0:35
Good afternoon everyone.
0:38
So what I'd like to do today is to walk you through how could we try basically to fill up the gap between preclinical model and clinic by trying to get model which I'm as translational as possible.
0:51
As you know, most of the strategy which are currently developed in immunotherapy are involving compounds are highly selective to the human targets.
0:59
They have been raised again such as bispecifics CAR-Ts etcetera.
1:04
And those therapeutics most of the time act on both lymphoid and myeloid cells.
1:08
So how could we have a preclinical model that will enable the assessment of therapeutic which are highly specific to human target while reflecting what would happen ultimately in humans.
1:18
As you know, humanised mice model has been used for years in order to try to provide models with a human immune system.
1:27
This is a way to try to overcome the species specificity issue when testing bispecific antibody.
1:34
However, the main limitation of this model is the fact that the immune system doesn't really fully reflects what happened in human.
1:43
During today's talk, I will stay focused on this part of the preclinical part and I will have an emphasis on efficacy study and safety studies using a humanised mouse model.
1:55
So on the next slide I just highlighted the model which are the more advanced which exhibit both lymphoid and myeloid cells.
2:04
Because as I said humanised models were developed almost 2 decades ago.
2:09
They first had models which have mainly T&B cells and starting from 2010 some new model kicks in.
2:16
This model also has myeloid components.
2:19
The myeloid compartment development is boosted through the over expression of human cytokines.
2:25
They are myeloid cells that develop, but the downside of it is that those animals have a very short lifespan and they do develop inflammatory reactions which create some quite heavy phenotype in those mice.
2:38
The second model which has been reported was a NOG-EXL.
2:42
Same strategy.
2:43
They also over expressed human cytokine, but this time they took off the human CSF.
2:48
The result was a longer lifespan in the animals, a less severe inflammation.
2:54
However, the side effects are still reported and those animals display anaemia.
2:59
Our analysis of this situation was that most likely those side effects are due to the fact that the cytokine which are getting used to drive the human myeloid cell compartment development are expressed in a super physiological manner which create the inflammatory reaction that is seen.
3:14
So we decided to up to another system which is the BRGSF model.
3:19
These mice have a knockout in mouse Flt3L which free up a niche for the human cells to develop upon engraftment.
3:28
Upon engraftment the model would have developed both lymphoid and myeloid cells and the human myeloid cells could be boosted transiently by Flt3L injections.
3:36
This system has also recently been reported by the Jacks and by Taconic where they generated indeed a quite similar approach.
3:46
But again they decided to supply the FlT3L transgenically using non physiological promoters such as a CAG promoter which is a strong ubiquitous promoter.
3:56
Consequently they reported that the myeloid cells are present but they decrease after a few weeks and among the explanation that they provided one was the exhaustion of the myeloid cells due to the human FLT3 that was expressed at the high level.
4:11
So when we started to work with this model, people were quite sceptical that we could have myeloid cells without any side effect and more precisely anaemia.
4:19
So we run this study jointly with Janssen where we basically compare the BRGSF model, which has received Flt3 in green here to the NOG-EXL, which is this model.
4:28
And as you can clearly see here the body weight loss, whereas those animals are still gaining weight as they aged.
4:35
So what is the immune system in those mice?
4:38
So the model has no T cell, no B cell, no NK cells, and there is a knockout on the FLT3 gene.
4:50
As I said it previously, mice are reconstituted with cold blood cells on human origin and reconstitution is done into newborn animals.
4:58
The newborn mice will then develop a human immune system which will be first dominated by human lymphoid cells.
5:09
Upon Flt3 ligand injection to those mice, you boost the human myeloid compartment.
5:14
As you can see here, this is a pan boosting of the human myeloid cells, which acts both on DCS, macrophages, any kind of myeloid cells.
5:24
Basically that boosting is transient in live animals, but I'll get back to this later.
5:30
First, let's try to have a look to at the immune cells that develop in these mice.
5:33
So this is just a slide where we look at the kinetic of the human cells development in order to see how those immune population will evolve with time.
5:46
So same animals are followed in a longitudinal studies over time over 31 weeks and we breed those mice at different time points and we just look at their frequency of different immune cell population.
5:58
And to make a long story short, while the myeloid compartment remains stable over time, you're clearly seeing here that your T cell compartment increases and reaches a plateau around 20 weeks old while your B cell compartment decreases, which is a good thing because in any his model you have much more B cell than T cell and that doesn't reflect neither mice nor human.
6:19
So we are getting into a more, let's say acceptable ratio of T:B cells.
6:24
What type of T cell do we have in those mice?
6:28
The T cell that we have are just the standard immune cells which are known for T cells.
6:38
So basically you characterise if you look in the spin of those animal you would see different type of T cells, CDA, T-Reg etcetera.
6:46
One thing which is noticeable compared to other HIS mice is that we have TCR gamma delta T cells which you don't generally have into humanised mouse model.
6:55
If we look into the T cell population, you have the T regs which are seen here, you have your gamma delta T cells which are seen here.
7:02
And if you have a deeper look into the gamma delta T cell, you find a different subset delta 1, delta 2, delta 3.
7:07
And we've been able to show that they could get activated by targeting compound which are specific to either gamma 9 or delta 1 and they respond similarly to human PBMCs.
7:16
But I'm not going to get into those data today, but I'll be more than happy to discuss this if needed.
7:21
What about the B cell compartment?
7:23
Now B cells have basically different maturation stage of B cells from previous cells to plasma cells as you could see.
7:34
And when those B cells are activated as vivo with an anti CD4D and a cocktail of cytokines, they basically get activated and they switch to an IgG neg phenotype 2 and IgG plus IgM plus phenotype.
7:50
If you look into the blood of these mice and you just measure the IG, you would see that you would detect human IgGs and those human IgGs are on different isotypes.
8:02
So basically the B cell are capable of can switch up to IgE.
8:06
So they seems to be fully functional.
8:08
So what we decided to do then was to see if this model could be used in the context of an inflammatory disease.
8:15
So the first studies that we run here was a study on a DTH study using a vitamin D analogue which is called MC903.
8:24
MC903 drives the Th2 response and that will drive recruitment of a eosinophils and an increase in IgE in the serum.
8:32
So this is a protocol that was used.
8:34
Mice received Flt3L ligand for a week, then they had the topical administration of MC903 on each lab of the ear.
8:45
And we just looked at the inflammation on the ear through the ear sickness of the animal.
8:50
And as you can see here, the more MTC03 you give, the higher is the information in the ears of the animals.
8:57
What we wanted to do is to see if we could have a Th2 dependent information process.
9:01
So we use a different system which is called oxazolone which is an atom and we followed this protocol boosting of the myeloid compartment sanitization with oxazolone on the abdomen of the animals and then re sanitization on the ears at different time point here.
9:20
And over 2 weeks what happened is the following.
9:23
You start to see an increase of the sickness of the skin again and to show that this was T cell dependent, we decided to treat the animal with Orencia which is an antibody that blocks CTLA for interaction with CD80.
9:38
CD86 study was performed and as you could see you clearly see here the increase of the sickness of the animals upon treatment with oxazolone.
9:49
And as soon as you treat the mice with Orencia, you mitigate that information.
9:53
So this is information is clearly T cell treated.
9:58
Immune cells are present, they seem to respond to known stimulus.
10:03
How would they react to the context of tumour bearing mice?
10:06
So do we have the plasticity of interaction between tumour and immune cells and that's how would this drive the recruitment of the immune cells in the tumour microenvironment?
10:15
So in the study that I'm sharing with you here, we took mice deriving from 5 different donors and we just take one time point which was 400 to 500 cubic millimetre and we compare the TME composition depending on the tumour type.
10:27
Several tumour models were assessed, MDA-MB-231, so a triple negative breast cancer cell line.
10:33
As you could see about 2/3 of the cell recruited in the TME are of human origin in red versus mouse origin.
10:42
HPAF2 here, that's the opposite, 2/3 of mouse origin, 1/3 is of human origin.
10:49
A549. Again you have a higher mouse infiltrate than human infiltrate, but human cells are still present.
10:56
What happened?
10:57
Now if we start to look into the immune cells subset which are present in the TME, these models recruit mainly myeloid cells.
11:07
You also see NK cells, T cells and B cells.
11:12
The HPAF 2, it's highly T cell driven, whereas the A549 has approximately a majority of T cells and NK cells, but also have other type of cells such as myeloid cells.
11:27
So you're clearly seeing from this picture that your immune cell recruitment into the TME will be tumour dependent and that will also drive the nature of the immune cells which are getting recruited.
11:41
What about the activation status now?
11:43
So we decided to go a little bit further and with the MDA-MB-231 model, we better analyse those cells looking at their activation level.
11:52
So we focus on the two sort of cells that are a human origin.
11:55
So the human CD45 positive cell here in Reds, the majority of those are myeloid cells as you could see here in brownish.
12:03
Those myeloid cells are mainly CD14 positive cells.
12:06
They are M2 like macrophages and they also show activation markers such as CD80, CD86 and they also express human specific receptors.
12:18
So we went further.
12:20
We look for PD-L1 expression among those M2 macrophages, most of them are PD-L1 positive, about 80% of them are PD-L1 positive and this remains stable over time.
12:33
This regardless of the size of the tumour.
12:36
If you look now into the MFI, you clearly see that PDL one expression increases on the cells as the tumour grows.
12:44
On the other hand, we look into T cell and looking at the T cell, most of them are PD1 positive for the CD8 cells and PD1 increases with the tumour size on the CD4 and the T regs.
12:59
So we decided to look a little bit better on that kinetic of activation looking this time in the same tumour model at different time points.
13:07
So we did 200 cubic millimetre, 500 and 800 cubic millimetres.
13:13
What this slide tells you, it's really the kinetic that we're seeing.
13:16
The majority of the cells that come in first are human cells with this system model.
13:20
Then the mouse cells come in and the mouse cells reach a plateau around 37% approximately.
13:26
If you now look into those mouse cells, they will mainly be here myeloid cells as said previously, M2 macrophages.
13:34
Then you have the NK cells that comes into the tumour as well as the T cells and the Tregs and then you get a stabilisation of your tumour microenvironment which is still dominated by M2 like macrophages.
13:47
So the kinetic of recruitment is as shown here and two macrophages are present along the tumour development.
13:55
Then your T cells kick in and you see your Treg getting recruited as well.
14:02
How would this be in a model which is slightly different A549 which were highly infiltrated with NK cells.
14:09
Same study was conducted and I said previously we have a majority of mouse cells here in blue and you clearly see that when the tumour reaches 8000 cubic millimetres, it's mainly mouse dominated.
14:25
However, if we focus on the human cells which are recruited in the TME, we see that the majority of those human cells are NK cells here and T cells, the NK cells would be present all of the tumour growths.
14:42
And if we look into the NK cell subset, we clearly characterise standard subset which are known NK1 NK2, and NK3. NK1 and NK2 being present at early time of tumour development and then they decreases and then the NK3 kicks in.
14:57
So the cytotoxic NK are present at this time and we don't have to add any IL15.
15:01
They are naturally produced and they get present in the model.
15:05
So are those NK cells functional?
15:07
We did the study with an NK cell engager which targets CD20.
15:11
So the idea here was to try to deplete B cell in those mice.
15:15
So mice were injected with a compound by in IV.
15:19
We tested 2 dose, 10 milligramme per kgs and 30 milligramme per kgs.
15:22
And we look into the blood of those mice at 48 hours.
15:25
And what you're seeing here is that the depleting agent in blue show a nice reduction of the B cell circulation in the blood of the animals, whereas the FC silent version of that antibody which is not capable of inducing ADCC shows no effect compared to control.
15:46
So we clearly see here that we have an ADCC which is FC gamma receptor dependent.
15:55
So we have immune cell in the TME, they get present, they get activated.
15:59
Could they get moderated in order to drive an anti-tumour response?
16:05
So several studies were performed.
16:07
I'm just going to share with you three of them here.
16:10
One study I did jointly with Light Chain Bioscience in Geneva.
16:16
This guy had a bispecific antibody targeting CEA and CD3.
16:26
It's known that pancreatic tumours are resistant to CEA CD3 co-targeting.
16:33
So they decided to test the combination of that bispecific to an antibody targeting CD28 and PD-L1.
16:40
So in other word providing signal 1 and signal 2 to T cells, would that provide a better tumour growth control than single one only?
16:50
This is a protocol that we use.
16:52
Mice were engrafted and then treated by IV twice and the last treatment was IP.
16:57
And as you could see here, the CD3 CEA in grey has no effect compared to control.
17:03
And you clearly see here a decrease of the tumour growth when the combination therapy is provided.
17:09
And you have the other individual mice that are being assessed here.
17:14
So T cells are able to respond and are able to control tumour outgrowth.
17:19
Could you repolarize your T cells in this work?
17:23
The did together with Teva in Israel they developed an anti PD1 antibody which is not a PD-1 blocker.
17:31
This antibody binds to PD-1 high cells and is linked to an attenuated IL-2.
17:38
So it will reactivate T cells which are PD1 high.
17:44
And the outcome in this Melanoma tumour model is a control of the tumour out growth compared to nivolumab alone.
17:50
Here the study was prolonged but I'm we're just showing this time point here and the tumour control was much greater later.
17:59
But what is quite impressive is how you totally change the ratio of your CD8 TReg in the TME upon treatment.
18:07
With this attenuated IL2 anti PD-1 therapeutics.
18:12
Finally, could we repolarize the myeloid compartment?
18:17
This is also a study done with a biotech company located here in London.
18:21
These guys develop antibody against LILRBs.
18:23
LILRBs are inhibitory receptors that are involved in the immune escape mechanisms.
18:29
They develop a panel LILRB which would bind to LILRBs and turn that immune suppressive signal into a immunomodulatory signal.
18:37
And as you can see here that result into tumour growth control.
18:42
So one of the key challenges when dealing with therapeutic that's controlled the immune system is how could we prevent immune related others effects.
18:53
The standard way that people are using is to use PBMC reconstituted mice and treat those mice with that compound and see whether that's induces cytokines.
19:03
We were a little bit, let's say sceptical because we all know that PBMC reconstituted mice only have, well mainly have T cells and do not have myeloid cells.
19:13
So what we did in this study was to compare the immune system of CD34 positive reconstituted mice to the immune system of PBMC reconstituted mice.
19:22
If you look carefully here, you're clearly seeing that the PBMC reconstitution is highly dominated by T cell in blue and had also B cells, whereas the CD 34 reconstitution shows cells of different origin which are spread all over the spectrum of the immune cells that you find in those mice.
19:42
Now if you focus on the T cells, you'll see that in the CD34 reconstituted mice, most of your T cells are naive cells here 30%, whereas only 3% in the PBMC reconstituted mice are naive.
19:53
And in contrary, most of the T cell in this model are affected T cells.
19:58
They are already pre activated.
20:00
This pre activation is seen here in this graph where you clearly see that those cells are high into LAG3 and CD25, showing that they get reactivated despite the fact that you have no sign of GvHD.
20:19
So what we decided to do was to assess a cytokine release syndrome in animals that have myeloid cells because the myeloid cell compartment will be key in the cytokine storm that should be triggered upon treatment with therapeutics and that was detailed in this paper.
20:35
In summary, we treated mice with OKT3 and we observed cytokines as you could see here.
20:43
We repeated the same study in mice that received Filt3 ligand at the grid curve and as you can see here we have an increase of cytokine.
20:53
We first have cytokine of T cell origin, then we start to have cytokine of myeloid cells origin and finally you see the myeloid cells activation that peaks here.
21:05
On top of this, we also detected some clinical sign that was associated with cytokine syndrome in green here, body weight loss and body attempt to drop these clinical signs were observed into this model that was repeated with other therapeutics.
21:21
And we wanted to see if they wrote really due to the human immune system and not the mouse immune system.
21:26
So what we did was to treat the mice with infliximab, which is an anti-human TNF alpha infliximab treatment in orange totally mitigates the response to the cytokine but also mitigates the clinical sign clearly showing that what we're seeing is human specific.
21:45
Finally, we decided to see whether we can understand the mechanism that support the toxicity of some compounds.
21:52
So we took an anti-VISTA that were developed by Janssen, which had an early stop in the clinic in 2014.
21:59
This antibody was tested into the system and we use OKT3 as a control.
22:04
The cytokine release was even greater than with OKT3 as you can see here.
22:09
Then we look into the immune cell population and we clearly see that this antibody depletes VISTA positive cells, monocyte and disease.
22:17
Finally, we decided to use this system to test a novel anti VISTA which works in the pH selective manner.
22:25
That antibody here doesn't trigger any cytokine release and doesn't deplete the VISTA cells.
22:31
This antibody went into clinic afterwards and in the clinic they did multiple dosing of the antibody.
22:39
And as you could see here, those multiple dosing did not result into any cytokine release at all.
22:44
So in other word, what we observe in the model was indeed translational compared to what they've seen in the clinic.
22:56
So just to conclude on this year, I'm going to finish with this last slide just to summarise what was done on cytokine release.
23:03
Two compounds which are known to be toxic have shown toxicity.
23:08
One compound which is less toxic show a much weaker cytokine release and no clinical sign and the SN101 which was totally safe in the clinic did show also the same safety here.
23:20
So this type of model also has limitations.
23:23
This is why we know their limitation and we also use a lot of syngenic models because those syngenic model would probably provide the better cross between immune cells and stroma which you won't have in those systems due to the cross reactivity issues between human and mouse cells.
23:40
I'm just going to finish by thanking all the different company that supported us over these years and thank you for your attention.
