Robot Updates

As part of ongoing attempts to get the system to work to our liking we run into two problem.

First, many of the steps in our yeast growth protocols use individual pipetting actions for each well. For this reason we have an arm with individual pipettes. Initially we configured the arm with four fixed tips and four disposable pipettes. The fixed tips are easier to use and do not consume plasticware. On the other hand to get perfect sterility we need to use plastic tips.

So far we learned that the fixed tips (with a wash between uses) is sufficiently clean for our purposes. And so we use these a lot. The problem is that the time it takes to "process" a plate (e.g., remove individually tailored amounts per well for growth dependent dilutions) was too long. The main time consuming steps is the tip washing. 

After consultation with Neotec people, we decided to switch two of the disposable pipette heads with fixed tips. This will give us more tips and thus less washes (as the washes are done in parallel and take the same amount of time for 4 or for 6 tips).

Six fixed tips in action

The second problem is much more serious - the robotic manipulation arm kept getting out of alignment. This lead to serious of failure in our longer experiments. The blame fell onto the plate shaker, which we use a lot in our protocol. The design of the shaker has to allow a robotic arm to place/remove a plate, but at the same time hold the plate firmly during the shaking. The solution in the shaker we had was a spring that held the plate in place. Every time the arm came down to the shaker it would press on a lever and release the plate from the spring. Apparently these repeated presses moved the arm out of place (not by much but sufficiently to cause problems)

The solution was to replace the shaker head to one with a different locking mechanism based on a higher stand that kept the plate sitting down. This required adaptation of a solution Tecan built for arms that hold the plate from above (and not from the side like ours).

New shaker head

So far it seems that the system is indeed more stable. It worked for two days without issues and we are counting.

Funny/Sad Moments

Part of learning to program a complex robot involves learning to debug programs that actually manipulate the real world. As a consequence "bugs" often have more impact than a simple error message on the screen.

Most these bugs are our fault, and so really have no one to blame. Once in a while there is one that seems to come out of the blue, and it took a while to find out what happened. Today, Ayelet tried a simple protocol that worked fine before, and it had dramatic results. I came in to check if this error was reproducible and so got it onto a movie (although a bit out of focus).

After some soul searching, finger pointing, and checks. We realized that the incubator was slighly moved when the FACS was being serviced. As a result the programmed position for lifting the plate was off, and thus the dramatic consequences.

Once we figured this out it was easy to recalibrate the position (we are getting good at that). Moreover, we had the incubator fixed into place the next morning. (We have had the request to do so out for a while, but now we made it clear this was urgent.)

Robotic NanoString Experiment

Today was finished our first large scale experiment using our robotic platform. 

Assaf programmed the robot to grow 90+ strains of yeast over two days to reach optimal conditions (as I previously described), then the robot stimulated them and harvested cells at 9 time points. Each harvesting step involved taking small amount of cells from the plate they were growing them and plunging them into a plate of cold liquid methanol. This immediately freezes the cells and "stops" them cold (literally).

Next, Assaf used the protocol he and Ayelet devised for extracting RNA from yeast cells, to process these 9 x 96well plates. These will be shipped to Ollie's lab for running the NanoString assay to measure exact RNA quantities.

This was the first time we had an actual experiment running from start to finish. Although we "debugged" various parts before, putting all of them together uncovered all types of problems. Assaf had to restart the growth several time before we had the whole thing running. But now we are smarter and know how to get things right the next time.

After few days of heavy robot use, Assaf collected all the plasticware he used. Here is a picture of triumphant Assaf at the end of the process.

Growing yeasts (Robotically)

The first task we set to do with the robot is grow yeast cells for experiments. This sounds easy? No?

To understand the issue. Lets review the typical yeast life cycle. Suppose you pick a small number of yeasts from a colony or a saturated culture and put in a fresh "rich" media with glucose (sugar). Moreover, suppose that the media is kept in nice temperature (yeast like 30C) and shaken to make sure yeast cells and nutrients keep mixing.

Initially it will take the yeast cells time to realize that they are not in the nutrient-poor environment they were in. They will revitalize themselves and prepare to grow. This phase is called lag phase. During this phase the number of yeast cells will not change.

Once out of lag phase the yeast will start to grow on glucose. They will work hard to use this rich source of energy to grow as fast as possible. A typical yeast cell will divide every 90-120 min (depending on the exact conditions and temperature). This period is called doubling time as the number of yeast will double every fixed period. During this phase the yeast will experience exponential growth. For this reason this is known either as exponential phase or log phase (since it looks linear in logarithmic scale). 

After a while the yeast will start exhausting the sugar. If this was bacteria they will stop growing. But yeast has another trick up their sleeve. They switch from using sugar as fuel to using ethanol. During the fast growth phase the yeast use the glucose in a fast way by fermenting it to ethanol. This is what in human metabolism is known as anaerobic metabolism as it does not require oxygen. In nature this trick allows yeast to outgrow the competition (fast growth) and also kill it (by increasing ethanol concentration). Humans learned long time ago to use this property of yeasts to make alcoholic beverages. 

Returning to the yeast growth, the switch from glucose to ethanol is called the diauxic shift --- the yeast will go through it once it can no longer import glucose from the environment. Ethanol can be used for aerobic metabolism (or respiration), but requires more work to extract energy from it. As a result the yeast will grow slower. They still grow exponentially but the doubling time is much longer.  This phase is often referred to as saturated or early stationary phase although these description are inaccurate as the yeast still grows.

After a while (and this can take much longer), the ethanol reserves are consumed, and the yeast stops growing and enters in to stationary phase. The cells prepare for nutritional hardship and reduce their activity. 

When plot the number of yeasts in the tube during this phases we ideally see this type of curve:

For our experiment we want to take yeasts in the middle of the fast growing exponential phase. Moreover, to make sure that the yeast forgot its history, we want to make sure that there were several (>3) cell divisions since the lag phase. This means that we need to yeast to multiply itself by at least 8-fold from the initial amount.

Moreover, we want to make sure the yeast do not come close to diauxic shift, as this stage results in major changes in the yeast metabolism. This means avoiding over-crowded situation. Finally, we also want to ensure that we have sufficient number of cells to work with, so we do not want under-crowded cells either.

Sounds easy. If the lag phase is 60min and doubling time is say 90min, then we need 330min (5 1/2 hours) to grow the yeast. Calculate the desired amount at the end and seed the culture with 1/8th of that.

The problem is that we want to work with many strains of yeast. In fact, we want to grow 96 strains in one plate. Each strain has different lag time and doubling time. This means that while one strain has 90min doubling time, another might have 150min doubling time. This means that for the latter strain we need 450min to get 3 doubling (8-fold increase), but by then the fast strain has two more doublings and has grown by 32-fold from the original number of cells. Due to the exponential growth, small difference in growth rate can lead to dramatic differences in cell concentration.

So, how do we deal with the problem? Ideally, we can measure the relevant times for each strain and then plan the initial seeding to get things right. In fact, this is what we do, but using a robot.

During the last two weeks, Assaf and Avital developed a robotic protocol that grows the plate of yeast for 20 hours. Every half hour the robot took the plate out of the incubator, and put into a plate reader (spectrophotometer for plates) that measures the optical density (roughly equivalent to number of cells). After this incubation time most wells were past the diauxic shift. The program then used the plate to seed a new plate and again monitored growth for several hours. At this point Assaf and Avital's program computed what dillution it need to perform to each well to ensure that at the planned target time the cells will rich a desired density. The robot then applied a customized dilution step for each well.

At the end of this procedure we had a plate with 96 strains (with very different growth characteristics) all in roughly the same density. To our surprise/relief/joy the robot did all of this without a fault.

The end result can be seen like this. In this graph OD corresponds to yeast density, and each curve describe the density in one well on the plate. You can see the yeasts growing fast and then slowing down. You can also see the two dilution steps (the first dramatic one and then the "correction" step). Most importantly you can appreciate that in the end all the wells are fairly close to each other in density.

The nice growth can be more easily seen in a log-scale plot:

For some reason the empty wells (that do not grow :-) misbehave after the dilution step. Do not that before the final dilution there is a large variability in the density and that it mostly removed by the program.

And so now we can start doing experiment with tightly controlled yeast growth. Yey!

Robotic Arm Integration (and other news)

The last week I was away on travel. During the week Shai (from Neotec) together with Avital & Assaf worked on robotic programing and hopefully we will have updates from that front.

In addtion, Udi and Yoram from Neotec manufactured a nice station for handoff between the Tecan Liquid Handling Robot and the KiNEDx robotic arm. Today Shai together with Amir and Moshe (also from neotec) worked on fine tuning the integration of the robotic arm into the Tecan control software.

At the end of the day we managed to film a demonstration in which using EvoWare (the Tecan control software) we can take a plate from the deck of the Tecan to the microscope/hypercyt and back. This means that we made a significant step toward integrating the system.

It is impressive to see the KiNEDx arm in action. Its movements are fast and fluid.

From the movie you will also notice that we hang the plastic curtains that surround the Tecan and keep it clean and sterile. We still need to tailor a hole for the KiNEDx to move through, so for now the enclosure is not complete.

Just to get a sense of the flurry of activity that led to these results, here is a time-lapse of part of that day.

HyperCyt and Python Scripts

Today we had a long day in the robotic room. Yesterday Arik from Neotec came to calibrate the find details of the liquid handling arms on the Tecan which solves some issues we encountered on Tuesday.

Today Shai, who is responsible for "smart" robotic application came. We had a general discussion about different ways of integrating software to the robot controlling software. Afterward he, Avital and Assaf sat down and implement some pythons scripts that call the robot and others that the robot can call.

It seemed that this was very fast pace study as they reported success, and managed to fancy pipetting procedures on the fly.

In the meantime, Ariel and Jenia tried to get the HyperCyt to work in the new location. We decided to move the peristaltic pump onto the hypercyt deck. This however resulted in the arm pushing the pump off when the device was initialized. This was annoying, especially since the pictures of the device showed the pump sitting exactly where we put it. 

We searched the manual high and low, and finally realized that one of the figures mentions an L-shaped piece that serves as a stopper in such a configuration.  We used this as evidence that this is the right solution and installed the stopper, which indeed solved the problem.

However, now the coordinates of the arm were totally off. And so we learned how to "teach" it where the different locations are. This involved moving the tip of the needle very slowly with mouse controls until it was in the right location.

In the end the system was working, and we even managed to film it. As you can see the sampling needle goes into each well in succession. This means that it creates alternating bubbles of media (+ cells) and air in the tube. If you look carefully in the movie you can see these tubes.

After lunch break and group meeting Jenia run his nifty analysis software to break the long FACs stream of events to specific wells. It worked like a charm and immediately gave him detailed summary of each well. The investment in this software was definitely worth while.

First steps with the robot

Today we had our first training session with the Tecan robot. We learned how to create a script for doing various steps (e.g., pipetting, moving plate from one device to another and such). The concepts were relatively simple and intuitive, and by lunch we had our first working protocol. It takes a plate, seed it from another plate. Move the plate to the shaker for a minute and then measures optical density.

After lunch we moved to more complex protocols that used multiple plates (e.g., when you want to run something on a large number of plates). At the end of the day we felt that we can start making use of this device.

We also had a our down moments, including a crash when the robot arm that moves plates decided to go to the first available position, which was not available....

Busy busy day (Robots, FACS, cabinets and sinks)

So another one of these crazy days. So what did we have today.

* David and Ariel from Merkel Technologies came to move the FACS and HyperCyt into the new robotic room (see movie).

* Udi and Shy from NeoTec continued installing the KiNEDx robotic arm. They managed to move a plate from the Tecan to the microscope and back (see movie again).

* We had our first wet-lab group meeting in the room next to robotic room. The furnishing need some work, but the room is usable.

 

* Yoel, the carpenter from "Wooden Horse" (סוס עץ) brought the new storage cabinets/lockers for the corridor. He also brough cabinets for the robotic room, which made it interesting to work there for a while. The workmanship of the new cabinet was impressive.

* The plumbers came to install a water heater in small yeast preparation room. They also finished the sink in the robotic room.

* We had a regular group meeting in the CS building as well :-)

* And most importantly, Yael came to visit the lab to approve it.

And movie of some of the installation around the robot.... Note the sequence at the end where the arm moves the plate from Tecan to Microscope.

Robot is (finally) installed!

After two delays, today was finally the day. Early in the morning Udi and Arik from Neotec showed up for installation. After re-checking that the tables are flat, we set out to install the robot.

The first step was to unbox the robot from the huge shiping box.

All empty space inside the device was full of a box with various accessories. It took a while to realize that we don't have any hope of pulling it out as one box, and we resorted to removing the contents one-by-one.

Now we could start seeing the shape of the device, Tecan Evo 200, and admire it.

Next, Udi and Arik connect special handles to the robot frame.

We recruited Alon from the lab and Ayelet and set out to lift the Tecan (220kg according to Udi) onto the smallish cart that the Neotek people brought with them. 

Somewhat surprisingly, the cart held the weight, and we slowly moved the procession into the room and next to the table.

A bit of a coordinated heave, and the Tecan was on the table. Few more adjustments and it was located into place.

The curious people could now examine the details of the labels.

We then had a small unwrapping ceremony where we got a chance to remove all plastic protectors from the robot outer shell. 

 

Shy, who is the main software integrator showed up and joined the party. Arik set out to installing the device. This involved removing safety brackets and unlocking the  arms. It didn't take long for him to get to a stage where he set the device on a "Random Walk" mode. In this mode the robot tries moving the arms to different X/Y/Z locations to see that there are no obsticles or mechanical problems. It was fun to watch and gave a good impression of what the machine can do.

While the main Tecan was playing at random walk, Udi and Arik started assembling the movable arm. Its called Peak Robotic KiNEDx. It has a flexible gripper that can hold plates and move them from the Tecan to the microscope.

Once Udi managed to get the arm to move, we tested how far it can reach. We had to move the microscope a bit, but now the stage is in reach of the arm.

During this whole procession, Shy started working on interfacing with the external device. He had a quick success in interfacing with the microscope and in no time managed to show that he can control it from the driver he wrote. So, we are hopeful that the integration would be smooth.

And finally, the obligatory stop-motion film of the day's highlights. Given requests from the audience, I edited the sequence to be short and include the main interesting points. Enjoy!

Microscope, Robotics, Tables, and a Window

So the big day has arrived. Today the robot is finally moving into our hands. Things did not turn out exactly as planned.

 

The day started early, with Dimitry from the Eisenberg Brothers who came in to move the microscope from the molecular lab to its new station in the robotic room.

While Dimitry was moving things, a nice guy from the communication department showed up to talk about connecting the ethernet ports in the new room. I then got a long phone call from a collaborator.

A bit later, Udi Oz from Neotec and his crew showed up. They brought the Peak Robotics arm. As they started working, Danny from the computer system showed up to talk about communications and potential place to put wireless. 

Then Michel the carpenter suddenly came in to ask which doors need safety stops. As he was talking to me the Neotec people realized there is a problem with the new tables for the robot. They were not flat, and the rail for the robot arm was not sitting properly because of that.

We started consultation as to what to do about that when the delivery truck with the robot showed. The huge robot box almost blocked the corridor. 

But since the tables are not flat, we had to move to a side area so traffic will not be jammed.

As we were doing this the aluminum profile person showed up and took the temporary window in the robotic room. Michel came in and tried to see if he can help us with the tables. In the end we decided to wait until the PVPlast team shows up and fix the issues. 

Just to get a sense of the problematic aspect here, we took pictures of a straight piece of wood as one end was touching the table, the other is seen to be far from it.

These events were squeezed into five intense hours. Just to get a sense, here is a time-lapse movie of this part of the day.

This left me a bit of time to talk some issues with Udi, grab a sandwich and go to group meeting. On returning we found out that the new window is installed.

It has integrated shades inside, between two glass panels, and so allows to close or open the window easily.

Robotic Platform - A choice is made

Our aim in the long run is to establish a platform to "phenotype" (that is measure behavior of) many promoters in different genetic perturbations. The plan is to use the microscope and the FACS to measure the activity of these promoters in living yeast cells using what is called a fluorescent reporter. (I will expand on this point in a future post.)

To establish such a platform we need to be able to continuously feed the microscope with new samples so that we maximize the throughput in terms of the number of measured phenotypes. We are aiming at numbers in the 10,000s, which means that even when we work with 384well plates, we still need many runs.

An important ingredient in achieving this goal is a robotic platform that will allow us to prepare samples for the microscope and then move them onto it. The goal is that the robot will integrate several devices:

• Incubator
• Plate reader for spectrophotometer reading of cell density
• Microscope
• FACS loader

In addition the robot should be able to do basic manipulations such as moving liquid from one plate to another, add media, to a plate, and so on. Doing this requires that the robot also have what is called liquid handling capabilities. In other words, that the robot should be able to perform pipette operations: from one well to another, or from multiple wells in parallel. It also means that we the robot needs places to store pipette tips, plates, and media.

During the last year we spent a lot of effort in trying to define our needs and to understand the options available for devices that integrate all of these capabilities. We had long interactions with three companies, Eisenberg Brothers who represents Perkin Elmer, Neotec who represents Tecan, and Agilent, where we interacted directly with their European office.

The process was interesting as we learned a lot about the capabilities of different devices, different ways of using them, and ways of integrating them. This included site visits to installations in Israel, Germany, and the US. The main issue with the choice of robotic platform how to integrate the multitude of devices in a way that will allow seamless operations.

I will not go through the gory details. Agilent has dropped out of the discussion sometime in September as they felt they cannot meet the target price we aimed at. Neotec/Tecan and Eisenberg Brothers/Perkin Elmer stayed in the running and we had long correspondence with people from both. We got to financial negotiation stages with both and finally, after much delays we decided to go with Neotec/Tecan solution.

In the next few weeks I will provide outline of the robot and how we plan to use it.