|This diary is a follow-up to the previous one, wherein I abreacted to the way that the adjuvanted swine flu vaccines (with MF59 from Novartis, and AS03 from GSK) were approved by the European Medicines Agency or EMEA. Here I will expand on issues around cytokine response to these adjuvanted vaccines, and, in particular, paying special attention to one cytokine, interleukin-6 or IL-6 for short. This is important because it is now known that IL-6 is involved in the pathophysiology of a large number of diseases, and there is evidence to suggest that vaccination with at least the MF59-adjuvanted influenza vaccine but possibly also the AS03-adjuvanted one, may result in increased IL-6 response, with adverse consequences in some people and/or in certain situations.
Now, if you think that last sentence appears to be too complicated and full of qualifiers, you're right, but it isn't so much an indication of uncertainty as indication of complexity. Every single word of that sentence is there for a reason. I'm giving everyone fair warning right up front, that this is an important exploration of issues with very serious potential consequences, but the subject matter requires a fair amount of patience and learning, in order for you to fully appreciate the significance of the issues raised. Hence also the length of this diary - I simply cannot find ways of splitting it up and still give a coherent account that you can read in one sitting!
This story is complicated to tell also because there are (at least) 2 stories happening side by side. One is the development of vaccine adjuvants, with companies trying their best to get their products licensed and marketed. The other is the astonishing pace of new discoveries in immunology. While those working on the first, the vaccine part, have been keenly aware of developments on the immunology front (and may have been adjusting their positions accordingly), the other - the immunologists - have been largely oblivious. But such 'ignorance' has not stopped them from making major discoveries that are immensely consequential for vaccine and especially adjuvant development.
First, some basics. A cytokine is a molecule made by one cell that has effects on other cells. They are produced when certain cells (eg macrophages) are activated by the appropriate stimuli (eg ingestion of a foreign antigen). They have effects in the local area - the immediate environment where they are produced, but can and often do spill over into the blood if large quantities are made, in which case they can have effects on distant organs. Cytokines have very short life-spans, ie they are made, they do their job, and they disappear, but they can trigger cascading and self-perpetuating reactions, or switch the direction of critical processes, such that the effects remain even when they are gone. Depending on the type of processes affected and host conditions, even transient changes in cytokine levels can sometimes produce long-lasting or irreversible effects.
We'll come back to the science later. Let's look at what information we can find, about cytokine responses from the 2 adjuvanted flu vaccines.
Let's start with the GSK one. In their document submitted to the EMEA, in response to a query about the use of their vaccine in early pregnancy, GSK argued that
the adjuvant is not expected to have adverse effects on early pregnancy primarily because the innate immune response is local to the site of injection and on injection of the adjuvanted vaccine, there is no systemic cytokine response detectable.
Since no data is presented in those documents, nor AFAIK in any peer-reviewed publications regarding the systemic cytokine response to AS03 adjuvanted vaccines, animals and humans included, and since we have already encountered some rather questionable claims from industry representatives, those remarks in the EMEA file should IMHO be considered unsupported by evidence, until they are.
BTW this may well be consistent with corporate policy. GSK also markets vaccines adjuvanted with AS04, an adjuvant that contains both alum and MPL, a 'detoxified' version of the bacterial endotoxin LPS. AS04 is used in a hepatitis B vaccine Fendrix which was approved in the EU in 2005, and the HPV vaccine Cervarix approved in 2007. In a recent presentation to the FDA seeking licensure of Cervarix in the US, company representatives first said that the vaccine was "not associated with a systemic cytokine release." (p.24 of meeting transcripts) Then, during Q&A, they modified that to "just a little bit of IL-6" (p.168) in response to questions from an advisory committee member, an Obs/Gyn doc from Mass General Hospital who said
One of the things that troubled me a little bit about the spontaneous abortion issue as an obstetrician, and that is why I asked the question about detoxification of LPS, because there are very good animal models for inducing abortion and early delivery using LPS, presumably mediated by IL-6, TNF alpha and IL-1 beta.
Again, no data was presented. But they did say that data demonstrating no/low serum cytokines has been submitted to journals for publication, in 2009. That is 4 years after Fendrix was approved, 2 years for Cervarix, after millions of people have been vaccinated. So, I don't expect them to act differently for AS03.
So much for AS03. Now let's look at MF59.
In 1994, scientists from Chiron (before it became Novartis) published a study with findings that, I suspect, may have troubled many people ever since, albeit for different reasons. We'll go through the relevant findings in some detail, not least because we'll need those details to make sense of more recent data from the company that appear to suggest, almost but not quite, contradictory findings.
Here's the abstract of the paper:
Valensi 1994 Systemic cytokine profiles in BALB/c mice immunized with trivalent influenza vaccine containing MF59 oil emulsion and other advanced adjuvants.
ABSTRACT: We have studied serum cytokine profiles in BALB/c mice after immunization with influenza vaccine alone or combined with the following adjuvants: alum; MF59 emulsion; MF59 containing the muramyl peptide N-acetyl-muramyl-L-alanyl-D-isoglutaminyl-L-alanine-2-(1,2- dipalmitoyl-sn-glycero-3-(hydroxyphosphoryloxy)) ethylamide (MTP-PE); MF59 plus the lipid A analogue monophosphoryl lipid A; MF59 plus the Quil A saponin fraction LTC; or LTC alone. Pooled mouse sera were analyzed by ELISA at various times after immunization for IL-1 alpha, IL-2, IL-3, IL-4, IL-5, IL-6, IL-10, IFN-gamma, and TNF-alpha. In naive mice, vaccine alone induced low levels of IL-3 and IL-5 only; vaccine plus alum induced a low IL-6 response as well. The MF59-based adjuvants significantly increased the IL-5 and IL-6 levels, whereas Quil A LTC induced strong IFN-gamma and measurable IL-2 responses, in addition to moderate IL-5 and IL-6. In previously infected mice, MF59 and MF59/MTP-PE were capable of generating IFN-gamma responses, as well as IL-5 and IL-6. All of the cytokine responses were rapid (peaking 3 to 12 h postimmunization) and short lived. In naive mice, the MF59 adjuvants induced serum cytokine profiles that are consistent with a primarily Th2-type response, whereas the Quil A LTC induced cytokines associated with both Th1 and Th2 responses. Ab analyses indicated that, although the adjuvants strongly affected the magnitude of the humoral response, there was no obvious correlation between the cytokine profile observed and the subclasses of Ab induced.
There were several parts to these experiments. First, they vaccinated mice with the flu vaccine either on its own, or with MF59 or other adjuvants. Then they measured the serum cytokine response. This was done with 3 different groups of mice:
- naive mice after 1 injection
- naive mice after 3 weekly injections
- mice previously infected with influenza and recovered, then given 1 injection
The last group is interesting because it mimics host immune status in seasonal vaccination. It is also likely to resemble a substantial portion of vaccinees in the current pandemic, since many people receiving these vaccines may have already been infected without being aware of it, in the 6 months since the beginning of this pandemic. In fact, in the EMEA file, serum samples from the GSK test subjects for H1N1 vaccine, taken this summer, showed 44.6% of those aged 18-40 were seropositive before vaccination, falling to 29% for those aged 51-60. Those numbers can only be higher now.
So they measured a number of cytokines. For simplicity I'm only going to show the results on the 2 most important ones, IL-5 and IL-6, formatted as a composite table. MF59-0 indicates vaccine adjuvanted with MF59 only, and is the one we are interested in.
- First of all, notice the unadjuvanted seasonal flu vaccine (under Adjuvant > None) induced almost no or very low serum cytokine responses, in all groups.
- For Group 1, naive mice with one vaccination, you can see some small increase in both IL-5 and IL-6, peaking at 12 hours and 6 hours respectively. Notice now quickly the cytokines appear and then disappear, all within <24 hours. This is what the companies mean by 'transient', and they are right. Serum cytokine changes from vaccinations are generally transient.
- The next table (Group 2) shows serum cytokines after 3 weekly vaccinations. You can see that the levels are higher. Also, IL-6 is now just a little higher at 3 hours instead of at 6 hours.
- For Group 3 - previously infected and recovered, receiving one dose - As you can see, IL-6 is now much higher than the previous 2 groups, with a very high peak at 3 hours, then disappearing within 24 hrs. Remember this group only received 1 dose of vaccine, but the difference is quite dramatic. But for IL-5, it's only higher than group 1, but not higher than group 2.
- Peak serum IL-6 levels for the 3 groups were 94, 500, and 1655 pg/ml respectively.
To summarize, both prior vaccination (Group 2) and prior infection (Group 3) increased the systemic cytokine response, but the quality of that change is different in the 2 groups. Prior infection seems to have a very strong effect on IL-6 but not as strong on IL-5. We don't know why that is the case here, except we do know that cytokines interact in complex and multiple ways, with cells of the immune system and affecting each other, as shown in this cool interactive diagram from the NIH, for instance.
Two other findings from this paper are of interest here:
- They repeated the experiments several times, in order to test whether the results were consistent and reproducible:
The magnitudes of most of the cytokine responses were quite reproducible....IL-5 and IL-6 responses were detected at relatively high levels in all experiments, with IL-5 showing a range of 1208 to 2466 pg/ml and IL-6 showing a range of 520 to 1646.
- They also injected some mice with adjuvant only, and compared the serum cytokine responses to those receiving vaccine+adjuvant (ie adjuvanted vaccine). Here are the results:
You can see that for all adjuvants, there was much less to almost no cytokine response with adjuvant alone, despite 3 weekly injections, but levels were much higher when the vaccine and adjuvant were given together. eg for MF59-adjuvanted flu vaccine vs MF59 alone, peak levels for IL-5 were 2996 vs 567 pg/ml, and levels for IL-6 were 2774 vs 430pg/ml respectively. These are very big differences, significant especially since they are consistent for all adjuvants tested.
There are a few more things I want to point out about this paper, the significance of which will be apparent when we move on to subsequent publications:
- This is an in vivo experiment, ie you take a live animal, inject it with whatever you are testing, then some time later, you find out what happens to the animal, by blood tests, look at the tissues, autopsies, whatever. As opposed to in vitro tests, where you put some cells in a test tube, mix them with whatever you're testing, and off you go. Because biological systems are complex, networked, and interactive (see the NIH chart above), it is very hard to create an in vitro model that resembles the whole organism, which means that in general cell culture results are likely to be less reliable than animal studies.
- The main results were obtained from adjuvanted flu vaccines, and not just the adjuvants. This is a very important distinction to keep in mind, because we can see from these results, that the addition of a vaccine antigen dramatically raises the serum cytokine response, compared to the adjuvant alone.
- The authors mentioned 'Th2 responses'. Let me explain, because much of the rest this discussion depends on your understanding of this issue.
Th refers to T-helper cells, a subset of T-lymphocytes or T-cells which when activated, help to direct the nature of the immune response. Until recently, it was thought that there were 2 types of Th cells and Th responses - Th1 and Th2 (reviewed in this 1994 paper) - which were important for immunity against intracellular and extracellular pathogens respectively. As recently as 2002, scientists still held this somewhat simplistic view, as expressed in this paper in Nature Rev Immunology (Murphy 2002)
Th1 cells are responsible for cell-mediated immunity, whereas Th2 cells are responsible for extracellular immunity. In terms of the control of parasites, Th1 immunity provides protection against intracellular protozoa, such as Leishmania species and Toxoplasma gondii, whereas Th2 immunity is associated with protection against intestinal helminths. As well as their protective roles in host defence, both subsets of Th cell have been implicated in pathological responses. Th1 cells can mediate organ-specific autoimmunity, while Th2 cells have been implicated in the pathogenesis of asthma and allergy.
Different cytokines were classified according to the quality of response (ie Th1 or Th2) they elicited. Some cytokines clearly belong to one group than the other, but some were debatable. Some authors (below) attributed IL-6 to a Th2 response, but others remained agnostic, and did not classify IL-6 either as Th1 or Th2. In general, it would have been accurate to label IL-6 as either Th2 or Th0, ie neutral to these responses.
At the time the MF59 results were published in 1994, this was the accepted wisdom. Based on their results and the prevailing concepts of immunology at the time, it would have been logical, in 1994, to conclude that the systemic cytokine response from MF59-adjuvanted flu vaccines was unlikely to trigger autoimmune diseases.
However, in the last few years (starting with this seminal paper Harrington 2005), the Th1/Th2 paradigm was expanded to include a Th17 cell (named because of its association to the IL-17 cytokine), plus the inclusion of regulatory T cells or Treg, into this model. There is now abundant evidence that Th17 cells and Th17 responses play major roles in the pathogenesis of autoimmune diseases (see below). But the most intriguing discovery above all, is that IL-6 is responsible for switching from tolerance to autoimmunity against self-antigens. Here's how:
Bettelli 2007 Th17: the third member of the effector T cell trilogy.
Th17 cells have recently emerged as a third independent T cell subset that may play an essential role in protection against certain extracellular pathogens. However, Th17 cells with specificity for self-antigens are highly pathogenic and lead to the development of inflammation and severe autoimmunity. A combination of TGF-beta plus IL-6 and the transcription factors STAT3 and RORgammat were recently described to be essential for initial differentiation of Th17 cells and IL-23 for the later stabilization of the Th17 cell subset.
IL-6 plays a pivotal role in dictating whether the immune response is dominated by pathogenic Th17 cells or protective T-reg cells.
Afzali 2007 The role of T helper 17 (Th17) and regulatory T cells (Treg) in human organ transplantation and autoimmune disease.
..that many inflammatory or autoimmune diseases present clinically as episodes of inflammation (flares),with periods of quiescence in between these episodes, argues for the presence of intervening periods of 'equilibrium' where the immune system displays tolerance to self-components (i.e. that proinflammatory components are 'regulated'). During acute flares, a state of 'disequilibrium' ensues in which immune responses against self components are dysregulated. The possibility arises that during these episodes Th cell phenotypes become skewed towards proinflamatory lineages (Th17 and Th1) and away from anti-inflammatory phenotypes (Treg) on the basis of the local cytokine environment and DC populations. The mechanism of such a change could either be loss of skewing towards Treg phenotypes (with default towards Th1/Th17) or a primary shift towards the proinflammatory pathways. The central cytokine in this pathway, on the basis of the mouse data, may be IL-6, which is known to be elevated in most inflammator conditions. Presumably, the balance is redressed during the recovery from flares and the equilibrium re-establishes itself.
A picture is worth a thousand words. Here's a reasonably simple one, from Pernis 2009
A naïve or uncommitted Th cell can go in any one of several different directions, depending on which cytokines are present in the immediate micro-environment where the immune response is taking place, eg vaccination site and/or local lymph nodes. You can see the Th1 and Th2 cells, as per the older paradigm, but the important story is in the lower half of the diagram. Notice that the presence of TGF-beta favors the formation of regulatory T cells (Treg), whose job it is to suppress any excessive inflammation and/or inappropriate responses directed towards self antigens. However, the addition of IL-6 to that environment switches it towards producing Th17 cells instead, which as quoted above are highly pathogenic in inducing autoimmunity.
In addition, Th17 cells themselves produce IL-6, which in turn induces more Th17 cells, thus establishing a self-reinforcing loop:
Langrish 2005 IL-23 drives a pathogenic T cell population that induces autoimmune inflammation.
IL-6 is predominately produced by myeloid cells and is an important factor promoting inflammatory responses. The production of IL-6 by Th-17 but not Th1 cells is intriguing and may explain the hyperencephalogenicity of Th-17 cells.
As we have seen before, autoimmune diseases typically start from random triggers which once activated can lead to self-perpetuating cycle of tissue damage and increasing immune activation. Since regulatory T cells are one of the most important mechanisms for immune regulation (Fontenot 2005, Awasthi 2009) and since local inflammation eg after vaccination, results in tissue damage, cell death, and release of self-antigens that can activate T-cells (as described here), then a surge of IL-6 in the microenvironment may well be critical for turning a normal or harmless immune response eg to vaccination, to a destructive and pathological one, particularly in subjects with genetic predispositions to autoimmune diseases.
(See also Bettelli 2006, Mangan 2006, Xu 2007, Romagnani 2008.)
With this type of emerging new understanding, then the results of the 1994 study now carry dramatically different implications.
It's generally good practice in science, to have other researchers repeat experiments to see if they can reproduce those same results. It is also good practice, as a student of science, to search for additional research on the subject, and see if other, preferably independent researchers, have replicated or negated those findings. Unfortunately, because these adjuvants are proprietary products, you generally don't see independent researchers conducting such studies.
In between 1994 and now, Novartis has published a large number of papers, some of which offering great insight on the mechanisms of action of MF59 (Dupuis 1998, 1999, 2001), but as far as I can determine, the only other papers that touch on the cytokine response are 2 from 2008, ie after a gap of 14 years. Let's take a look (both papers are free). The first one is Seubert 2008 The Adjuvants Aluminum Hydroxide and MF59 Induce Monocyte and Granulocyte Chemoattractants and Enhance Monocyte Differentiation toward Dendritic Cells.
This is an in vitro study by cell culture, using different cell types obtained from blood samples from human donors. The 'human' part is good, we need to get information on humans. However, this may be more than negated by several issues.
- The first is that these cells are functioning in an artificial environment outside of the body. The importance of finely tuned regulation of the immune micro-environment in maintaining protective vs destructive responses is now firmly established. (Belkaid 2008, Stout 2004, Celli 2008). Some of the cells studied, like the dendritic cells (DCs) are very difficult to find in peripheral blood, and require extensive manipulation (Yi 2005) to create, including prolonged incubation with various cytokines like this one. If you take a look at their Materials and Methods section, the question arises: after so much manipulation, to what extent can the test conditions (and therefore the results) be relevant to the real-life situation of vaccine-in-arm? The limitations of cell culture models are widely known, (O'Neill 2004), ie not unique to this paper, but nevertheless we need to take that into account.
- The second is that experiments were done with just the different adjuvants, without any vaccine antigen in them. As we have seen from Valensi et al, the presence or absence of an antigen can make a huge difference to the quality and degree of response. Since people are going to receive the adjuvanted vaccine and not pure adjuvant, this adds another layer of limitation to the usefulness of the data.
- The third issue is IMO the critical one, and that is the absence of supporting data for some of the most important claims that the authors made. Now, if you read enough journal papers, you will come across occasionally authors making a remark peripheral to the issues under discussion, and rather than clutter the paper with yet more data, they might say 'data not shown'. This is acceptable practice for minor issues not germaine to the main discussion, but seldom have I come across papers where this happens more than once or twice. Well, Seubert et al is astonishing in that there were 13 instances of "data not shown", including where core issues are involved, eg what kind of cytokines does MF59 induce or not induce. Let me show you 3 examples:
In contrast, MF59 did not induce cytokine release in endothelial and epithelial cells lines or in monocyte-derived DCs, which are models for tissue-resident DCs (data not shown)
In contrast, LPS induces, among others, a number of proinflammatory cytokines such as IL-1-beta, IL-6, and TNF-alpha in endothelial and epithelial cells, macrophages, monocyte-derived DCs, and PBMCs. However, these were never found after incubation with MF59 (data not shown).
Taken together, our data indicate that MF59 induces a range of chemoattractants but not classical proinflammatory cytokines and that macrophages, monocytes, and granulocytes represent its three principal target populations. Alum, in contrast, induces a wider range of chemokines, including proinflammatory cytokines like IL-1beta (data not shown), and appears to act mainly on monocytes and macrophages and not on granulocytes.
The paper is very complicated; I admit to being utterly confused by some parts of it, but they did state very clearly that MF59 does not stimulate 'proinflammatory cytokines' such as IL-6, but no data is presented, so take it FWIW...
The second paper on MF59, is this one, in PNAS, Mosca 2008 Molecular and cellular signatures of human vaccine adjuvants There are several parts to this paper, but we'll focus on the local and systemic cytokine responses:
- First, the authors injected 1 single dose of different adjuvants (again without vaccine antigen) into mice and then measured the cytokine response in the muscle by gene expression profiling, which is a really cool way of measuring the activity of thousands of genes in one go. Changes in gene expression reflect changes in the molecular functioning of cells, including the induction of cytokines. Here's a diagram that shows the main cytokines and chemokines induced. Notice that IL-6 is induced (but only for a short time, for one reading at 3 hours after injection, as per additional chart and results in paper). This is compatible with the 1994 results for adjuvant-only injections, which elicited a much lower and more transient response compared to adjuvant+antigen.
- The authors also measured systemic cytokine response. They found that MF59 caused increased IL-5 levels peaking at 12 hours. However, they didn't comment on any other cytokines for MF59, so we don't know whether they tested for IL-6 and found it was not raised, or they did not test for it. It's an unfortunate oversight of a very important issue, but that's just how it is. Let's assume they tested for it and didn't find any.
Because they worked on >1,600 genes, it was of course not possible to list them in the paper itself, so additional datasets were provided in the supporting information. I poked around in there, and found the following which may or may not be significant (since I'm not an immunologist!)
- Among the many genes upregulated was one IL17c. When I looked it up, I found IL17c is synonymous with IL21, another proinflammatory cytokine (Nurieva 2007) secreted by Th17 cells. This was upregulated only for MF59 and no other adjuvants, starting at 24 hours, peaked at day 2, and fell slightly by day 4 ie lagging behind the IL6 response. So I looked up IL21 - it is activated by (and therefore operates downstream of) IL6, which sort of makes sense, since IL6 was induced much earlier, at 3 hours. More interestingly, I discovered that genetic polymorphisms in a common gene loci associated with IL-21 are strongly linked to a number of autoimmune diseases including type 1 diabetes, rheumatoid arthritis, celiac disease, inflammatory bowel disease, multiple sclerosis (Zhernakova 2007, Lettre 2008, Garner 2009, Glas 2009)
In summary, they found that MF59, given as adjuvant only, induced many cytokines and chemokines, some of which may be proinflammatory, with IL-6 among them, in the muscle after injection of plain adjuvant. But they didn't mention whether they tested for IL-6 in the serum or not. Assuming they did, a negative finding is not surprising given the transient induction in the muscle - there may simply not be enough IL-6 to spill into the circulation, under these test conditions.
So, what do we take from these 2 papers? Well, I have to tell you that I read them (and others) backwards and forwards I don't know how many times, and the only conclusion that I can come to, is that there is nothing in them to negate the findings from the 1994 study. That's just me, personally...
BTW I found some lists of diseases affected by IL-6, together with citations, from a nifty review of the rapidly expanding body of knowledge on this cytokine, (free!) by Hong 2007 Interleukin-6 and its receptor in cancer: implications for Translational Therapeutics. I have to warn you they may blow your mind...;-)
- Human illness and IL-6:
- IL6 and cancer:
- and, finally, just to show you the extent of scientific interest in this cytokine, examples of drugs under development, that target/block IL-6 specifically:
IN SUMMARY, we now know that MF59-adjuvanted flu vaccines induce high serum IL-5 and IL-6 in mice, especially for mice that have some immunity from prior infection. We also know that IL-6 can switch the immune environment from tolerance to autoimmunity. The list of diseases in which high or dysregulated IL-6 has a role, is long and constantly growing. We do not know whether AS03 has similar effects, since no data is available. (AS03 is explored in more detail in part 2 The Meaning of 'Reactogenicity')
In the next diary, we'll look at the role of IL-6 in the acute phase response and in vaccination reactions, to see if we can gain more useful insights, to help those of us who may shortly be offered this vaccine in their country. Those of you who have followed our recent debates on vaccine safety, may be familiar with the limitations of clinical trials and post-marketing data, here and here. The risk of vaccination has to be balanced against the risk of not getting vaccinated, but right now many people are faced with having to make such determinations with scant information.
Non-clinical studies like the ones we discussed, are important in giving us insights into disease causation, aka how things work, and are increasingly recognized by regulatory authorities as important tools especially to resolve complex safety issues, eg involving delayed and/or subtle outcomes, or risks that affect only certain subsets of populations. Unfortunately, efforts to set standards or regulatory requirements for non-clinical safety assessment of vaccines and adjuvants have proceeded at snail's pace, compared to the proactive attitude of industry and the astonishingly large amount of new knowledge continuously generated by researchers on immunology. In other words, governments are falling behind both industry and academia, but they are the ones who have their fingers on the button, as to whether certain vaccines should be approved for routine or mass vaccinations. Personally, I don't believe this is a good or sustainable formula for safer vaccines and/or improved public confidence, but as always your mileage may differ.
That said, there are intense debates going on, efforts to establish standard protocols for testing. One big problem is that there are few tools available to test for excessive or abnormal immunostimulation. A number of models are proposed. Interestingly one that is frequently suggested by industry itself, as a screening test for new products, is serum IL-6! (Kaplanski 2008, Lindblad 2007, Lebron 2007)
Now if that doesn't tell you something...
UPDATE: Part 2 of this series is here The Meaning of 'Reactogenicity'