Limitations of pupil diameter as a proxy for modes of locus coeruleus activity

Technical Summary: Limitations of Pupil Diameter as a Proxy for Modes of Locus Coeruleus Activity

Problem Statement
The locus coeruleus-norepinephrine (LC-NE) system is hypothesized to regulate higher brain functions through distinct modes of activation: a "phasic" mode (moderate baseline, high evoked response) associated with optimal engagement, and a "tonic" mode (high baseline, low evoked response) associated with distractibility. These modes are thought to follow an inverted-U relationship with arousal and performance (Yerkes-Dodson law). Because direct measurement of LC activity is difficult due to its small size and deep brainstem location, researchers frequently rely on pupil diameter as a physiological proxy. Previous work suggests that pupil size correlates with LC activity and that baseline and evoked pupil modulations exhibit an inverse relationship similar to that of LC activity. However, it remains unclear whether these within-modality relationships (e.g., baseline pupil vs. evoked pupil) reliably predict cross-modal relationships (e.g., baseline pupil vs. evoked LC), which is necessary to infer LC activation modes solely from pupil measurements.

Methodology
The study utilized simultaneous recordings of single-unit LC spiking activity and pupil diameter in two awake monkeys (Oz and Ci) performing a passive fixation task.

• Subjects and Data: The dataset comprised 42 single units from 32 sessions in monkey Oz and 63 units from 51 sessions in monkey Ci.
• Task: Monkeys maintained fixation for 1–5 seconds. On approximately 25% of trials, a startling auditory beep (1 kHz, 0.5 sec) was presented to evoke LC and pupillary responses.
• Measurements:
  • Baseline: Defined as the period from fixation onset to 20 ms before the beep (or end of fixation for no-beep trials).
  • Evoked: Defined as the response window 0–200 ms post-beep for LC activity (baseline-subtracted) and the maximum change in pupil diameter relative to baseline for pupil size.
• Analysis: The authors analyzed trial-by-trial correlations within and across modalities. They corrected for linear drifts over sessions using residuals. Statistical significance was assessed using Wilcoxon rank-sum tests on partial Spearman correlation coefficients across sessions. Additional analyses tested for non-linear (quadratic) relationships and pooled data across sessions to address potential underpowering.

Key Results

1. Within-Modality Coupling: The study replicated established findings showing reliable positive correlations between baseline LC activity and baseline pupil diameter, as well as between evoked LC responses and evoked pupil responses. Furthermore, within each modality, a significant negative correlation was observed between baseline levels and evoked responses (i.e., higher baseline activity predicted smaller evoked responses), consistent with the phasic/tonic mode hypothesis.
2. Cross-Modal Decoupling: Crucially, the study found no reliable relationships when comparing baseline measures in one modality to evoked measures in the other.
   • Evoked pupil responses did not reliably predict baseline LC activity (p = 0.894 for Oz; p = 0.904 for Ci).
   • Baseline pupil diameter did not reliably predict evoked LC responses (p = 0.915 for Oz; p = 0.008 for Ci, though the latter was not consistent across the full dataset in a manner supporting the hypothesis).
3. Robustness Checks: These null findings for cross-modal inference held true even when:
   • Analyzing raw data without drift correction.
   • Testing for non-linear (inverted-U) relationships, which were found to be rare in the dataset.
   • Pooling data across all trials, sessions, and units.
   • Stratifying data based on the direction of the within-session baseline coupling.

Key Contributions

• Empirical Validation of Proxy Limitations: The paper provides direct electrophysiological evidence that while pupil size and LC activity are coupled within the same epoch (baseline vs. baseline, or evoked vs. evoked), they are not reliably coupled across epochs.
• Refutation of Cross-Modal Inference: The authors demonstrate that one cannot infer the "mode" of LC activation (phasic vs. tonic) based on the relationship between baseline pupil size and evoked pupil size, nor can one infer baseline LC state from evoked pupil responses.
• Clarification of Mechanisms: The results suggest that the inverse relationship between baseline and evoked responses observed in pupil size is likely driven by physical constraints (e.g., the limited dynamic range of the pupil causing saturation at large diameters) and multiple neural inputs, rather than being a direct, linear reflection of LC firing modes.

Significance and Claims
The authors conclude that pupil modulations, while reflecting LC-NE activity, should be used with caution when making inferences about specific LC activation modes (phasic vs. tonic). The lack of reliable cross-modal coupling implies that the "inverted-U" relationship often attributed to LC function based on pupil data may not accurately reflect the underlying neural dynamics of the LC-NE system. The paper asserts that the two systems are insufficiently coupled to allow moment-to-moment pupil changes to serve as a proxy for LC firing modes. The authors suggest that future work requires targeted experiments involving pharmacological and task-specific manipulations to disentangle the contributions of different neuromodulatory systems and physical constraints before pupil diameter can be reliably interpreted in terms of LC-NE function.