Do olfactory bulb projection targets determine what odors mice encode?

Tracing output neurons by where their axons end

The study asked whether the olfactory bulb sends different information to different cortical areas, rather than treating all output neurons as one population. To separate neurons by destination, the authors injected a retrograde viral tracer carrying the calcium reporter GCaMP6f into either the AON or the APC of Pcdh21-Cre mice, where the Cre line restricts reporter expression to olfactory bulb output neurons. Because cortical cells in these areas do not express Cre, labeling stayed confined to bulb neurons whose axons reached the injection site, avoiding labeling of top-down fibers.

This design matters because mitral cells (MCs) and tufted cells (TCs) both project to the AON and APC, with overlapping rather than absolute segregation. The authors confirmed injection sites histologically and kept only correctly targeted mice. They then imaged odor-evoked calcium signals in the bulb, first at the population level with widefield imaging and then at single-cell resolution with in vivo two-photon microscopy at depths of 150 to 500 µm across the dorsal bulb surface.

Population odor maps do not separate the two targets

At the population level, the two projection-defined streams looked remarkably alike. Peak odor-evoked GCaMP6f signals ranged from 24.5% to 129.7% ΔF/F for APC-projecting populations and 29.2% to 129.9% ΔF/F for AON-projecting populations, with no significant difference. Autocorrelation widths of the response maps were also statistically indistinguishable (APC-traced 158.5 ± 32.1 µm, n = 5 mice; AON-traced 157.7 ± 31.6 µm, n = 6 mice; p = 0.992), indicating a similar diffuse component.

Inhalation-linked kinetics followed the same pattern: amplitudes and latencies differed strongly between anterior and posterior bulb regions, but not between AON- and APC-traced populations at matched positions. This is an important negative result. It means the projection-specific differences reported later are not simply a consequence of where the labeled cells sit along the bulb's spatial axes, and they are not visible in coarse population maps. The informative comparison required resolving individual cells.

AON-projecting cells respond to more odors

Single-cell two-photon imaging exposed a first functional difference. A similar fraction of AON- and APC-traced mitral/tufted cells responded to at least one odor in an eight-odor panel at 1% vapor pressure (74.9%, 170 of 225 cells for AON; 70.2%, 158 of 227 cells for APC; p = 0.82). But on a cell-odor basis, AON-traced cells produced a significantly larger share of significant responses: 38.6% of cell-odor pairs (696 of 1816) versus 28.2% (507 of 1800) for APC-traced cells (p < 0.001).

The authors note that inhibitory responses were rare, seen in only 2.5% of AON-traced neurons and none of the APC-traced neurons, so all further analysis used excitatory cell-odor pairs. That restriction is a real boundary: the paper cannot rule out that inhibitory differences between the populations exist. Within the excitatory data, the broader responsiveness of AON-projecting cells is the first sign that the two streams carry differently shaped odor information.

APC-projecting cells are more selective for odor identity

Selectivity was measured with lifetime sparseness, which describes how narrowly a single cell responds across odors, and population sparseness, which describes how sparse the population response is to individual stimuli. Both differed between projection-defined populations, with APC-traced cells showing higher selectivity than AON-traced cells (lifetime sparseness p = 0.002; population sparseness p = 0.015, Mann–Whitney U tests). The cumulative probability of responding to many odors was markedly higher in AON-traced cells.

The paper's summary is that AON-projecting cells are more broadly tuned, while APC-projecting cells are more selective. This is a comparison of tuning breadth, not a claim that one stream encodes identity and the other does not. The distinction matters for readers: the evidence supports a difference in how sharply odors are separated across the responsive population, not an absolute division of labor between the two cortical targets.

APC-projecting cells track concentration more strongly

When odors were presented from 0.1% to 4% saturated vapor pressure, a similar fraction of AON- and APC-traced cells responded to at least one concentration (58.7%, 105 of 179 for AON; 59.9%, 109 of 182 for APC; p = 0.99). On a cell-concentration basis, however, APC-traced cells had a significantly higher share of significant responses (33.04%, 421 of 1274) than AON-traced cells (25.9%, 325 of 1253; p = 0.001).

Response amplitude grew with concentration in both populations, but far more steeply for APC-traced cells: a 189.09% mean normalized amplitude increase from lowest to highest concentration, versus 33.01% for AON-traced cells. The area under the concentration-response curve was also higher for APC-traced neurons (504.18 ± 1049.26 versus 347.35 ± 354.76; p < 0.001). Differences were clearest at low concentrations, where only APC-traced cells showed significant amplitude changes between 0.1% and 0.25% or 0.5%. The authors interpret this as two different strategies for transmitting concentration-dependent signals rather than one stream simply being more sensitive.

Projection target classifies better than cell type

Because MCs and TCs have their own response differences, the authors checked whether cell-type composition could explain the target-dependent results. Dual-color tracing from AON and APC showed that the populations were separable but not pure: 49.2 ± 7.4% of MCs were double-labeled, versus 28.9 ± 5.9% of TCs. Most AON-traced cells (80.6%, 79 of 98) sat in the external plexiform layer, while most APC-traced cells (82%, 55 of 67) sat in the mitral cell layer, confirming target-dependent segregation.

Removing double-labeled cells sharpened the AON-versus-APC differences in lifetime and population sparseness, which the authors read as evidence that singly projecting neurons carry highly distinct coding properties. Splitting the traced populations by cell type also produced a clear pattern: APC tufted cells showed the strongest concentration-dependent response changes, and AON tufted cells showed very promiscuous odor tuning. Critically, both MCs and TCs showed similar target-dependent shifts, so cell identity alone did not account for the effect. A support vector machine classifier reinforced this: cross-validated accuracy for AON versus APC was 0.65, clearly above the 0.5 chance level, while classifying MCs versus TCs irrespective of target fell near chance.

A projection-defined split, with stated boundaries

The paper's conclusion is that olfactory bulb output is segregated not only by mitral versus tufted identity but also by innervation target, setting up parallel processing before information reaches higher olfactory cortex. The strongest single comparison is the classifier result: projection target carried more separable information than cell type, and the weakest classification was APC mitral versus APC tufted cells, where accuracy dropped to chance. That ordering, rather than any single tuning statistic, is the paper's most decision-relevant finding for readers comparing routing hypotheses.

Several boundaries travel with that claim. The work is in anesthetized mice with an artificial inhalation paradigm, uses a limited odor panel and concentration range, and analyzes excitatory responses because inhibitory responses were too rare to compare. The tracing is also not absolute: roughly half of MCs and about a third of TCs were double-labeled, so the two streams overlap anatomically. The authors' interpretation is that the bulb already separates sensory information by target; the data establish a strong functional difference between projection-defined populations, not a complete anatomical partition or a demonstrated behavioral consequence.

Sources used in this answer