Week 1: Multiple-Photon Microscopy

What does intrinsic signal imaging actually measure, and why is that "indirect"?

It measures brain activity through blood flow, not through electrical activity itself. It's indirect because it relies on the vascular response neurons trigger, rather than recording their firing directly.

Walk through the chain of events from "neuron becomes active" to "blood flow changes."

Active neuron → increased metabolic/energy demand → more O2 consumed → local vasodilation (blood vessels widen) → more blood pulled into that region. So activity is inferred from a downstream vascular consequence, not measured directly.

What crosses the blood-brain barrier during neuronal activity, and where does it end up?

Oxygen crosses blood-brain barrier + is consumed by neurons; converted into CO2, which is then exhaled. This O2-for-CO2 exchange is the metabolic cost of firing that intrinsic imaging ultimately detects.

Why can you optically distinguish oxygenated from deoxygenated blood?

Because HbO2 (oxygenated) and Hb (deoxygenated) hemoglobin have different light absorption spectra — at certain visible wavelengths their absorption differs sharply, so measuring reflected light at those wavelengths tells you the local oxygenation state.

Which is more opaque to light — oxygenated or deoxygenated hemoglobin — and why does that matter?

Deoxygenated Hb absorbs light strongly (very opaque); oxygenated HbO2 absorbs much less. This differential absorption is the actual physical signal intrinsic imaging detects — more deoxy-Hb means more light absorbed, less reflected.

What does it mean to image at a wavelength where Hb and HbO2 absorb equally?

At that "isosbestic" wavelength, the signal no longer reflects oxygenation — instead it becomes proportional to total blood volume, since both hemoglobin forms absorb the same amount regardless of their ratio.

What do orientation maps and retinotopic maps (from intrinsic imaging) reveal about the cortex?

They show that neurons respond selectively to specific stimulus features (e.g. edge orientation, position in visual field), and that this selectivity is spatially organised into map-like patterns across the cortical surface — not randomly scattered.

Why is intrinsic signal imaging good for maps but bad for tracking individual neurons quickly?

Because it's slow (vascular responses take seconds) and has poor depth resolution (signal is averaged over hundreds of µm), it can only reveal large-scale, spatially averaged functional organisation — not fast, single-cell dynamics.

In wide-field imaging through a mouse's thinned/opened skull, what do bright white regions and "interrupted" vessels indicate?

Bright/white regions are dense with nerves and vasculature; an "interrupted" artery means the vessel is diving directly into the tissue below (a penetrating vessel), not that blood flow has actually stopped.

Why do the fast fluorescence spikes and the slow reflectance dip in wide-field recordings look so different in speed?

The fluorescence (calcium) signal reflects fast electrical/neuronal activity (millisecond scale), while the reflectance signal reflects the slower hemodynamic response (second scale) — this mismatch is a direct illustration of why vascular signals lag far behind neuronal ones.

What's the fundamental trade-off of wide-field imaging?

You can map which broad brain regions respond to which stimuli, but spatial resolution is low — you're imaging a wide field of tissue at once, not resolving individual cells within it.

Why does resolving smaller structures require shorter wavelengths of light?

Because of the diffraction limit — a light microscope's resolving power is fundamentally tied to wavelength, so finer detail can only be resolved with shorter wavelengths.

What are the two key numbers to remember: light microscope resolution, and typical neuron cell body size?

Conventional light microscope resolution ≈ 0.2 µm; typical neuron cell body ≈ 10–20 µm. So a light microscope has more than enough resolution to see a whole cell body clearly, but seeing sub-cellular detail pushes toward that 0.2 µm limit.

What's the problem with focal depth in standard wide-field microscopy, and why does it matter?

A wide-field image is a "smushed" combination of light from many depths at once — light from above and below the true focal plane still contributes to the image. This blurring, with no true optical sectioning, is exactly the problem confocal microscopy was invented to solve.

In one sentence, how does confocal microscopy change the approach compared to wide-field?

Instead of passively collecting whatever light comes back from a broadly illuminated volume, confocal actively shapes both the illumination and the detection path so only in-focus light is recorded.

What is the pinhole doing in a confocal microscope, mechanistically?

It sits in front of the detector and physically blocks light that isn't coming from the focal plane — because out-of-focus light diverges more by the time it reaches the pinhole's plane, it gets blocked, while in-focus light converges through the small aperture. This minimises focal depth and maximises spatial resolution.

How do you choose a fluorescent label for a confocal experiment, and why does the label's identity matter?

You pick a dye/protein (e.g. GFP) that's known to bind or attach to your structure of interest, and whose excitation/absorption wavelength you know — that's what tells you which laser wavelength will actually excite it.

Why must confocal microscopes scan point-by-point rather than capturing a whole image at once?

Because only one point is illuminated and detected at a time (that's the whole point of the pinhole approach) — you scan along a line, then step laterally, gradually building up a full 2D (and with depth, 3D) image pixel by pixel.

What sets the lateral resolution of a confocal scan?

How finely you sample — i.e. how many pixels you collect along the scanned line. Denser sampling = finer lateral resolution, at the cost of scan time.

List confocal's three core requirements, and explain why each is needed.

(1) Strong laser light — because the pinhole throws most photons away, so you need enough photons to begin with; (2) point-by-point scanning — because only one spot is imaged at a time; (3) computer power — to control the scan and digitally reconstruct the image from all the individual points.

Why is confocal microscopy inherently slow, and how does this get worse for 3D imaging?

The image is built pixel by pixel, so scanning already takes time; adding more depth slices (for a 3D z-stack) multiplies the number of points that must be scanned, so time resolution degrades further the bigger/deeper the volume you want.

Why can confocal microscopy damage the sample, and what's the trade-off with laser power?

Higher laser power gives a stronger, cleaner signal — but sustained high power can burn or photobleach the sample (destroying the fluorophore or damaging tissue), so the improvement is short-lived. There's a hard trade-off between image quality and specimen survival.

What is confocal's practical depth limit in brain tissue, and what physically causes that limit?

About 100 µm. Beyond this, light scattering in the tissue disrupts both directions of the light path: excitation photons get deflected away from the intended focal spot, and emission photons get scattered on their way back out.

Why does scattering specifically break confocal's pinhole strategy, even for correctly-focused light?

Because the pinhole can't tell the difference between "genuinely out-of-focus light" and "light that was correctly generated at the focus but got scattered on its way out." Both arrive at the pinhole from the wrong angle/position and get rejected — so confocal loses real signal in scattering tissue, not just noise.

Why does normal (single-photon) excitation generate fluorescence outside the focal plane too?

Because the excitation beam is cone-shaped — it's bright enough to excite fluorescence anywhere along that cone, above and below the true focus, not just at the narrow tip where it's most concentrated.

What is multiphoton microscopy's core strategic difference from confocal's pinhole approach?

Instead of generating fluorescence everywhere and then filtering it after the fact (confocal's pinhole), multiphoton microscopy restricts the excitation itself to a tiny focal volume in the first place — so no filtering is needed, and all collected emission light can be assumed to be real signal.

Mechanistically, what has to happen for a molecule to be excited via multiphoton absorption?

Instead of absorbing one high-energy photon, the molecule absorbs several lower-energy photons nearly simultaneously, and the total energy delivered equals the sum of their individual energies — e.g. two long-wavelength (infrared) photons together supplying the energy of one shorter-wavelength photon.

Why is multiphoton absorption described as a "nonlinear" process, and what does that have to do with confinement?

probability depends on square of light intensity + it requires an extremely high density of photons.

laser intensity drops off sharply away from the focal point, the threshold for absorption is only met at the exact geometric focus, eliminating out-of-focus excitation above + below that tiny volume

Why is a Ti:Sapphire pulsed laser essential for multiphoton excitation, rather than a continuous laser?

It delivers very short, high-intensity pulses, creating brief bursts of extremely high instantaneous photon density (needed for multiphoton absorption to occur) while keeping the average power low — protecting the tissue from the sustained heating/damage a continuous high-power beam would cause.

Give two separate reasons multiphoton microscopy causes less tissue damage than confocal.

(1) Excitation only happens momentarily, at the focus, during each ultrashort pulse — the tissue isn't continuously bombarded with high-density photons; (2) at all other depths/locations along the beam path, photon density is too low to cause any excitation at all, so surface tissue in particular is spared.

Why does multiphoton microscopy penetrate deeper into brain tissue than confocal?

It uses infrared excitation light, which is both less absorbed and less scattered by brain tissue than the visible light confocal uses — infrared light simply travels further through tissue before being lost.

How does the dichroic mirror exploit the wavelength difference between excitation and emission in multiphoton setups?

Since excitation is infrared and emission is visible, the dichroic mirror is built to pass infrared straight through while reflecting visible light sideways to the detector — cleanly separating the two paths without extra filtering steps.

On the Jablonski diagram, what are the two routes to reach the excited state S1/S2, and roughly how fast is each?

Absorbing a single photon (a fast, essentially instantaneous/femtosecond event), or absorbing two photons nearly simultaneously (multiphoton route) — both reach comparable excited states, just via a different number of photons.

What are 1 femtosecond and 1 attosecond, and why does the lecture bother naming these units?

1 femtosecond = 10^-15 s; 1 attosecond = 10^-18 s. They're introduced because multiphoton absorption and pulsed-laser timing operate on the femtosecond scale — you need units this small to even talk meaningfully about pulse duration and absorption timing.

Explain, step by step, why pulsed (not continuous) lasers are needed for efficient 2-photon absorption.

2-photon absorption probability scales w/ square of instantaneous intensity. If you spread a fixed number of photons evenly over time (continuous laser), instantaneous intensity stays low + 2-photon events are rare. If you concentrate the same photons into brief, intense pulses, instantaneous intensity spikes during each pulse, boosting 2-photon absorption; while damage + power stays the same/lower.

Describe what you'd actually see with single-photon excitation of a fluorescent sample in a cuvette, and why.

A visible glowing cone or line running the whole length of the beam through the sample — because the beam is bright enough to excite fluorescence everywhere along its path, not just at the focus.

Describe what you'd see with two-photon excitation of the same sample, and explain the contrast with single-photon.

Just a single tiny bright point, exactly at the focal plane — because only there is the photon density high enough for two photons to be absorbed nearly simultaneously; everywhere else along the beam, density is too low for any excitation to occur at all.

Why does multiphoton microscopy not need a pinhole, given what you know about where fluorescence is generated?

Since fluorescence can only be generated at the tightly confined focal volume (not spread along the beam), any photon reaching the detector — however it travelled to get there — can be assumed to have originated at the focus. There's nothing "out of focus" left to filter out, so light can be collected broadly and efficiently instead of being thrown away by a pinhole.

Explain "spatial compression" vs "temporal compression" of photons in multiphoton microscopy, and how they work together.

Spatial compression: objective lens focuses a wide cone of light into tiny spot, concentrating photon density in space. Temporal compression: the femtosecond laser delivers photons in brief/dense bursts instead of steady stream, concentrating photon density in time. Both increase instantaneous photon density + needed at the same time for absorption efficient at focal point.

Why does a multiphoton objective need a high numerical aperture (NA), and what two separate benefits does this give?

(1) Resolution: high NA focuses excitation light onto a smaller spot, reduces focal depth (tighter PSF); (2) light collection: high NA means a wider aperture angle, which captures more of the scattered emitted fluorescence coming back out of the tissue.

Why does a multiphoton objective also favour low magnification, and how does that help?

Low magnification gives a large field of view, and typically comes packaged with a large lens diameter — which, like high NA, improves light collection by physically capturing more of the scattered emission light exiting the tissue.

Why is an objective's light-collecting power just as important as its focusing precision in multiphoton microscopy?

Because multiphoton relies on gathering all emitted photons from the tissue (no pinhole to discard "extra" light) — so any photon the objective fails to capture is signal permanently lost, making collection efficiency a first-order design concern, not an afterthought.

What's the difference between "filling" and "underfilling" the objective's back aperture, and what does each produce (confocal + multiphoton)?

Filling: the laser beam diameter matches or exceeds the back aperture, using the objective's full NA and producing a small, tightly confined focal spot. Underfilling: the beam is narrower than the aperture, so only part of the NA is used, producing a larger, less well-confined (poorer resolution) focal spot.

Given the filling/underfilling trade-off, why isn't "always fill completely" the obvious answer?

lens can’t transmit whole wider beam, so aperture cuts off outer parts.

less laser power reaches sample.

system must balance sharper focal spot against loss of laser power.

What happens to emitted fluorescence photons as they travel through scattering tissue on their way to the objective?

They get deflected multiple times, spreading out over a progressively wider angle and area the further they travel — by the time they reach the objective, some fall outside its collection angle and are lost unless the NA/lens diameter is generous enough to still capture that spread.

What are the two distinct physical sources of light scattering in brain tissue, and which matters most for in vivo two-photon imaging?

(1) Refractive index variations within the tissue itself, and (2) scattering off blood cells. The second matters especially for in vivo two-photon imaging because the cortex is so densely vascularised — blood cells scatter excitation photons on the way in and emission photons on the way out.

Below what wavelength do hemoglobin forms dominate light absorption in tissue, and what happens above it?

Below ~650 nm, oxy- and deoxy-hemoglobin dominate absorption; above that, moving into the near-infrared, absorption by tissue chromophores generally drops — which is exactly why Ti:Sapphire's near-infrared output penetrates tissue so much better.

Compare absorption vs scattering of blood in blue-green light versus red light.

Blue-green: absorption and scattering are of roughly similar strength — both contribute comparably to loss. Red/NIR: absorption drops to a much lower level while scattering stays comparatively high and flat — so in the red/NIR range, scattering (not absorption) becomes the dominant obstacle.

Trace the excitation light's path through a multiphoton setup, component by component.

Ti:Sapphire laser generates the pulsed beam → intensity control (neutral density filter or Pockels cell) attenuates power → beam expansion (scan lens + tube lens) widens the beam to fill the objective's back aperture → scan mirrors steer the beam in x-y → dichroic mirror passes the infrared beam through → objective focuses it into the sample.

Why does the beam need to be expanded before reaching the objective?

So that it properly fills the objective's back aperture (recall: filling, not underfilling, gives the tightest, best-resolved focal spot) — the scan lens and tube lens jointly perform this expansion.

Name the two ways to control/attenuate laser intensity, and how do they differ?

A neutral density filter (a simple, static, passive way to reduce power) or a Pockels cell (an electro-optic device that allows fast, computer-controlled, dynamic modulation of intensity during scanning).

Name the three types of scan mirror technology and one distinguishing feature of each.

Galvo scanners (galvanometer-driven mirrors, standard/flexible), resonance scanners (oscillate at a fixed resonant frequency, much faster, used for high-speed line scanning), acousto-optical deflectors/AODs (deflect light via sound waves in a crystal, enabling very fast, random-access, non-raster scanning of arbitrary points).

What does the dichroic mirror separate in the detection path, and how?

It separates the infrared excitation beam (which passes straight through) from the visible emitted fluorescence (which is reflected off toward the detectors) — exploiting the wavelength gap between excitation and emission.

Why are photodiodes/photomultiplier tubes used as detectors rather than a standard camera sensor?

Because emitted fluorescence signal is often very weak, so you need detectors with especially high sensitivity/efficiency at collecting low light levels — PMTs in particular offer high gain for this purpose.

Compare the three 2D laser scan patterns: raster, line, and random access.

Raster scan: back-and-forth lines covering whole plane (full spatial coverage, slower). Line scan: repeatedly scanning along single fixed line (fast temporal resolution, only 1D spatial info). Random access: jumping between specific pre-selected points of interest, skipping empty space entirely (fast + targeted, but only samples chosen points).

What's the difference between "3D regular" and "3D free" scan patterns, and when would you use each?

3D regular: the same scan types (raster/line/random-access) repeated systematically across multiple stacked depth planes — good for structured volume coverage. 3D free: arbitrary, flexible trajectories through a volume that follow the actual 3D shape of the structure being imaged (e.g. a winding dendrite) — better when you only care about specific real structures, not a whole regular grid.

How do you build a full 3D image with a point-scanning multiphoton microscope?

Combine lateral (x-y) scanning at a given depth with mechanically moving the objective up/down (axial, z) to shift the focal plane — repeating the lateral scan at each depth builds up a full 3D volume.

What physically happens when a neuron fires an action potential, and how does an electrode detect it?

A voltage pulse is generated and propagates down the neuron toward its synapses; an electrode detects this because ions moving across the membrane carry an electric charge, which the electrode picks up as a change in voltage/current.

What is the physical/chemical basis that lets calcium dyes report neuronal firing optically?

Every time a neuron fires, calcium ions enter the cell; calcium-sensitive dyes change their fluorescence (absorption/emission properties) in response to this rising intracellular calcium concentration — so a calcium influx becomes a visible change in brightness.

What are the two ways of getting a calcium indicator/sensor into neurons, and what's the key advantage of the second method?

(1) direct injection of a synthetic dye into the tissue; (2) genetically modifying animal so that cell type expresses a genetically encoded calcium indicator; advantage=selectively label only defined cell population, rather than whatever happens to take up the injected dye.

Why does calcium imaging show activity as distinct "packets" or transients, rather than one continuous signal?

Because calcium dyes have relatively fast (though not instant) binding and unbinding kinetics — each burst of calcium entry produces a rise-then-decay fluorescence transient, so individual bouts of electrical activity appear as separable events over time.

What does drawing a region of interest (ROI) on an imaging dataset let you do, and why is that useful?

It lets you convert the 2D (or 3D) fluorescence data within that region into a single one-dimensional signal over time — turning a complex image into a simple time-series that's much easier to analyse (e.g. detecting calcium transients).

Why is a chronic (long-term) preparation necessary for studying things like synaptic change over time, instead of an acute one?

Because tracking the same neurons or synapses over days to months requires being able to return to the exact same field of view repeatedly — this needs a stable, long-lasting optical window into the brain, which an acute (one-time, short-duration) preparation can't provide.

What do the bright puncta seen along a GCaMP6-expressing dendrite represent, and why does imaging them matter?

They represent dendritic spines — small protrusions where excitatory synapses are located. Imaging them at high enough resolution and speed lets you watch synapse-level (spine-level) calcium activity change in real time, revealing fine subcellular functional detail beyond just whole-cell activity.

Why can genetically encoded calcium sensors label large neuronal populations more easily than injected dyes?

Because they're delivered via genetic targeting rather than physical injection, they can be expressed broadly + non-invasively across a whole defined cell population, enabling large-scale simultaneous imaging of many neurons at once.

Structurally, what is a genetically encoded calcium indicator, and how does it produce a fluorescence change?

fusion protein combining a fluorescent protein w/ a calcium-binding domain; when calcium binds, the protein's conformation changes, which changes its fluorescence — directly linking calcium concentration to optical signal

What is channelrhodopsin, where does it come from, and what makes it useful to neuroscientists?

light-activated ion channel in green algae. It's useful because, unlike a neuron's normal voltage-gated channels, it opens in response to light rather than voltage — giving an artificial/externally controllable "on-switch" for neurons.

Describe exactly what happens ionically when channelrhodopsin is exposed to blue light.

The channel opens, allowing Na+ ions to flow into the cell (increasing/depolarising the membrane voltage) while K+ ions flow out — together producing a measurable electrical current and pushing the neuron's voltage in the depolarising direction, toward firing.

Explain how normal voltage-gated ion channels work, and contrast this with channelrhodopsin.

Voltage-gated channels open or close depending on the membrane's own voltage — sodium channels opening depolarise the cell (Na+ in), potassium channels opening repolarise it (K+ out) back toward equilibrium. Channelrhodopsin instead responds to light, not voltage — letting an experimenter trigger the same kind of ion flow artificially, on demand, regardless of the cell's natural voltage state.

Break down the term "optogenetics" into its two parts and explain each.

"Opto" refers to using light-activated proteins (like channelrhodopsin) to control cells; "genetic" refers to targeting the relevant DNA specifically to genetically defined cell types — so only the cells you've genetically chosen express the light-sensitive channel and respond to light.

In an all-optical stimulation-and-imaging experiment, what does each of the two laser paths do?

A stimulation laser path is tuned to activate an opsin (e.g. channelrhodopsin) in specific targeted cells. A readout path reads out activity via a calcium sensor across the same field of view. Together they let you both control + observe neurons optically.

In the combined stimulation+imaging experiment, how do you know the stimulation actually worked?

Each time targeted cells are stimulated → clear increase in intracellular Ca²⁺ shortly after; nearby cells that weren’t targeted show little/no change in Ca²⁺; so, stimulation successfully activated intended cells without strongly affecting neighbouring cells.

What is the single biggest advantage of combining optogenetics with calcium imaging, compared to classic electrophysiology?

You can selectively activate individual, genetically defined cells with light and simultaneously confirm their activation optically — achieving both precise manipulation and readout without inserting any physical electrodes into the tissue.

In the brain's hierarchy of spatial scales, what does "coupling to behaviour" mean, direction-wise?

It means moving upward from neural circuits toward larger, more integrative scales — maps, then whole systems, then ultimately observable behaviour.

In the brain's hierarchy of spatial scales, what does "coupling to elements" mean, direction-wise?

It means moving downward from neural circuits toward their smaller constituent parts — individual neurons, then synapses, then molecules.

Why do neural circuits sit at the conceptual centre of the brain's scale hierarchy?

Because they must be understood in two directions at once: in terms of the elements that compose them (synapses, neurons, molecules) below, and in terms of how they give rise to larger-scale phenomena (maps, systems, behaviour) above — making circuits the pivotal, hardest-to-isolate level of analysis.

Ultimately, why is understanding the brain so difficult, according to this scales argument?

Because no single technique or single scale of analysis is sufficient on its own — real understanding requires methods that can span many orders of magnitude in both space (molecules to whole brain) and time (milliseconds to months), which no one tool currently does alone.