Purpose

The mapping of brain activity across cortical layers is critical to investigate laminar processing (e.g., feedforward and/or feedback pathways) for understanding the information flow within brain functional network¹. Unlike widely-used gradient-echo (GE)-BOLD fMRI, CBV-weighted fMRI is specific to the neuronally active site across the cortical depths^2-4. However, its underlying neurophysiological source (local field potential or multi-unit activity) is unknown. Here, we investigated whether laminar fMRI responses reflect neuronal input- or output-driven activity by comparing somatosensory-evoked laminar CBV and BOLD activity with cortical depth-dependent electrophysiological data.

Materials & Methods

To examine the laminar specific fMRI responses and neural activities in somatosensory cortex of mice, three different fMRI studies were designed: 1) CBV-weighted fMRI, 2) BOLD-fMRI, and 3) multi-channel electrophysiological recordings.
All fMRI experiment were performed on 15.2T with high spatial resolution of 156×156×500μm³ under ketamine-xylazine anesthesia⁵. CBV-weighted fMRI data (n=22 mice) were acquired using gradient echo (GE)-based imaging (FLASH) with temporal resolution of 2s (TE=3ms) after the injection of monocrystalline iron oxide nanoparticles (MION, 45 mg/kg), while BOLD-fMRI data (n=7 mice) were acquired using single-shot GE-EPI sequence with temporal resolution of 250ms (TE=11ms)⁶.
To investigate the neural sources of laminar specific fMRI findings, electrophysiological data (n=6 mice) were recorded using a 16-channel electrode with 50μm inter-channel spacing. The electrode was inserted into the center of S1FL predefined by CBV-weighted optical imaging to a depth of 1 mm spanning most of the cortical depth.
For somatosensory stimulation, the left forepaws of the mice were electrically stimulated with current intensity of 0.5mA, pulse width of 0.5ms, and frequency of 4Hz. Functional trial consisted of a 60-s (30s for electrophysiology) prestimulus, 20-s stimulus, 60-s interstimulus, 20-s stimulus, and 60-s (30s for electrophysiology) poststimulus period, and 15 fMRI trials (5-8 trials for electrophysiology) were obtained for signal averaging.
To measure layer-specific fMRI responses, cortical areas responding to stimulation were flattened by radially projecting 22 lines perpendicular to the cortical edges⁷ (Fig.1A-i). The cortical depth profiles were resampled to double using bicubic interpolation, leading to a nominal resolution of 78µm (Fig.1A-ii). Group-averaged percent change maps were calculated, and cortical profile was obtained in the primary forelimb somatosensory area (S1FL) (bluish-ROI in Fig.1A-ii). Laminar boundaries were defined as a cortical thickness distribution based on the Allen mouse brain atlas.
To extract local field potential (LFP) and multiunit activity (MUA), the extracellular recording traces were band-pass filtered (1 < f < 280Hz) and high-pass filtered (> 300Hz), respectively. Then, LFP waveforms and spike trains during stimulation were aligned with a 250-ms moving window and then averaged (Fig.2A). To determine the location of layer 4 (L4), inversion current source density (iCSD) analysis was performed with LFP waveforms (Fig.2B).

Results

In the cortical depth-dependent fMRI responses, somatosensory-evoked CBV-weighted response peaked at input layer 4 (L4) of S1FL (≈ depth of 390~546µm; Fig.1B), while BOLD response peaked at the cortical surface (Fig.1C). CBV-weighted response became increasingly specific to L4 over time (Fig.1B-ii and iii), whereas the BOLD response first appeared in L4 (Fig.1C-ii and iii).
In the spatiotemporal dynamics of sensory-evoked neural activity, the synaptic inputs were observed as an evident current sink at depths of 0.4-0.56 mm from the surface (Fig.2B), indicating L4. The sensory-evoked LFPs showed the highest changes at L4 (Fig.2C), which is expected for thalamocortical inputs, while the earliest MUA responses occurred ~15ms after stimulus onset at L4 but were stronger in L5 later (Fig.2D).
To compare laminar fMRI responses with neurophysiology data, average magnitude of supragranular (L2/3), granular (L4) and infragranular (L5/6) responses were determined (Fig.3). CBV changes were the strongest at L4 (Fig.3A), which is somewhat consistent with synaptic input LFPs (Fig.3C), but not with spiking output activity MUA (Fig.3D). The BOLD changes were highest in the upper layer, where large draining veins are concentrated (Fig.3B), while the strongest MUA was observed at L5/6 (Fig.3D).

Discussion & Conclusion

We demonstrated laminar specific fMRI responses (CBV-weighted and BOLD) and cortical depth-dependent neural activity (LPF and MUA) in the mouse S1FL to investigate neuronal source of hemodynamic fMRI signals. Sensory neural information generated by forepaw stimulation are first projected through the ventral thalamus into L4 of S1FL and then rapidly spread to other layers⁸. In this study, somatosensory-evoked CBV responses peaked at thalamocortical input layer (L4) with high specificity over stimulus time, while early BOLD response was observed in L4, but later showed poor layer specificity on the cortical surface due to the spreading of non-specific pial vein contributions. This is consistent with previous fMRI observations that CBV responses were preferentially peaked in the layers in which synaptic activity is selectively induced, independent of the baseline blood volume³. Laminar profile of LFP showed the greatest response to sensory stimulation in L4, whereas the evoked MUA showed highest neuronal firing in L5/6, where the thick-tufted pyramidal cells are located⁹ and provide the major cortical output¹⁰. Since LFP and MUA reflects the input to neuronal network and spiking output, respectively¹¹, the laminar specific CBV results suggest that the blood supply is tightly regulated by microvessels coupled to synaptic input-driven activity, rather than the output response.

Acknowledgements

This work was supported by IBS-R015-D1.