The unique face of anxious depression: Exaggerated threat but preserved positive valence reactivity

Introduction

Nearly half of individuals with Major Depressive Disorder (MDD) also have an anxiety disorder (Kessler et al., 2015). Thus, comorbid MDD and anxiety disorders (e.g. generalized anxiety disorder, social phobia, panic disorder, or simple phobia) are among the most common presentations for mental health providers. Yet, the underlying brain and behavioral processes that characterize anxious depression are still incompletely understood. A better understanding of the underlying process dysfunctions is an important first step in the development of more mechanistically oriented interventions. Anxious depression is associated with greater treatment resistance, (Ionescu et al., 2014), quicker symptom relapse (Fava et al., 2008), poorer treatment outcomes, (Penninx et al., 2011) and higher levels of suicidal ideation (Kessler et al., 2015) than non-anxious depression. However, treatment engagement is reported to be higher (Kessler et al., 2015). Typical treatments include traditional antidepressants augmented with benzodiazepines or atypical antipsychotics. This can be problematic because of the risk for use disorder and unfavorable side effects profiles, which can be increased in anxious depression (Gaspersz et al., 2017).

MDD and anxiety have some overlapping dysfunction in neurocognitive processes (McTeague et al., 2020). However, there are important features which make them distinct. According to Clark and Watson’s seminal model of depression and anxiety hyperarousal is a key feature of anxiety and blunted positive responding or anhedonia is a key feature of depression (Clark & Watson, 1991). Adding to this, the emotion context insensitivity model (Rottenberg et al., 2005) reports attenuation of both positive and negative affect in MDD, proposing a general level of emotional blunting in MDD. It is unclear what happens when this emotional blunting and hyperarousal are combined in anxious MDD. Additionally, there is presumably a healthy level of “arousal” that lies between blunted non-anxious MDD and “hyperaroused” anxiety/anxious MDD. This means that when those with anxious and non-anxious MDD are merged into a single clinical sample, comparisons on measures of arousal with healthy people are not helpful, as these two clinical groups lie on either end of the arousal spectrum. Furthermore, studies comparing anxious versus non-anxious MDD are rare, which may be impeding treatment selection and development.

Following on from this, hyperarousal is a feature of anxiety and anxious MDD, that is not likely to be present in non-anxious depression and therefore may be a useful candidate target mechanism. Hyperarousal in anxious MDD is associated with dysregulation of stress circuitry (Powers et al., 2016) and cortical thinning in prefrontal areas associated with top-down aspects of emotional regulation (Zhao et al., 2017). Threat responses are particularly associated with anxiety (Bar-Haim et al., 2007) and have been shown to engage a corticolimbic circuit (Mobbs et al., 2009; Mobbs et al., 2007). In a functional magnetic resonance imaging (fMRI) study comparing anxious and non-anxious MDD, limbic responses to emotional conflict (presentation of emotional facial expressions and non-task relevant incongruent emotional words) were similar across anxiety and MDD but prefrontal regulation of emotional conflict was absent in anxiety/ anxious MDD but preserved in non-anxious MDD (Etkin & Schatzberg, 2011), suggesting that deficits in top-down aspects of emotional regulation may be key to this comorbidity. Emotional regulation is associated with structural and functional connectivity of frontal and limbic regions (Kim et al., 2011). MDD is associated with reduced corticolimbic functional connectivity (Kaiser et al., 2015), although findings are mixed (Williams, 2017), perhaps due to the fact that MDD is comprised of heterogeneous subgroups. Examining the anxious MDD subgroup may remove some of the variance, but few studies have examined this. In a study of depressed patients, amygdala functional connectivity mediated the relationship between anxiety and MDD (He et al., 2019). Together, this suggests that top-down control of defensive responses to acute threat (and the associated circuits) may be a suitable cognitive target specific to anxious versus non-anxious MDD.

The startle reflex is a defensive eyeblink response to an intense stimulus. The plasticity of this reflex in relation to positive or negative affective context makes this an ideal experimental approach to examine alterations of the defense system (Boecker & Pauli, 2019). In healthy participants startle reflex is attenuated by appetitive stimuli and potentiated by aversive stimuli; a phenomenon called affective startle modulation (ASM) (Bradley et al., 1990; Lang et al., 1990); explained by motivational priming of the appetitive/defensive systems (Lang, 1995). Hereby congruent aversive motivational states prime the defensive system, potentiating the defensive startle response, whereas incongruent appetitive states attenuate the response. Deviations from the expected pattern of ASM are thought to reflect abnormal functioning of the underlying motivational system. A recent review of ASM and psychopathology (Boecker & Pauli, 2019) suggests that the defensive/appetitive systems operate independently and characterize depressed and anxious psychopathology in the following ways: 1) increased affective startle potentiation (ASP) to aversive stimuli in anxiety (Cook et al., 1992; Cook et al., 1991; Garner et al., 2011; Temple & Cook, 2007) and to phobic stimuli in phobia (Globisch et al., 1999), and 2) general hyporeactivity to aversive and appetitive affective stimuli in MDD, with blunted affective startle modulation in severe MDD (Kaviani et al., 2004). Interestingly, in a study examining a range of anxiety patients, responses were blunted in participants with more pervasive disorders such as multiple trauma PTSD and co-morbidity with MDD (McTeague & Lang, 2012), suggesting disengagement of the defense system in response to chronic disease course. These opposing forces of depression and anxiety on startle make the examination of anxious MDD necessary and potentially useful. In practical terms, startle paradigms are more clinically feasible to implement than neuroimaging and have recently been employed as a way to screen novel anxiolytics (Grillon & Ernst, 2020).

Preclinical literature has established a key role for the defensive and appetitive neuronal systems in the modulation of startle reflex, namely the bed nucleus of the stria terminalis (BNST), the amygdala (Davis et al., 1999) and the nucleus accumbens (Koch et al., 1996). Concurrent neuroimaging and startle electrophysiology is technically difficult, historically limiting human findings to lesion studies of the amygdala showing elimination of negative ASM (Buchanan et al., 2004). However, recent advances allowed converging evidence in human fMRI of valence-specific triggered amygdala responding, with individual level associations between startle magnitude and neural activation strength, suggesting that startle measures could be used as a direct read-out of neural activation of the central amygdala (Kuhn et al., 2020), allowing measurement of a key component of the corticolimbic circuit implicated in anxious MDD.

ASM to aversive stimuli is dimensionally associated with trait fear (Vaidyanathan et al., 2009), a self-report measure of threat sensitivity. The goal of this investigation was to determine whether anxious MDD shows a unique profile of exaggerated defense related processes (i.e. hyperarousal) using data from a large transdiagnostic sample collected as part of the Tulsa 1000 study (Victor et al., 2018). The basic approach was to compare a propensity-matched sample of depressed and anxious depressed individuals using a multi-level approach focused on positive/ negative valence and threat using symptoms and physiological levels of analysis. Based on previous findings, we hypothesized that individuals with anxious depression but not those with non-anxious depression would show an increased threat-related startle response pattern characterized by an altered ASM response to negative stimuli compared to positive stimuli.

Methods and Materials

Procedure

General procedures included a clinical interview session, a neuroimaging session and a behavioral (electrophysiology) session, completed within two weeks on average. Although the parent project (i.e., T1000) consisted of a broader range of protocols, only details relevant to the current study are presented here. See protocol paper (Victor et al., 2018) for full details.

Study staff administered the MINI clinical interview. During this session, participants also provided self-reported information on demographics. For the current study we focused on dimensional measures of anxiety/threat sensitivity (Anxiety Sensitivity Index; ASI), depression (PROMIS depression scale) and as a follow up, approach and avoidance motivation (Behavioral Inhibition/ Behavioral Activation scale; BIS/BAS).

Participants completed an emotional reactivity task (Lang et al., 1990) during EMG eyeblink recording (Fig. 1). Facial EMG was recorded to assess physiological startle from the orbicularis oculi in accordance with published guidelines (Blumenthal et al., 2005). Two electrodes were attached below the lower eyelid of the left eye, one below the outer edge and one below the centre of the eye (distance approx. 15 mm). Participants viewed appetitive, neutral, and aversive images from the International Affective picture series (IAPS) (Lang et al., 1999) for 6 s. Noise probes (95 dB) were presented between 2500 and 4500 ms after picture onset to elicit startle blink responses during 24 of these image presentations (8 per valence). Electrocardiogram (EKG) data were also collected at a baseline session (during rest) and during the emotional reactivity task. Analyses were carried out on the first 500 participants of the T1000 sample and were not preregistered. Confirmatory analyses from the second 500 participants will be preregistered.

• Download figure
• Open in new tab

Figure 1: Emotional reactivity task:

A) Each trial begins with a 20-26s fixation period, followed by presentation of one image for 6s, during which a startle probe is presented. After each image, the participant makes valence and arousal ratings on a 7 point scale. (IAPS images blurred in this schematic for copyright reasons; first example image is a pancake (appetitive), second example image is a building (neutral)) B) Startle EMG data collected from electrodes placed on the orbicularis oculi.

Results

Startle response

Mixed effects linear regression results showed a significant valence X group interaction (F(2,310) 3.460, p = 0.03, R² = 0.022). Planned contrasts showed that the Dep group had no modulation of startle response from positive or negative valence (pairwise comparisons; all p > 0.5), whereas the Dep+Anx group showed negative potentiation compared to positive (negative > positive) (t(184) = 4.608, p < 0.001) and positive attenuation compared to neutral (positive < neutral) (t(548) = -5.018, p < 0.001) (Fig. 2). The Dep+Anx group also had lower startle response during appetitive stimuli compared to the Dep group (t(397) = -2.273, p = 0.02). Supplemental analyses comparing to healthy controls are included for completeness in the Supplement.

• Download figure
• Open in new tab

Figure 2:

Blunted affective startle modulation in depression versus comorbid depression and anxiety. Shaded area represents the 95% confidence interval.

Dimensional analyses

Dimensional analyses in the same sample examined the effects of self-report anxiety sensitivity (ASI) and depression (PROMIS) on raw startle response using linear regression. There was a significant ASI X Depression interaction. This indicates that depression moderated the effect of anxiety sensitivity (F(1,178 = 4.609 p = 0.03, R² = 0.032). For the lower three quartiles of self-report depression, ASI was significantly correlated with overall startle reflex (r(112) = 0.23, p = 0.01), but this relationship was not observed in the top quartile (r(72) = -0.16, p = 0.18). This interaction was present when examining mean startle across all trials (Fig. 3) and during appetitive (p = 0.05) or aversive (p = 0.01) stimuli separately but not when examining difference scores of aversive versus appetitive or neutral (all p > 0.7).

• Download figure
• Open in new tab

Figure 3:

Moderation effect of self-report depression scores on effect of self-report anxiety sensitivity on overall startle response. PROMIS: Patient-Reported Outcomes Measurement Information System; ASI: Anxiety Sensitivity Index.

Discussion

This multi-level investigation aimed to examine whether anxious MDD is characterized by a unique profile of exaggerated defense related processes relative to propensity matched participants with non-anxious MDD. There were four main results: First, at a physiological level of analysis the Dep+Anx group showed preserved ASM whereas the Dep group showed no ASM; Second, at a behavioral level of analysis, the Dep group had slower RTs to arousal ratings for aversive cues than Dep+Anx; Finally, at a self-report level of analysis, depression moderated the effects of anxiety sensitivity on overall startle reflex where the most severely depressed individuals did not have an association between anxiety sensitivity and startle reflex. We further characterized the Dep+Anx group as having greater self-report behavioral avoidance. Together, these results support the hypothesis that anxious depression is characterized by affective modulation of appetitive and defensive systems relative to those with non-anxious depression and that non-anxious depression only is associated with valence independent blunting of emotional reactivity.

Previous work based on symptom patterns supports a tripartite model (Clark & Watson, 1991), consisting of a general negative affect factor and specific MDD and anxiety factors. The general factor is increased negative affect, common to both MDD and anxiety. In our findings this would be represented by the self-report depression scale which was similar in Dep versus Dep+Anx, showing that this general factor is associated with blunted overall startle. The specific MDD factor in the tripartite model is the absence of positive affect (anhedonia). In the emotional reactivity task the positively valenced images prime the appetitive system. As expected from the tripartite model, we see a lack of positive affective modulation from appetitive images in Dep. However, this was not the case in the Dep+Anx group, where we found preserved positive affective modulation. This suggests a novel key difference that may help explain greater treatment engagement in anxious depression (Kessler et al., 2015). The anxiety specific factor in the tripartite model is hyperarousal. In the emotional reactivity task the negatively valenced images prime the defensive system. This is illustrated by EKG analyses showing increased sympathetic activation in this task compared to rest. As expected from the tripartite model the Dep+Anx group show hyperarousal, represented by higher affective modulation from negative compared to positive images, which is not observed in Dep, suggesting another key difference. Hyperarousal is also represented by self-report anxiety sensitivity, which was also significantly higher in Dep+Anx versus Dep. These two measures of hyperarousal were significantly correlated in all but the most depressed participants, suggesting a moderating effect of depression.

This blunting of negative emotional reactivity aligns with the ECI model of MDD (Rottenberg et al., 2005), which seems not to apply to MDD when it is comorbid with anxiety. The striking difference between these two groups argues that they should not be combined for trials that target measures of hyperarousal, as differences from HCs will be obscured by the depressed and anxious clinical groups residing on either end of the “hyperarousal spectrum”. To illustrate this, supplemental analyses from a smaller HC group suggest that the HC ASM profile is between the Dep+Anx group and the Dep group. This may reflect a “normal” level of arousal between blunted Dep and hyperaroused Dep+Anx. This could be interpreted as anxiety “rescuing” the impairment from depression. However, this interpretation contrasts with the clinical observations that comorbid MDD and anxiety disorders are associated with worse clinical outcomes than MDD alone. Therefore, additional studies will be necessary to elaborate on the connection between startle dysfunction and clinical outcomes in individuals with Dep+Anx.

The approach-withdrawal model helps tease apart the behavioral factors, with depression being associated with reduced approach and anxiety being associated with increased avoidance (Davidson, 1998). In the present study, reduced approach and withdrawal could be compared to a lack of appetitive/approach or defensive/withdrawal priming in Dep, which is preserved in Dep+Anx. Speculatively, this suggests that treatments targeting the appetitive system (e.g. antidepressants acting on the dopamine system such as bupropion) may not be as effective in anxious depression, as (in terms of startle modulation at least) this system seems to be somewhat preserved. Conversely, treatments that target hyperarousal (e.g. GABAergic treatments) may not be necessary in non-anxious depression and may, we speculate, contribute towards further emotional blunting. Future research should examine how those with anxious versus non-anxious MDD respond to treatments separately targeting the appetitive and defensive systems.

These results add to prior work (Ironside et al., 2023) showing increases in interoceptive and nociceptive reactivity in anxious versus non-anxious MDD. Together, these findings have led us to propose a process model for anxious MDD to help distinguish disease modifiable processes that could be targeted with treatment. Anxiety may be characterized by excessive defense system activation which may be highly taxing and result in exhausting affective processing capacities ultimately resulting in MDD. This is in line with data showing that two-thirds of individuals with lifetime comorbid anxiety disorders and MDD reported an earlier age-of-onset of their anxiety disorder than their MDD (Kessler et al., 2015). In comparison, individuals with non-anxious MDD may be characterized by a primary lack of reactivity to affective stimuli, which results in a lack of anxious responding even when threat occurs. Therefore, whereas the primary disease-modifying process for anxious MDD would be to attenuate threat-related processing, the primary disease-modifying process for non-anxious MDD would be to enhance valence-related processing in general. Reward responsivity is a well-studied process in MDD, but we propose a more general factor of blunted valence responding, first proposed by the ECI model of MDD (Rottenberg et al., 2005) and supported by these data. This would modify the tripartite model, transforming the anxiety specific factor of hyperarousal into a general factor “arousal” with depression and anxiety working in opposition. If this is the case, the ECI model is insufficient to explain anxious MDD and further characterization is needed.

This study had several limitations. First, case-control designs and cross-sectional studies cannot arbitrate between cause and effect. Thus, mechanistic explanations require future experimental testing. Second, although the paradigm has been used extensively in prior investigations, compared to other experimental paradigms this version had fewer trials. Third, the Dep+Anx group was twice as large, which increases the ability to detect effects in the Dep+Anx versus the Dep group. Fourth, for dimensional measures, the interpretation of the affective modulation of startle is limited as we see similar effects across valence, but we suggest that when the groups are collapsed the blunting effect of depression on overall startle overshadows affective modulation. However, we strongly feel that propensity matching within sub-types of MDD aims to address a crucial gap in the literature and can contribute to solving the crisis of heterogeneity in MDD research. Blunted startle could be related to a single impairment downstream from these systems in non-anxious depression. For example, a sensory deficit (Serafini et al., 2017) or a motor deficit (Sobin & Sackeim, 1997). Sensory deficits are difficult to investigate with the current data but our lack of valence independent group differences on RTs suggests that these groups may not have a single downstream motor deficit that could drive our startle findings. In addition, a recent review (Boecker & Pauli, 2019) suggests that the appetitive and defensive systems operate independently and are affected in opposing ways by anxiety and depression. Although behavioral inhibition was significantly different between the groups it was not associated with startle, although we would argue that this is unsurprising as behavioral inhibition is more related to cognitive arousal. The ASI was the best measure of anxious arousal that we had in our self-report data. Future studies of threat sensitivity should use more specific self-report scales than the ASI, for example the Trait Fear Scale (Kramer et al., 2020). Finally, there were no group differences on heart rate variability, although the emotional reactivity task was not designed to probe this.

In sum, these findings suggest that those with anxious and non-anxious MDD have distinct neurocognitive profiles and thus, may require different treatment approaches.

Exaggerated defensive responses in anxious MDD may promote avoidance behavior, whereas a lack of positive affect may be driving the same outcome in non-anxious MDD. The preservation of positive affective modulation suggests that anxious MDD is not as simple as depression plus anxiety and that there may be an interactive effect of depression and anxiety that could be key to understanding and treating this comorbidity. To our knowledge, this is the first time that findings on preservation of the appetitive system in anxious MDD have been presented; enabled by our propensity matching approach using both categorical and dimensional analyses. This appetitive system preservation may be a useful tool for developing strategies to regulate negative affect and hyperarousal in this large patient group.

Author Contribution Statement

Maria Ironside: Conceptualization, formal analysis, writing – original draft

Rayus Kuplicki: Formal analysis, methodology, software, writing – review and editing

Ebony A. Walker: Data curation

Tate Poplin: Data curation, formal analysis

Cheldyn M. Ramsey: Data curation

Katherine L. Forthman: Formal analysis, software, writing – review and editing

Melissa E. Nestor: Software

Robin L. Aupperle: Conceptualization, writing – review and editing

Salvador M. Guinjoan: Writing – review and editing

Sahib S. Khalsa: Conceptualization, writing – review and editing

Jonathan Savitz: Conceptualization, writing – review and editing

Jennifer L. Stewart: Conceptualization, writing – review and editing

Teresa A. Victor: Conceptualization, writing – review and editing

Martin P. Paulus: Conceptualization, supervision, methodology, funding acquisition, writing – review and editing