Reconciling functional differences in populations of neurons recorded with two-photon imaging and electrophysiology

Joshua H. Siegle, Peter Ledochowitsch, Xiaoxuan Jia, Daniel Millman, Gabriel Koch Ocker, Shiella Caldejon, Linzy Casal, Andrew Cho, Daniel J. Denman, Séverine Durand, Peter A. Groblewski, Greggory Heller, India Kato, Sara Kivikas, Jérôme Lecoq, Chelsea Nayan, Kiet Ngo, Philip R. Nicovich, Kat North, Tamina K. Ramirez, Jackie Swapp, Xana Waughman, Ali Williford, Shawn R. Olsen, Christof Koch, Michael A. Buice, Saskia de Vries

Abstract

Extracellular electrophysiology and two-photon calcium imaging are widely used methods for measuring physiological activity with single-cell resolution across large populations of neurons in the brain. While these two modalities have distinct advantages and disadvantages, neither provides complete, unbiased information about the underlying neural population. Here, we compare evoked responses in visual cortex recorded in awake mice under highly standardized conditions using either imaging or electrophysiology. Across all stimulus conditions tested, we observe a larger fraction of responsive neurons in electrophysiology and higher stimulus selectivity in calcium imaging. This work explores which data transformations are most useful for explaining these modality-specific discrepancies. We show that the higher selectivity in imaging can be partially reconciled by applying a spikes-to-calcium forward model to the electrophysiology data. However, the forward model could not reconcile differences in responsiveness without sub-selecting neurons based on event rate or level of signal contamination. This suggests that differences in responsiveness more likely reflect neuronal sampling bias or cluster-merging artifacts during spike sorting of electrophysiological recordings, rather than flaws in event detection from fluorescence time series. This work establishes the dominant impacts of the two modalities’ respective biases on a set of functional metrics that are fundamental for characterizing sensory-evoked responses.

The Modality Mismatch in Systems Neuroscience

In systems neuroscience, conclusions are often tied to the sensor used. For decades, researchers have noted that extracellular electrophysiology (ephys)—recording electrical spikes—and two-photon calcium imaging (ophys)—recording fluorescent proxies of activity—frequently yield conflicting results. Ephys typically reports a large fraction of active neurons, while ophys suggests a smaller, more "selective" (picky) population. This study aims to determine if these discrepancies arise from biological effects, hardware physics, or data processing algorithms.

A Standardized Comparison Framework

To minimize experimental noise, the authors utilized the Allen Brain Observatory's standardized data pipeline. This allowed for a direct comparison of ephys and ophys data recorded from the same visual cortical areas in mice using identical visual stimuli and experimental rigs.

Standardized experimental workflow and recording rigs used to compare Neuropixels electrophysiology and two-photon calcium imaging.

The study leveraged two primary datasets: ~41,500 neurons from 170 mice (ophys) and ~11,000 neurons from 52 mice (ephys). A new validation dataset was also collected from GCaMP6f-expressing mice to ensure the calcium indicator did not fundamentally alter electrical behavior.

Visual stimuli (Gabors, flashes, gratings, etc.) and the structured session timelines used across both recording modalities.

Reconciling Discrepancies through Forward Modeling

The researchers employed "Forward Modeling" using the MLSpike biophysical model to transform ephys spike trains into "synthetic" calcium traces. This test determined if the slow, non-linear nature of calcium indicators was responsible for the perceived higher selectivity in imaging.

Key findings include:

  • Selectivity as an Optical Filter: Calcium indicators act as a non-linear sparsifying filter, amplifying bursts while missing single spikes. Forward modeling reduced the similarity gap (Jensen-Shannon distance) in selectivity distributions from 0.41 to 0.14.
  • The Responsiveness Gap: Forward modeling could not explain why ephys detects more responsive neurons. This discrepancy likely stems from sampling bias; ephys favors highly active neurons. Reconciling the two required sub-selecting only the top 7-39% of the most active ophys neurons.
  • Spike Sorting Artifacts: Signal contamination in ephys (merging spikes from multiple neurons) artificially inflates responsiveness and decreases selectivity. Units with zero detected contamination aligned much more closely with synthetic ophys data.

Conclusion and Future Directions

While functional preferences (e.g., preferred motion direction) are robust across modalities, metrics like responsiveness and selectivity are heavily influenced by recording bias. Future work with higher-density probes like Neuropixels Ultra and more simultaneous ephys/ophys recordings will be essential to further refine the "translation manual" between these two pillars of neuroscience.

Summary updated

Abstract

Abstract Extracellular electrophysiology and two-photon calcium imaging are widely used methods for measuring physiological activity with single-cell resolution across large populations of neurons in the brain. While these two modalities have distinct advantages and disadvantages, neither provides complete, unbiased information about the underlying neural population. Here, we compare evoked responses in visual cortex recorded in awake mice under highly standardized conditions using either imaging or electrophysiology. Across all stimulus conditions tested, we observe a larger fraction of responsive neurons in electrophysiology and higher stimulus selectivity in calcium imaging. This work explores which data transformations are most useful for explaining these modality-specific discrepancies. We show that the higher selectivity in imaging can be partially reconciled by applying a spikes-to-calcium forward model to the electrophysiology data. However, the forward model could not reconcile differences in responsiveness without sub-selecting neurons based on event rate or level of signal contamination. This suggests that differences in responsiveness more likely reflect neuronal sampling bias or cluster-merging artifacts during spike sorting of electrophysiological recordings, rather than flaws in event detection from fluorescence time series. This work establishes the dominant impacts of the two modalities’ respective biases on a set of functional metrics that are fundamental for characterizing sensory-evoked responses.
BibTeX
@inproceedings{siegle2021reconciling,
  author = {Joshua H. Siegle and Peter Ledochowitsch and Xiaoxuan Jia and Daniel Millman and Gabriel Koch Ocker and Shiella Caldejon and Linzy Casal and Andrew Cho and Daniel J. Denman and Séverine Durand and Peter A. Groblewski and Greggory Heller and India Kato and Sara Kivikas and Jérôme Lecoq and Chelsea Nayan and Kiet Ngo and Philip R. Nicovich and Kat North and Tamina K. Ramirez and Jackie Swapp and Xana Waughman and Ali Williford and Shawn R. Olsen and Christof Koch and Michael A. Buice and Saskia de Vries},
  title = {{Reconciling functional differences in populations of neurons recorded with two-photon imaging and electrophysiology}},
  booktitle = {eLife},
  year = {2021},
  doi = {10.7554/elife.69068},
}
Figure