1957) were among the first to report that some AEDs (e.g., phenytoin, methion) can increase levels of 5-HT within the rat brain (Bonnycastle et al., 1957). Carbamazepine, antiepilepsirine, and loreclezole increase 5-HT within the hippocampus, thalamus, and striatum of GEPRs (Ahmad et al., 2005; Dailey et al., 1997a; Dailey et al., 1996; Okada et al., 1992). Anticonvulsant doses of carbamazepine similarly produce hippocampal 5-HT release in non-epileptic rats (Dailey et al., 1998; Dailey et al., 1997b; Kaneko et al., 1993). At high dosages, carbamazepine promotes prolonged synaptic 5-HT by blocking uptake of 5-HT into hippocampal synaptosomes which may contribute to increased 5-HT levels (Dailey et al., 1998). 5-HT depletion decreases the anticonvulsant effect of both carbamazepine and antiepilepsirine (Dailey et al., 1996). Similarly, suppression of pilocarpine induced seizures by oxcarbazepine is associated with increases in hippocampal 5-HT in mice (Clinckers et al., 2005). However, perfusion of the 5-HT_1A antagonist WAY-100635 into the hippocampus abolishes the anticonvulsant effects of oxcarbazepine, implicating a 5-HT_1A receptor mechanism (Clinckers et al., 2005).

Chronic phenytoin administration increases levels of 5-HT within the motor cortex. However, it decreases levels of 5-HT within the cerebellum in rats (Meshkibaf et al., 1995). Compared to oxcarbazepine, alterations to 5-HT neurotransmission is likely not a major mechanism of phenytoin action. However, the anticonvulsant zonisamide increases 5-HIAA in the striatum and both 5-HT and 5-HIAA within the hippocampus in seizure-naïve rats, implicating enhancement of 5-HT neurotransmission in its mechanism of action (Kaneko et al., 1993; Okada et al., 1992).

Lamotrigine, an anticonvulsant and mood stabilizer, dose-dependently inhibits 5-HT uptake in human platelets and rat brain synaptosomes likely via effects on monoamine transporters (Southam et al., 1998). Another anticonvulsant and mood stabilizer, sodium valproate, dose dependently increases extracellular 5-HT within the hippocampus and caudate in seizure-naïve Wistar rats, although levels of 5-HIAA remain stable (Biggs et al., 1992; Murakami et al., 2001). Similarly, valproic acid acts as an AED and mood stabilizer and increases 5-HT levels within the amygdala of seizure-naïve rats (Kempf et al., 1982).

Consistent with the observation that increases in 5-HT are seizure protective in epilepsy patients and in animal models, 5-HT may be critical in the anticonvulsant action of a variety of AEDs. Whether the increases in 5-HT are primary or secondary to the mechanisms underlying the anticonvulsant action of AEDs is unknown. It is also possible that increases in 5-HT contribute to the mood stabilizing properties of some AEDs.

Seizure associated mortality

People with epilepsy are more likely to die prematurely than the general population due to an increased risk of motor vehicle accidents, falls, burns, drowning, aspiration pneumonia, suicide, status epilepticus, and SUDEP (Devinsky et al., 2016b; Weiss et al., 2010). Epilepsy patients, especially those with generalized tonic-clonic seizures, are also likely to endure accidental mild (dental injury, burn, head injury) and severe (skull fracture, hemorrhage, subdural/epidural hematomas, brain contusion) injuries (Asadi-Pooya et al., 2012; Dabla et al., 2018; Wirrell, 2006). These injuries put epilepsy patients at a greater risk for death or future medical complications (Chang et al., 2012). Epilepsy patients are 10× more likely to drown than the general population, with deaths occurring most commonly in the bathtub (Bain et al., 2018). Because pulmonary edema may occur in drowning and SUDEP, some cases of accidental drowning without submersion may be undiagnosed SUDEP (Cihan et al., 2018).

SUDEP is the leading cause of death in patients with refractory epilepsy (Devinsky et al., 2016b). SUDEP is diagnosed as “a sudden, unexpected, witnessed or unwitnessed, nontraumatic and non-drowning death with or without evidence for a seizure and excluding documented status epilepticus, in which postmortem examination does not reveal a toxicologic or anatomic cause of death” (Nashef et al., 2012). The American Academy of Neurology and American Epilepsy Society presently identifies SUDEP risk at 1.2/1000 patient years in adults and 1/4500 patient years in children (DeGiorgio et al., 2019; Donner et al., 2018; Harden et al., 2017; Sveinsson et al., 2017; Thurman et al., 2014). However, recent research suggests that SUDEP may have equal incidence in pediatric and adult patients (Keller et al., 2018; Sveinsson et al., 2017). SUDEP incidence is likely underestimated, as the International Classification of Disease codes used in death certificate data did not consider SUDEP a separate entity and autopsy recommendations in suspected SUDEP cases have been lacking (Middleton et al., 2018). SUDEP risk is greatest in patients with refractory epilepsy (Devinsky et al., 2016a; Dravet et al., 2013; Hesdorffer et al., 2011). Severe myoclonic epilepsy of infancy, or Dravet syndrome, is associated with a particularly high incidence of SUDEP (Dravet, 2011). The most consistent risk factor for SUDEP is frequency of generalized tonic clonic seizures (Harden et al., 2017; Hesdorffer et al., 2011; Lamberts et al., 2012; Ryvlin et al., 2019; Walczak et al., 2001). Patients with poor AED compliance and patients undergoing polytherapy are also at an increased risk of SUDEP (Hesdorffer et al., 2011; Tomson et al., 2008; Walczak et al., 2001). This is likely reflective of the fact that patients taking multiple medications have a more severe form of epilepsy (Hesdorffer et al., 2013; Ryvlin et al., 2019).

In a multi-center MORTality in Epilepsy Monitoring Units Study (MORTEMUS) of SUDEP incidents in epilepsy monitoring units, all cases of SUDEP featured terminal respiratory arrest prior to terminal asystole (Figure 1) (Ryvlin et al., 2013). This is recapitulated in animal seizure models such as the Dravet and MES seizure models, where breathing cessation precedes cardiac dysfunction and death (Figure 1) (Buchanan et al., 2014; Faingold et al., 2011b; Feng et al., 2017; Goldman et al., 2017; Hajek et al., 2016; Kim et al., 2018). Seizures frequently occur during sleep (Derry et al., 2013). SUDEP frequently occurs at night and nocturnal seizures are considered a SUDEP risk factor (Lamberts et al., 2012; Latreille et al., 2017; Nobili et al., 2011; Purnell et al., 2018). Sleep and circadian related effects on breathing and cardiac activity likely contribute to the uneven temporal distribution of SUDEP (Buchanan, 2013; Purnell et al., 2017; Purnell et al., 2018). The cardiorespiratory, circadian, time-ofday, and arousal mechanisms potentially involved in SUDEP are summarized in the following sections.

Mechanisms of death

Respiratory

Numerous animal models of epilepsy experience respiratory dysfunction, both related and unrelated to seizures. DBA/1 and DBA/2 mice experience respiratory arrest and death following audiogenic seizures (Faingold et al., 2010). Similarly, MES-induced seizures in mice often result in terminal apnea which precedes terminal asystole (Figure 1) (Buchanan et al., 2014). Mechanical ventilation can significantly reduce death in both models (Figure 2) (Buchanan et al., 2014; Venit et al., 2004). Respiratory phenotypes, including hypoventilation, apnea, and a diminished hypercapnic ventilatory response (HCVR) are observed in mouse model of Dravet syndrome caused by a mutation in the Scn1a gene (Kuo et al., 2019). Severe peri-ictal respiratory dysfunction, including ataxic breathing and prolonged hypoventilation has been observed in Dravet patients (Kim et al., 2018). Another epilepsy model, the Kcna1 null mutant mouse, experiences spontaneous seizures as well as progressive respiratory dysfunction that worsens as the animals age (Simeone et al., 2018). Similar to the SUDEP cases examined in the MORTEMUS study (Ryvlin et al., 2013), Kcna1 null mice experience seizure-induced respiratory dysfunction prior to cardiac dysfunction (Dhaibar et al., 2019).

Seizures themselves cause alterations in breathing, such as coughing, hyperventilation, apnea, increased bronchial secretions, laryngospasm, respiratory arrest, and neurogenic pulmonary edema (Kennedy et al., 2015; Nakase et al., 2016; Rugg-Gunn et al., 2016). Oxygen saturation below 90% occurs in about a third of generalized and focal seizures (Bateman et al., 2008). Moreover, post-convulsive central apnea occurs in 22% of generalized clonic seizures and is associated with SUDEP and near-SUDEP cases (Vilella et al., 2019). Although SUDEP excludes status epilepticus, status epilepticus triggers central apnea and hypoventilation in seizure prone sheep, in some cases resulting in death (Johnston et al., 1997).

Seizure duration also appears to play a major role in deleterious respiratory phenotypes. Longer seizures are associated with rises in left atrium and pulmonary artery pressures, the presence of chest x-ray abnormalities such as pulmonary edema, and are correlated with the extent of oxygen desaturation and rise of end tidal CO₂ that occurs during and after seizures (Bayne et al., 1981; Kennedy et al., 2015; Seyal et al., 2010).

In humans, there is also evidence that different types of epilepsy and seizures confer different types or degrees of respiratory dysfunction. Temporal and frontal lobe epilepsy account for the majority of focal epilepsies associated with apnea, with TLE having an 8-fold greater association with apnea compared to frontal lobe epilepsy (Lacuey et al., 2018). Ictal central apneas are also particularly common when both hemispheres are involved in the seizure, but also occur in approximately half of focal impaired awareness seizures (Lacuey et al., 2017; Rugg-Gunn et al., 2016).

The exact mechanism behind respiratory failure in SUDEP is yet unknown. Central apnea and oxygen desaturation occurs in epilepsy patients when seizures spread to the amygdala and when the amygdala is directly stimulated (Dlouhy et al., 2015; Lacuey et al., 2017; Nobis et al., 2018). Patients were not aware of the stimulus-invoked apnea and could voluntarily breathe when prompted, suggesting that the apnea was due to the loss of involuntary ventilatory drive rather than dysfunction in respiratory motor output pathways or musculature (Dlouhy et al., 2015). Disruption of brainstem areas responsible for breathing output by seizures may be another mechanism of respiratory failure. This is supported by the observation that seizures provoke a decrease in the firing of medullary serotonergic neurons during and after seizures concurrent with suppression of breathing (Zhan et al., 2016). Increases in 5-HT may provide a protective mechanism against seizure-induced apneas, as lower postictal serum 5-HT has been associated with postictal central apnea (Murugesan et al., 2019).

Airway obstruction apart from central apnea may be another risk factor for postictal respiratory arrest and SUDEP. An increase in recurrent laryngeal nerve firing during a seizure can cause partial or complete obstruction of airways due to laryngospasm (Nakase et al., 2016). Postictal immobility may prevent repositioning of the head in patients that are prone, which could exacerbate respiratory distress if air exchange is impeded. These phenomena may cause a positive feedback loop, whereby the respiratory distress potentiates postictal immobility, further increasing respiratory distress (Peng et al., 2017). The hypoventilation that occurs during a seizure, when in conjunction with ictal/postictal apneas and postictal hypercapnia and hypoxia, might lead to blood gas derangements that cause fatal cardiac dysfunction and ultimately death (Massey et al., 2014).

Cardiac

Epidemiological studies have consistently shown that individuals with epilepsy have a higher prevalence of structural cardiac disease compared to those without epilepsy (Shmuely et al., 2017; Strine et al., 2005). Long QT syndrome (LQTS), which portends increased risk for cardiac arrhythmias, is common in epilepsy, with 29% of patients with LQTS exhibiting a seizure phenotype (Johnson et al., 2009). Mutations that cause LQTS are found in 7% of SUDEP cases, and another 15% have gene variants that may predispose them to cardiac arrhythmias (Bagnall et al., 2016). Oxygen desaturation below 90% greatly increases the likelihood of QT prolongation, something that may be particularly dangerous in patients who already possess a LQTS phenotype (Seyal et al., 2010).

Some specific forms of epilepsy have accompanying cardiac irregularities. Drug-naïve TLE patients have depressed apnea-mediated heart rate variability (HRV) during sleep. Reduction in HRV is a considered a marker of disease and aging. This decrease may suggest an alteration in reflex baroreceptor activation in TLE patients, predisposing them to cardiac-mediated SUDEP (Nayak et al., 2017). Mouse models of Dravet syndrome experience pathological electrocardiogram (ECG) manifestations during seizures including tachycardia, bradycardia, and a LQTS phenotype (Auerbach et al., 2013; Kalume, 2013; Kalume et al., 2013). Interictal cardiac alterations have also been observed in epilepsy patients, including faster interictal heart rate and a decrease in baroreflex sensitivity (Nayak et al., 2017; Sevcencu et al., 2010).

Seizures have numerous effects on cardiac function including asystole (Figure 3) (Schuele et al., 2008). Sinus tachycardia is seen in up to 80% of seizures (Nashef et al., 1996; Sevcencu et al., 2010). There is evidence that ictal tachycardia occurs with right-side lateralized seizures localized to the temporal lobe (Rugg-Gunn et al., 2016). Seizures may also be accompanied by less common cardiac abnormalities, such as transient myocardial ischemia and ictal bradycardia, which has a higher potential to progress to cardiac asystole than tachycardia (Nayak et al., 2017; Rugg-Gunn et al., 2016).

Figure 3.

Seizures can provoke potentially fatal cardiac arrhythmias and asystole. Twomin ECG traces from two epilepsy patients who experienced ictal asystole independent of respiratory arrest. Used with permission from Schuele et al. 2008.

One potential mechanism for cardiac dysfunction in epilepsy is channelopathy. Most patients with Dravet syndrome carry variants in Scn1a, which encodes a sodium channel expressed in the heart and brain (Depienne et al., 2009). Induced pluripotent stem cells derived from cardiac myocytes from Dravet syndrome patients exhibit increased sodium currents (I_Na) and spontaneous contractions (Frasier et al., 2018). The subject whose myocytes showed the largest increase in I_Na was clinically evaluated and cardiac abnormalities were discovered (Frasier et al., 2018). Altered I_Na properties in Dravet syndrome SCN1A^R1407X/+ knock-in mice may be the result of an increase in the number of Nav1.5 channels, leading to altered cardiac function and a susceptibility to arrhythmogenesis (Auerbach et al., 2013).

KCNQ1 mutant mice express abnormal potassium channels in the heart and brain leading to pathological electrical discharges in both the brain and the heart (Goldman et al., 2009). In some cases these mice experience SUDEP resulting from increasingly irregular cardiac activity (Goldman et al., 2009). Similarly, Kcna1 null mice display increased interictal atrioventricular conduction blocks and bradycardia which may predispose these animals to SUDEP (Glasscock et al., 2010). It is possible that ictal or postictal hypoxemia may be sufficient to trigger life-threatening cardiac arrhythmias in susceptible human patients such as those with cardiac channelopathies (Park et al., 2017). A final factor that may augment deleterious cardiac phenotypes is, ironically, the use of antiepileptic drugs AEDs. It has been suggested that AEDs may trigger conduction abnormalities, which could contribute to the co-existence of arrhythmias in epilepsy and may likewise contribute to autonomic regulation in patients with TLE (Nayak et al., 2017). This may be less problematic if the seizures become well-controlled, but it is possible that multiple AEDs taken by patients with refractory epilepsy may contribute to seizure-induced cardiac dysfunction.

Arousal/PGES

Arousing in response to hypercapnia is an important adaptive response in normal homeostatic sleep processes (Dempsey et al., 2010; Guyenet et al., 2013). During the postictal period arousal to blood gas changes, especially a rise in CO₂, may be a lifesaving reflex (Richerson et al., 2011). 5-HT neurons are thought to be important for respiratory and arousal responses to CO₂. Mice with a >99% reduction in 5-HT neurons in the central nervous system do not arouse to hypercapnic challenges, but do arouse to hypoxic challenges and to auditory and tactile stimuli (Buchanan et al., 2010; Zhao et al., 2006). An individual who is prone following a seizure and whose 5-HT system either recovers quickly or is less depressed should arouse and change position when they begin to rebreathe CO₂. Death can occur when seizures interfere with this ability to arouse (Blumenfeld, 2012; Tao et al., 2010). After repeated rebreathing of CO₂, the prone patient would become hypoxic and acidotic, contributing to cardiorespiratory failure (Buchanan, 2019).

A potential SUDEP risk marker is the protracted postictal generalized EEG suppression (PGES) that follows some seizures (Figure 4) (Lhatoo et al., 2010). PGES is characterized by low amplitude, low frequency electrographic activity that persists after the termination of the seizure and gradually recovers (Lhatoo et al., 2010; Theeranaew et al., 2018). Accurate identification of PGES is challenging even among expert reviewers (Theeranaew et al., 2018). The origin and significance of PGES is not well understood. PGES may represent an electrographic marker of decreased arousability consequent to ictal DRN 5-HT network dysregulation. During PGES, patients experience postictal stupor and are more likely to be immobile, unresponsive, and require resuscitative measures, making the immediate environment dangerous (Kuo et al., 2016; Semmelroch et al., 2012). This hypothesis is supported by a recent study in humans indicating that interictal serum 5-HT levels are inversely related to PGES duration (Murugesan et al., 2018). Although an increase of 5-HT in serum does not reflect brain parenchymal 5-HT levels, it does support the idea that 5-HT may be involved in PGES.

Figure 4.

Postictal generalized EEG suppression (PGES) occurs following some seizures in human epilepsy patients and animal models. (A) Prolonged PGES (148 seconds) and postictal slowing following a generalized seizure in a monitored epilepsy patient who later dies of SUDEP. Used with permission from Lhatoo et al. 2010. (B) Prolonged PGES occurring following non-fatal seizures induced with MES during wake (n = 6) or NREM sleep (n = 4). Fast Fourier transform is plotted relative to baseline subtraction (1 min of wake baseline). Used with permission from Hajek and Buchanan 2016.

The relationship of PGES to SUDEP has been somewhat controversial. Early studies of SUDEP cases demonstrate that PGES duration may be an independent predictor of SUDEP with longer PGES duration correlating with increased SUDEP risk (Bozorgi et al., 2013b; Lhatoo et al., 2010). PGES preceded all instances of SUDEP captured in epilepsy monitoring units in the MORETMUS study (Ryvlin et al., 2013). Other studies have not observed a connection between PGES and SUDEP (Surges et al., 2011). However, PGES is associated with several SUDEP risk factors (Rajakulendran et al., 2015). PGES is associated with a lower postictal oxygen desaturation nadir and a higher level of end tidal CO₂ in epilepsy patients following convulsive seizures (Seyal et al., 2012). Moreover, PGES often follows convulsive seizures (Lamberts et al., 2013a; Lhatoo et al., 2010; Moseley et al., 2015; Poh et al., 2012; Seyal et al., 2013; Seyal et al., 2012; Surges et al., 2011). Fitting with the nocturnal predisposition of SUDEP, PGES is more likely to occur following seizures from sleep, a time of potentially cardiorespiratory instability (Lamberts et al., 2013b; Latreille et al., 2017; Peng et al., 2017; Purnell et al., 2017). PGES following MES seizures during NREM sleep is longer than seizures induced during wake in mice (Figure 4) (Hajek et al., 2016). PGES is also associated with increased sympathetic tone and decreased parasympathetic tone, both of which can precipitate potentially fatal cardiac arrythmias (Bozorgi et al., 2013a; Moseley et al., 2013; Poh et al., 2012). However, PGES was not associated with heart rate variability in another study, but this does not preclude inadequate perfusion (Lamberts et al., 2013b).

Association between SUDEP and the night

SUDEP occurs more frequently during the night (Ali et al., 2017; Hesdorffer et al., 2011; Lamberts et al., 2012; Langan et al., 2005; Purnell et al., 2018; Ryvlin et al., 2013; van der Lende et al., 2018). There are several possible explanations as to why SUDEP may occur at night, including reduced nighttime supervision (Blachut et al., 2015; Witek et al., 2017). However, it is likely that circadian and sleep-wake influences on 5-HT contribute to SUDEP.

The nocturnality of SUDEP is often attributed to differences in seizure frequency or severity occurring across vigilance states. Vigilance state at the time of seizure onset is often inferred based on time of day and proximity of the patient to their bed (Ali et al., 2017; Kloster et al., 1999; Lamberts et al., 2012). These studies consistently find that SUDEP is sleep related, but the absence of EEG recording precludes verification that the fatal seizure emerged from sleep. However, EEG data recorded from patients who died in epilepsy monitoring units supports that SUDEP occurs more frequently during sleep (Ryvlin et al., 2013).

Animal research also supports that vigilance state may be integral to SUDEP. MES seizures in mice are universally fatal when induced during REM sleep (Hajek et al., 2016). 5-HT may be involved in the uneven temporal distribution of SUDEP, as 5-HT neuron activity is modulated by vigilance state. 5-HT neurons are most active during wakefulness, less activity during NREM sleep, and nearly quiescent during REM sleep (Trulson et al., 1979). Downregulation of 5-HT activity during sleep may result in a reduction of 5-HT tone following a seizure and an increased risk of SUDEP.

That SUDEP happens more often during the night could also speak to an independent role of circadian rhythms (Purnell et al., 2017; Purnell et al., 2018). Many aspects of mammalian physiology are regulated by circadian rhythmicity. Sleep-wake behavior is subject to circadian regulation; however, circadian rhythms occur independent of sleep (Borbely et al., 2016). Patients who die of SUDEP are more likely to have a history of seizures during the night (Borbely et al., 2016). SUDEP has a nonuniform distribution regardless of sleep-state of origin. SUDEP inferred to have happened while awake occurs between 0800 and 1200 whereas SUDEP inferred to have happened during sleep occurs between 0400 and 0800 (Borbely et al., 2016). This suggests that there is a circadian component to the timing of SUDEP.

In several animal models, day-night differences in seizure severity and susceptibility have been observed and attributed to circadian differences in 5-HT and norepinephrine (Agren et al., 1986; Schreiber et al., 1971). In both humans and rodents 5-HT is highest during the day, lower during the night, and reaches its nadir in the early pre-dawn hours (Agren et al., 1986; Mateos et al., 2009; Rao et al., 1994; Rosenwasser et al., 1985). In the Kv1.1 null mouse and the Scn1a^R1407X/+ Dravet syndrome mouse, seizure induced death is more frequent during the night, with peaks at the light dark transition points (Moore et al., 2014; Teran et al., 2019). The normal circadian reduction in 5-HT during the late part of the night might make epilepsy patients more vulnerable to SUDEP during this time. Additionally, the effect of vigilance state and circadian phase may compound to cause further reductions in 5-HT tone. For example, if 5-HT tone is reduced during the late part of the night and it is reduced by sleep, sleep during the late part of the night might be a particularly hazardous time for a seizure to occur.

Similarities between SIDS and SUDEP

SUDEP shares similarities with other sudden death entities, most notably the sudden infant death syndrome (SIDS; Table 2). SIDS is defined as the unexpected, sudden death of an infant less than one year of age, for which no explanation is uncovered by exhaustive review of the clinical history, investigation of the death scene, and a complete autopsy (Willinger et al., 1991). Dysregulation of respiratory, cardiac, and arousal systems have been implicated in SIDS and SUDEP (Buchanan, 2019; Kinney et al., 2009; Richerson et al., 2011). Due to its role in regulation and/or modulation of these systems, 5-HT has been implicated in both SIDS and SUDEP (Buchanan, 2019; Richerson et al., 2011). Abnormalities in the brainstem 5-HT system have been identified on postmortem examination of the brains of several independent cohorts of babies that died of SIDS (Bright et al., 2017; Duncan et al., 2010; Haynes et al., 2017; Kinney et al., 2001; Kinney et al., 2003; Panigrahy et al., 2000; Paterson et al., 2006). As seen below, 5-HT abnormalities are beginning to be found in SUDEP cases.

Table 2.

SUDEP shares commonalities with other sudden death syndromes such as SIDS. Modified with permission from Buchanan 2019.

	SIDS	SUDEP
Diagnostic method	Diagnosis of exclusion	Diagnosis of exclusion
Baseline health	Normal	Normal except seizures
Male sex	More common in males	More common in males
Age at death	<1 year	Any age
Incidence	~0.6 per 1000 live births	0.2–1.2 in 1000 persons with epilepsy
Routine autopsy findings	Normal	Normal; pulmonary edema
Proposed mechanism of death	Respiratory, cardiac, arousal impairment, thermoregulatory dysregulation	Respiratory, cardiac, arousal impairment
Circumstances of death	Often found prone	Often found prone
Occurrence during sleep	Common, often unwitnessed	Common, often unwitnessed
Link to 5-HT	Yes	Yes
Microscopic hippocampal abnormalities	Yes	Yes

Additional evidence of a role for 5-HT in SUDEP

Human studies

Although it is impossible to determine if and when a patient will suffer SUDEP, it is possible to study high-risk patients and post-mortem samples to better elucidate SUDEP physiology. Volume loss is observed within the periaqueductal gray matter and 5-HT producing raphe nuclei in MRI imaging of epilepsy patients at high-risk for SUDEP. This is consistent with the role of the raphe nuclei in measures critical for SUDEP including breathing, sleep/wake regulation, and arousal. The magnitude of that deficit was greater in the two patients who later died of SUDEP and extended into the pontine and medullary brain regions (Mueller et al., 2014). Diminished TpOH and serotonin transporter (SERT) expression is also observed within the raphe nuclei in post-mortem evaluation of SUDEP cases compared to epilepsy controls (Patodia et al., 2018). Patients with TLE can express TPH2 gene polymorphisms that may affect cognition/mood, but it is also possible they affect predisposition to cardiorespiratory dysfunction (Bragatti et al., 2014; Richerson et al., 2011). Thus, preliminary findings in high-risk patients and post-mortem SUDEP victims are consistent with the epilepsy patient and animal studies implicating 5-HT broadly in seizure susceptibility.

Animal studies

5-HT may also play a role in seizure-induced mortality in animal models. Pharmacological increases in 5-HT may be protective against SUDEP in animal models. Fluoxetine suppresses S-IRA in DBA/1 and DBA/2 mice without influencing seizure severity or ventilation (Figure 2) (Tupal et al., 2006; Zeng et al., 2015). Fluoxetine may act on S-IRA in DBA/1 mice via a 5-HT₃ mechanism. Administration of SR 57227, a 5-HT₃ agonist, blocks S-IRA without affecting seizures. Ondansetron, a 5-HT₃ antagonist, abolishes the protective effect of fluoxetine on S-IRA (Faingold et al., 2016). The SSRIs fluvoxamine and sertraline more broadly increase 5-HT tone and suppress S-IRA in DBA/1 mice (Faingold et al., 2014; Faingold et al., 2013). Fenfluramine, a serotonin releasing appetite suppressant, not only protects against S-IRA in DBA/1 mice, but reduces seizure incidence and severity (Figure 2) (Tupal et al., 2019). Conversely, blocking 5-HT transmission with cyproheptadine increases S-IRA incidence in DBA/1 and DBA/2 mice that did originally not exhibit S-IRA (Faingold et al., 2014; Tupal et al., 2006).

Lmx1b^f/f/p mice lacking >99% of CNS 5-HT neurons have a lower seizure threshold and increased seizure-induced mortality in the acute pilocarpine and MES seizure models (Buchanan et al., 2014; Zhao et al., 2006). MES severity is likewise increased in C57BL/6 mice administered PCPA to reduce 5-HT to ~5–11% of original levels within the medulla and forebrain (Buchanan et al., 2014). As in the DBA mice, death occurs primarily via a respiratory mechanism in the MES model (Figure 1) (Buchanan et al., 2014). In wildtype mice, raising 5-HT levels with either citalopram, fluoxetine, duloxetine, the 5-HT₂ agonist DOI, or the 5-HT_2A agonist TCB-2 increased MES seizure survival (Figure 2) (Buchanan et al., 2014; Kruse et al., 2019). Similarly, administration of reboxetine, fluoxetine, DOI or TCB-2 prior to MES induction prevents death in Lmx1b^f/f/p mice (Figure 2) (Buchanan et al., 2014; Kruse et al., 2019). As in prior sections, it is seen that an increase in 5-HT tone is generally anticonvulsant and lowers mortality, whereas a deficit in 5-HT promotes seizure-induced mortality that can be reversed by increases in 5-HT.

Summary/Conclusions

5-HT plays a critical modulatory role in the pathology of seizures and epilepsy. In recent years, 5-HT has become a strong topic interest within the SUDEP research community. 5-HT is uniquely positioned because of its role in biological processes relevant to SUDEP including breathing, sleep/wake regulation, circadian rhythms, and arousal. Dysregulation of 5-HT may underlie certain epilepsies. Seizures may also exacerbate pre-existing 5-HT abnormalities.

SUDEP may result from a ‘perfect storm’ of intrinsic and extrinsic factors related to the fatal seizure (Figure 5). Interference with 5-HT neurotransmission may contribute to death given its multifaceted modulatory role. Propagation to brainstem centers may depress 5-HT respiratory nuclei, producing hypoventilation and apnea, and depress 5-HT arousal networks, limiting arousal in response to blood gas changes. The additional insult of a prone position may preclude successful autoresuscitation. Moreover, prolonged central apnea can exacerbate or initiate cardiac dysfunction, which in turn can worsen respiratory outcomes until terminal apnea occurs.

Figure 5.

Potential pathways by which seizures can lead to SUDEP. Propagation of seizures to cardiorespiratory centers within the brain can lead to potentially fatal cardiac arrythmias or apneas that can exacerbate respiratory and cardiac dysfunction respectively. Propagation of seizures to arousal centers and the occurrence of PGES following a seizure may prevent reflexive arousal to airway obstruction and lead to potentially fatal apneas. Risk factors including a patient’s lifestyle (adherence to medication, control of triggers), the time of day of the seizure, circadian factors, vigilance state, and postictal position can increase the likelihood of a seizure becoming fatal.

It is possible that future preventative “treatments” against SUDEP may involve adjunctive SSRIs or other pharmacological agents in high risk patients to improve postictal cardiorespiratory outcomes and reduce arousal impairment. However, future research into relevant 5-HT networks may also inform targets for deep-brain stimulation. Recovering baseline activity of 5-HT networks disrupted by seizures may be one way to protect patients at greatest risk for SUDEP. Regardless of its ultimate direction, research endeavors into 5-HT will continue to yield integral findings that will both benefit the lives of epilepsy patients and perhaps reduce SUDEP.

Abbreviations:

SUDEP

sudden unexpected death in epilepsy

5-HT

serotonin

TLE

temporal lobe epilepsy

5-HTT

serotonin transporter

5-HIAA

5-hydroxyindoleacetic acid

CSF

cerebrospinal fluid

AED

anti-epileptic drug

TPH2

tryptophan hydroxylase 2

TpOH

tryptophan hydroxylase

GEPR

genetically epilepsy prone rat

S-IRA

seizure-induced respiratory arrest

SSRI

selective serotonin reuptake inhibitor

MES

maximal electroshock seizure

PTZ

pentylenetetrazol

PCPA

para-chloroamphetamine

5-HTP

5-hydroxytryptophan

MORTEMUS

MORTality in Epilepsy Monitoring Units Study

LQTS

long QT syndrome

HRV

heart rate variability

ECG

electrocardiogram

PGES

post-ictal generalized EEG suppression

SIDS

sudden infant death syndrome