Chapter 2 – Fiber Photometry Recordings
If you've ever found yourself in a standoff with a fiber photometry setup (and let’s be honest, we’ve all been there), you know it's equal parts science and sorcery. Hi, I’m Jacob, and after 600+ hours of wrangling dLight recordings for my PhD, I’m here to share some hard-earned lessons. As every scientist knows, these are mostly from things going off the rails.

 

The Aha Moment: Why Fiber Photometry Simplified My Life
I like to think that the genius who invented fiber photometry (Ludwig Maximilian’s group was the first to publish the technique in 2005) had an ‘aha’ moment. They probably spent hours at the surgery station prepping animals for imaging GECIs and asked themselves: 'Wait, why am I making this so complicated when I could just beam light through a fiber optic to get what I need? Less mess, more science.'

I don’t know if this was the motivation for inventing fiber photometry, or if the researchers were intentionally looking for a way to monitor GECIs during freely moving behaviors. To be honest, it’s hard for me to tell. I would have expected fiber photometry to be published as its own “Nature Methods/Protocols” paper, but the actual technique is was first shared in a Brief Communication in Nature Neuroscience titled: “Cortical calcium waves in resting newborn mice.” Fiber photometry is mentioned in the first sentence of the abstract, but I did not expect that from the title.

 

In any case, this invaluable technique gave researchers the ability to monitor GECIs from a larger brain volume of tissue using simpler methods than they were previously. This technique has the added advantages of being able to monitor multiple volumes of tissue simultaneously in freely moving animals.
 

 

Oh, The Sensors You'll Record
After three years of chasing signals with fiber photometry, I’ve learned a few things—like how indicators can be your best friend, and sometimes your most annoying lab partner (looking at you, aberrant microwaves following AAV expression of indicators – shown with GCaMP, please let me know if this has been shown with other indicators too).

I’ve mostly recorded dLight for dopamine (some GRAB-DA), but I’ve also recorded:

  • nLight/GRAB-NE – norepinephrine
  • iSeroSnFR – serotonin
  • cLight – CRF
  • GCaMPs – calcium
  • deltaLight – endogenous opoiods that bind to the delta opioid receptor

Each of these sensors have super different response properties, especially in vivo – mostly because each of the ligands these sensors measure operate on super different timescales. Truth is, it doesn’t matter if you are using Fura-2, jRGECO, beta-testing one of the new Tian lab sensors, or even using a FRET sensor, the technique is always the same.

 

 

How to do a fiber photometry experiment.
1. Pick the right wavelength for your sensor.
You can think of it like tuning a guitar—hit the wrong note, and the song gives you a headache instead of a rush of dopamine.

You can get a good approximation of the optimal wavelength from fpbase.org.
Note: every sensor alters the structure of the fluorescent protein to some extent, so the excitation-emission curve of dLight is different from GCaMP, even though they use the same cpGFP.


2. Decide on the best control for your experimental setup and conditions.
I have some strong opinions about this, but I’ll present all the options.

 

a. Isosbestic channel.
The isosbestic channel is like the safety net of your experiment—it’s there to keep things stable, but it only works if you’ve set it up just right. Miss the correct wavelength by a few nanometers, and it can betray you, turning your controls into chaotic messes that ruin your actual signal.

A hair’s difference (not literally since a hair is >50,000 nanometers thick) between isosbestic used and the sensor’s true isosbestic wavelength and end up with ligand-dependent signal (or inverse signal) in your isosbestic channel. This is problematic when you go and do isosbestic correction in your analysis, as you’ll end up removing/inflating some of your ligand-dependent signal. That’s why I have strong feelings about controls.




b. Non-binding version of your sensor.
In the Tian lab we have “0” versions of all our sensors, which is a sensor with a “broken binding pocket.” Ligand doesn’t bind, so there is no ligand-dependent signal from this sensor.

Theoretically, this sensor should be mostly flat, but I’ve never experienced that. Instead, I get super stereotyped responses that are animal/brain region dependent; I don’t see any correlation between signals from the same brain regions in different animals or even between different brain regions in the same animal.

I spent months racking my brain on why that might be. If you’ve ever found yourself cursing at the 0 version of a Tian lab sensor, trust me, you’re not alone. Here’s what I think is going on: these sensors are too f***ing bright.

The 0 variants haven’t been optimized to get the basal brightness levels down. The brightness isn’t changing from any ligand signaling. Therefore, these sensors are super sensitive to processes that impact photon efficiency (like hemodynamics).

If there is a large blood vessel near my implant, then my implant is basically giving me “inverse-BOLD” signal since my recording is going to get darker when blood flows into the brain region because it’s more active and get brighter as is returns to that f***ing insane level of basal brightness.

^ I don’t really have a good way to test this hypothesis without a mouse MRI machine. I tried sticking implants in the brain without putting any sensor, and I observed a similar phenomenon – very stereotyped responses from each recording site. If you have ideas, please send them my way.


 

c. "Inert” fluorescent protein.
From data I've seen, this is probably the best control, but it is not one I have used before. (the CMOS-based photometry system I use has a hard time multiplexing different color). Basically, you inject two viruses: your sensor and a different colored fluorescent protein. (There are ways to optimize expression of the two. For example, you can use green membrane-localized dLight, and cytosol-localized mApple.) The second fluorescent protein should have a relatively low baseline fluorescence, such that the only changes in fluorescence intensity will be motion dependent.

There might be some ligand-dependent changes in signal, but those would mostly be due to changes in the environment surrounding the proteins (pH, degree of quenching, etc) that would be reflected in changes in their lifetime signals – not something that most people are measuring.

I first came across this method when I was looking for photometry analysis techniques, and came across this awesome Github page by Thomas Akam at Oxford. I haven’t seen this method used in too many other places, but I’ve spoken about it with Lin Tian at length, and we agree it’s probably the best control out there. I will write an addendum to this post after I’ve done some of my own recordings with this technique.

  1.  
  2.  

3. Setting up your photometry recording equipment.
Doesn’t matter if you are using CMOS-based detection with interleaved excitation and detection or continuous excitation with silicon-photodetector-based detection using lock-in modulation. You need to direct your excitation light through I dichroic mirror such that it will excite your sample, and the emitted light will travel through the dichroic to the photodetector. Just make sure you are using narrow bandwidth filters to keep everything close to your peak excitation/emission.


The simplest and cheapest commercially available systems will take care of all these concerns you (RWD and Doric). If you’re a person that likes to “get under the hood” a bit or want a system that allows maximum flexibility for recording new fluorescent proteins in the future, there are a couple of good options for you. The RZ10X from Tucker Davis Technologies uses “plug-n-play” LEDs, filters, and dichroics, so you can tailor things to your needs. ThorLabs offers “Kinematic” cubes and sells individual components to give you maximum customization and control over your system.
 

 

4. Coupling the ferrule patch cord to the ferrule of the implant.

The actual mechanics of this step depend on your implant and patch cord type, but this is the most important step of the recording. If the patch cord doesn’t stay perfectly coupled to the implant throughout the duration of the recording, then you will get signal loss/reduced photon efficiency.

I use patch cords/implants with metal or ceramic ferrules, and I use bronze sleeves to couple the implants to the patch cords. I find these sleeves keep the connection tight between the patch cord and the implant. I even have a hard time removing these sleeves at the end of my recordings sometimes.

I usually fill the patch cord with optical coupling gel before attaching it to the implant. At first, I believed it helped keep the light path intact in the event of patch cord slipping/rotating as the animal moves around the cage. I’m starting to think this might fall under the category of “super-secret handshakes you should do to get your experiment to work,” but it doesn’t do me or the animals any harm to use it, so I’m going to continue using it.

5. Doing the recording.

At this point, you have done everything you can to get the best recording you can, and it’s “up to the software.” I’ve only used Doric Neuroscience Studio for photometry recordings since I’ve always used a Doric system, but this is some messy software. It crashes all the time, it has tons of bugs, and it eats up all the memory on my computer.
 

I haven’t tried any alternatives because Doric keeps rolling out updates, and I keep on having faith that they are going to solve the problems. At this point, I’m just trying to get enough recordings to graduate. If the software crashes during my recording, then I’ll just record another day (if my experiment allows for this – some of the experiments I do have a “novelty” component, so I obviously cannot redo these recordings).
 

 

Conclusion
That’s enough rambling from me for one post. I was tempted to dive into analysis too, but trust me, it deserves its own chapter. Until then, I’d love to hear your photometry war stories—what’s worked, what’s failed miserably, and where you’ve cursed your setup like me. Drop me a line and let’s make the next 600 hours smoother for all of us.


 

Just one random related thing that I made up a good analogy to explain:
Patch Cords: The Unsung Heroes of Your Setup. Every photometry company will tell you to spend extra money on their low-autofluorescence patch cords, but is this necessary? To answer that question, we need to have a good understanding of what “autofluorescence” is.

Here’s the analogy I like to use: I have a special tube that allows two liquids to flow in opposite directions without having them mix, as long as there is a constant flowrate through the tube. One end of my tube is connected to a water source, and there’s a branch breaking off from this end where I can sip from a straw connected to the line. The other end of my tube has a can of concentrated orange juice, such that water will flow down through the tube, dilute some of the concentrate, and then the diluted orange juice flows back up the tube so I can drink it with my straw.

Throughout this process, some water and diluted orange juice molecules get stuck to the walls of my tube, and so the water doesn’t flow as smoothly, which causes my orange juice to have an inconsistent ratio of water:concentrate. If I were to plot this ratio vs time, it would be a jagged sawtooth pattern, and it would get more jagged the longer time goes on since more molecules would get stuck in the tube and disturb the flow the longer the tube is open.


 

That’s pretty much what happens with the photons traveling through the patch cord. Every now and then a photon will “forget” what direction it was traveling in and start bumping into the walls of the patch cord, disturbing the flow of the photons that are used to excite the sensor and those emitted by the sensor.

 

Now that we have a good understanding of what “autofluorescence” is, let’s try and understand “photobleaching.” This is where you make it so things can only flow one direction through the tube, and you blast water through the tube with a high pressure to make sure there’s no water molecules stuck to the tube walls. With photometry patch cords we blast high-powered light through the patch cord to clear them out.

 

Now we can try and answer the question: “do we need low-autofluorescence patch cords for fiber photometry recordings, and do we need to photobleach them?”


 
I have not run any experiments directly comparing the two, but my intuition (based on recording from brand new short and long patch cords without photobleaching them beforehand) is that these are unnecessary unless you have an ungodly long patch cord.

 

Let’s go back to our water/orange juice tube. Let’s say the tube was only the length of the straw. The ratio of water:orange juice concentrate should be consistent – it’s a small distance between the water, concentrate, and my mouth.

 

Now let’s swap out the straw for the trans-Saharan gas pipeline. Sadly, my orange juice is going to have a wildly inconsistent taste as I suck it out the tube. Same with my photometry signal if my patch cord were that long – the signal would be too noisy to discern “signal” from “noise.” Photons will get stuck in the patch cord, disrupting the steady flow of light.

 

If you have run a test comparing the two types of patch cords, please let me know. I hope this analogy helps you the next time the sales rep comes knocking on your lab door trying to sell you their "newest and lowest autofluorescence patch cord." Before purchashing, ask them how consistent is the orange juice concentration as it travels through the patch cord.

Chapter 2 – Fiber Photometry Recordings
If you've ever found yourself in a standoff with a fiber photometry setup (and let’s be honest, we’ve all been there), you know it's equal parts science and sorcery. Hi, I’m Jacob, and after 600+ hours of wrangling dLight recordings for my PhD, I’m here to share some hard-earned lessons. As every scientist knows, these are mostly from things going off the rails.

 

The Aha Moment: Why Fiber Photometry Simplified My Life
I like to think that the genius who invented fiber photometry (Ludwig Maximilian’s group was the first to publish the technique in 2005) had an ‘aha’ moment. They probably spent hours at the surgery station prepping animals for imaging GECIs and asked themselves: 'Wait, why am I making this so complicated when I could just beam light through a fiber optic to get what I need? Less mess, more science.'

I don’t know if this was the motivation for inventing fiber photometry, or if the researchers were intentionally looking for a way to monitor GECIs during freely moving behaviors. To be honest, it’s hard for me to tell. I would have expected fiber photometry to be published as its own “Nature Methods/Protocols” paper, but the actual technique is was first shared in a Brief Communication in Nature Neuroscience titled: “Cortical calcium waves in resting newborn mice.” Fiber photometry is mentioned in the first sentence of the abstract, but I did not expect that from the title.

 

In any case, this invaluable technique gave researchers the ability to monitor GECIs from a larger brain volume of tissue using simpler methods than they were previously. This technique has the added advantages of being able to monitor multiple volumes of tissue simultaneously in freely moving animals.
 

 

Oh, The Sensors You'll Record
After three years of chasing signals with fiber photometry, I’ve learned a few things—like how indicators can be your best friend, and sometimes your most annoying lab partner (looking at you, aberrant microwaves following AAV expression of indicators – shown with GCaMP, please let me know if this has been shown with other indicators too).

I’ve mostly recorded dLight for dopamine (some GRAB-DA), but I’ve also recorded:

  • nLight/GRAB-NE – norepinephrine
  • iSeroSnFR – serotonin
  • cLight – CRF
  • GCaMPs – calcium
  • deltaLight – endogenous opoiods that bind to the delta opioid receptor

Each of these sensors have super different response properties, especially in vivo – mostly because each of the ligands these sensors measure operate on super different timescales. Truth is, it doesn’t matter if you are using Fura-2, jRGECO, beta-testing one of the new Tian lab sensors, or even using a FRET sensor, the technique is always the same.

 

 

How to do a fiber photometry experiment.
1. Pick the right wavelength for your sensor.
You can think of it like tuning a guitar—hit the wrong note, and the song gives you a headache instead of a rush of dopamine.

You can get a good approximation of the optimal wavelength from fpbase.org.
Note: every sensor alters the structure of the fluorescent protein to some extent, so the excitation-emission curve of dLight is different from GCaMP, even though they use the same cpGFP.


2. Decide on the best control for your experimental setup and conditions.
I have some strong opinions about this, but I’ll present all the options.

 

a. Isosbestic channel.
The isosbestic channel is like the safety net of your experiment—it’s there to keep things stable, but it only works if you’ve set it up just right. Miss the correct wavelength by a few nanometers, and it can betray you, turning your controls into chaotic messes that ruin your actual signal.

A hair’s difference (not literally since a hair is >50,000 nanometers thick) between isosbestic used and the sensor’s true isosbestic wavelength and end up with ligand-dependent signal (or inverse signal) in your isosbestic channel. This is problematic when you go and do isosbestic correction in your analysis, as you’ll end up removing/inflating some of your ligand-dependent signal. That’s why I have strong feelings about controls.




b. Non-binding version of your sensor.
In the Tian lab we have “0” versions of all our sensors, which is a sensor with a “broken binding pocket.” Ligand doesn’t bind, so there is no ligand-dependent signal from this sensor.

Theoretically, this sensor should be mostly flat, but I’ve never experienced that. Instead, I get super stereotyped responses that are animal/brain region dependent; I don’t see any correlation between signals from the same brain regions in different animals or even between different brain regions in the same animal.

I spent months racking my brain on why that might be. If you’ve ever found yourself cursing at the 0 version of a Tian lab sensor, trust me, you’re not alone. Here’s what I think is going on: these sensors are too f***ing bright.

The 0 variants haven’t been optimized to get the basal brightness levels down. The brightness isn’t changing from any ligand signaling. Therefore, these sensors are super sensitive to processes that impact photon efficiency (like hemodynamics).

If there is a large blood vessel near my implant, then my implant is basically giving me “inverse-BOLD” signal since my recording is going to get darker when blood flows into the brain region because it’s more active and get brighter as is returns to that f***ing insane level of basal brightness.

^ I don’t really have a good way to test this hypothesis without a mouse MRI machine. I tried sticking implants in the brain without putting any sensor, and I observed a similar phenomenon – very stereotyped responses from each recording site. If you have ideas, please send them my way.


 

c. "Inert” fluorescent protein.
From data I've seen, this is probably the best control, but it is not one I have used before. (the CMOS-based photometry system I use has a hard time multiplexing different color). Basically, you inject two viruses: your sensor and a different colored fluorescent protein. (There are ways to optimize expression of the two. For example, you can use green membrane-localized dLight, and cytosol-localized mApple.) The second fluorescent protein should have a relatively low baseline fluorescence, such that the only changes in fluorescence intensity will be motion dependent.

There might be some ligand-dependent changes in signal, but those would mostly be due to changes in the environment surrounding the proteins (pH, degree of quenching, etc) that would be reflected in changes in their lifetime signals – not something that most people are measuring.

I first came across this method when I was looking for photometry analysis techniques, and came across this awesome Github page by Thomas Akam at Oxford. I haven’t seen this method used in too many other places, but I’ve spoken about it with Lin Tian at length, and we agree it’s probably the best control out there. I will write an addendum to this post after I’ve done some of my own recordings with this technique.

  1.  
  2.  

3. Setting up your photometry recording equipment.
Doesn’t matter if you are using CMOS-based detection with interleaved excitation and detection or continuous excitation with silicon-photodetector-based detection using lock-in modulation. You need to direct your excitation light through I dichroic mirror such that it will excite your sample, and the emitted light will travel through the dichroic to the photodetector. Just make sure you are using narrow bandwidth filters to keep everything close to your peak excitation/emission.


The simplest and cheapest commercially available systems will take care of all these concerns you (RWD and Doric). If you’re a person that likes to “get under the hood” a bit or want a system that allows maximum flexibility for recording new fluorescent proteins in the future, there are a couple of good options for you. The RZ10X from Tucker Davis Technologies uses “plug-n-play” LEDs, filters, and dichroics, so you can tailor things to your needs. ThorLabs offers “Kinematic” cubes and sells individual components to give you maximum customization and control over your system.
 

 

4. Coupling the ferrule patch cord to the ferrule of the implant.

The actual mechanics of this step depend on your implant and patch cord type, but this is the most important step of the recording. If the patch cord doesn’t stay perfectly coupled to the implant throughout the duration of the recording, then you will get signal loss/reduced photon efficiency.

I use patch cords/implants with metal or ceramic ferrules, and I use bronze sleeves to couple the implants to the patch cords. I find these sleeves keep the connection tight between the patch cord and the implant. I even have a hard time removing these sleeves at the end of my recordings sometimes.

I usually fill the patch cord with optical coupling gel before attaching it to the implant. At first, I believed it helped keep the light path intact in the event of patch cord slipping/rotating as the animal moves around the cage. I’m starting to think this might fall under the category of “super-secret handshakes you should do to get your experiment to work,” but it doesn’t do me or the animals any harm to use it, so I’m going to continue using it.

5. Doing the recording.

At this point, you have done everything you can to get the best recording you can, and it’s “up to the software.” I’ve only used Doric Neuroscience Studio for photometry recordings since I’ve always used a Doric system, but this is some messy software. It crashes all the time, it has tons of bugs, and it eats up all the memory on my computer.
 

I haven’t tried any alternatives because Doric keeps rolling out updates, and I keep on having faith that they are going to solve the problems. At this point, I’m just trying to get enough recordings to graduate. If the software crashes during my recording, then I’ll just record another day (if my experiment allows for this – some of the experiments I do have a “novelty” component, so I obviously cannot redo these recordings).
 

 

Conclusion
That’s enough rambling from me for one post. I was tempted to dive into analysis too, but trust me, it deserves its own chapter. Until then, I’d love to hear your photometry war stories—what’s worked, what’s failed miserably, and where you’ve cursed your setup like me. Drop me a line and let’s make the next 600 hours smoother for all of us.


 

Just one random related thing that I made up a good analogy to explain:
Patch Cords: The Unsung Heroes of Your Setup. Every photometry company will tell you to spend extra money on their low-autofluorescence patch cords, but is this necessary? To answer that question, we need to have a good understanding of what “autofluorescence” is.

Here’s the analogy I like to use: I have a special tube that allows two liquids to flow in opposite directions without having them mix, as long as there is a constant flowrate through the tube. One end of my tube is connected to a water source, and there’s a branch breaking off from this end where I can sip from a straw connected to the line. The other end of my tube has a can of concentrated orange juice, such that water will flow down through the tube, dilute some of the concentrate, and then the diluted orange juice flows back up the tube so I can drink it with my straw.

Throughout this process, some water and diluted orange juice molecules get stuck to the walls of my tube, and so the water doesn’t flow as smoothly, which causes my orange juice to have an inconsistent ratio of water:concentrate. If I were to plot this ratio vs time, it would be a jagged sawtooth pattern, and it would get more jagged the longer time goes on since more molecules would get stuck in the tube and disturb the flow the longer the tube is open.


 

That’s pretty much what happens with the photons traveling through the patch cord. Every now and then a photon will “forget” what direction it was traveling in and start bumping into the walls of the patch cord, disturbing the flow of the photons that are used to excite the sensor and those emitted by the sensor.

 

Now that we have a good understanding of what “autofluorescence” is, let’s try and understand “photobleaching.” This is where you make it so things can only flow one direction through the tube, and you blast water through the tube with a high pressure to make sure there’s no water molecules stuck to the tube walls. With photometry patch cords we blast high-powered light through the patch cord to clear them out.

 

Now we can try and answer the question: “do we need low-autofluorescence patch cords for fiber photometry recordings, and do we need to photobleach them?”


 
I have not run any experiments directly comparing the two, but my intuition (based on recording from brand new short and long patch cords without photobleaching them beforehand) is that these are unnecessary unless you have an ungodly long patch cord.

 

Let’s go back to our water/orange juice tube. Let’s say the tube was only the length of the straw. The ratio of water:orange juice concentrate should be consistent – it’s a small distance between the water, concentrate, and my mouth.

 

Now let’s swap out the straw for the trans-Saharan gas pipeline. Sadly, my orange juice is going to have a wildly inconsistent taste as I suck it out the tube. Same with my photometry signal if my patch cord were that long – the signal would be too noisy to discern “signal” from “noise.” Photons will get stuck in the patch cord, disrupting the steady flow of light.

 

If you have run a test comparing the two types of patch cords, please let me know. I hope this analogy helps you the next time the sales rep comes knocking on your lab door trying to sell you their "newest and lowest autofluorescence patch cord." Before purchashing, ask them how consistent is the orange juice concentration as it travels through the patch cord.