Predictors to Direct Neural Antibody Testing in Autoimmune Encephalitis/Epilepsy



Divyanshu Dubey, M.B.B.S.        

Associate Professor of Laboratory Medicine and Pathology & Neurology
Division of Clinical Biochemistry 
Mayo Clinic, Rochester, Minnesota



Hello, I'm Div Dubey. I'm one of the neurologists and lab medicine pathologists here at Mayo Clinic. My job includes seeing patients with autoimmune neurological disorders. Also, evaluating patients for these rare antibodies which can often provide us insights into these treatable conditions.

Today, in this talk we will be discussing various clinical and paraclinical features of autoimmune encephalitis, autoimmune seizures, and epilepsy which can help us in terms of ordering various panels, which is likely to increase the yield of these ordering practices and potentially point us towards characterizing patients which are likely to be seropositive.


These are some of my disclosures. 

A clinical approach to diagnosis of autoimmune encephalitis

One of the major advances in our field has been the proposal of an expert consensus diagnostic criteria for autoimmune encephalitis. These criteria were put forward by many of our colleagues who are actively involved in clinical as well as research on autoimmune encephalitis. They divided the autoimmune encephalitis diagnostic criteria based on the degree of association or degree of certainty for the underlying diagnosis into possible, probable, and indefinite. And nearly all the definite cases had to have an underlying autoantibody biomarker for it to be considered a definite autoimmune encephalitis.

The various subcategories on these criteria included patients with limbic encephalitis, acute disseminated encephalomyelitis, brainstem encephalitis, among others. Utilizing these criteria, we looked at the epidemiology of this condition and we found that this condition is not as uncommon as once suspected. At least in Olmsted County, Minnesota, where this population-based epidemiology study was done, the prevalence of autoimmune encephalitis was found to be similar to infectious encephalitis. And most importantly, if you look at this particular figure, the incidence rate of autoimmune encephalitis seems to be increasing quite significantly. Nearly tripling from 2006 to 2015 compared to the numbers in 1995 to 2005, and a lot of this growth has been in the definite antibody-positive autoimmune encephalitis group, highlighting the importance of antibody evaluation in our diagnosis of these conditions.

If we extrapolate this data from Olmsted County, Minnesota, on a worldwide scale, there are about one million autoimmune encephalitis patients worldwide, and 90,000 people get this condition every year. 

Antibodies and syndromes

As I said, a lot of this has come from the growth in the number of biomarkers that have been discovered and also made available for clinical testing through labs such as ours. You can see back in the early 1990s, the number of biomarkers could have been counted on your fingers. But now, it has grown and become much more complex.

In this particular figure, I've divided these biomarkers into those which are potentially just red flags of an underlying autoimmune disorder and not directly pathogenic. Those are the ones in green, which are targeting intraneuronal or intracellular antigens. Some of these are markers of cytotoxic T cell-mediated disorders, example being PCA-1 or CdR2L or Kelch11, amongst others, whereas the ones below are targeting cell surface antigens, and these are directly pathogenic. This information also helps us from the lab standpoint, because some of the cell surface antigens we try to develop assays such as live CBA or flow-based assay or cell-based assays for detection of these antigens, whereas those with intracellular or intraneuronal antigens, we often confirm by tissue IF followed by Western blot or immunoDOT. So, it's not only meaningful for clinicians but also for laboratorians. 

Techniques for autoantibody discoveries

And a lot of this growth has been through the growth of discovery mechanisms. So initially, most of the discoveries used to be made by immunoprecipitation mass spec. But nowadays, things like protein array, phage immunoprecipitation sequencing have led to the discovery of multiple antibodies, for example, Kelch11 through PhIP-Seq, and many others. Some of them which are still in the pipeline.

So when should we consider autoimmune encephalitis, or how to recognize it when we are seeing these patients in the emergency department, in the outpatient clinic, or in the hospital? 

How to recognize autoimmune encephalitis (AE)1-3

Some of the common things to think about are, one, the clinical phenotype, subacute or rapid progression is you should make us think about, "Could this be an immune-mediated phenomenon?"

This could be in the form of dementia, encephalopathy, or seizures. Specifically, when we talk about autoimmune encephalopathies, autoimmune dementias as well. And often these patients tend to have a multifocal presentation, example being, encephalomyelitis where patients have not just cortical or encephalopathic presentations but can also have coexisting spinal cord and in some cases, polyradicular involvement. And that is a high-risk cancer phenotype which we see with paraneoplastic antibodies. The other thing which can sometimes make us think about autoimmunity is viral prodrome. So, presence of runny nose, sore throat a couple of weeks before the onset of encephalitis. That is seen in patients with NMDA receptor encephalitis as well as patients with MOG. And then with NMDA recently, there have been multiple reports of HSV encephalitis and then few weeks later, patient comes in for worsening encephalopathy and those patients tend to have sort of NMDA or an apparent infectious autoimmunity, leading to NMDA receptor antibody production and an autoimmune encephalitis. And then the other common thing is asking the patients for history of cancer, because there have been quite a few cases where we’ve had things like testicular cancer seminomas who go on to develop paraneoplastic syndrome secondary to that. 

MRI in autoimmune encephalitis1,4-9

The other thing which is very helpful is looking for MRI features. So nearly all of our encephalitis patients, either in the ED or in an outpatient practice, end up getting MRIs upfront as a part of the initial workup. And there are certain MRI features which can help us in identifying autoimmune encephalitis. So these are subcortical, sometimes the encephalic changes we see with MOG, they look fluffy, they look sort of involving the white matter. They can also have brain stem involvement at the level of the peduncles or the brainstem itself. Other features are those involving the mesial temporal lobe, often bilateral but asymmetric medial temporal lobe involvement in the form of limbic encephalitis like we see with LGI-1, but also with high-risk paraneoplastic antibodies like ANNA1 or anti-Hu or Ma2 as well.

Then certain antibodies tend to have the encephalic involvement where they have deep gray nuclear involvement, an example being CRMP5. Other examples being PD10A. And then hypothalamic involvement, especially in young men presenting with sleep disorder, presenting with worsening encephalopathy, should make us think about Ma2 antibodies. Whereas periradial enhancements or the extending rays of sun like picture on contrast enhancement. At times, with coexisting meningeal inflammation should make us think about GFAP. And then there are certain antibodies where the MRIs look completely normal. Most of the NMDA cases that I've seen in my practice don't usually have an MRI abnormality. So a lack of MRI finding or MRI abnormality does not rule out autoimmune encephalitis, but looking for these features can help us pick out which patient is likely going to be antibody-positive or going to have an autoimmune encephalitis even without an antibody.

60-year-old male with encephalopathy

The other benefit of ordering these panels is in certain cases when you are seeing an MRI picture which looks like a typical autoimmune limbic encephalitis, detection of antibody can not only help us with diagnosis, it can also have some predictive value in terms of long-term outcomes as well as cancer surge.

So these are two patients, both of them presenting with encephalopathy, both having medial temporal lobe changes, asymmetric as we discussed, T2 FLAIR hyperintensity. The first patient had ANNA1 or anti-Hu antibodies, whereas the second patient who had LGI-1 antibodies. In this particular patient, I would be concerned about an underlying cancer, often small cell lung cancer, and I'm going to carefully look for cancer. If I can't find it on CT scans, I'll probably go for PET scan, and I may even follow them every six months to nine months doing serial scans for the period of the next couple of years. Whereas, LGI-1, the risk of cancer is much lower. Only 5% of patients with LGI-1 autoimmunity have underlying cancer.

The other benefit of detection of these antibodies is it can help me have the conversation with the patient or the family in terms of the potential natural history of these illnesses. And we'll talk about that through a couple of cases we'll show next. 

Neural antibodies can inform clinical course and prognosis

So this is a gentleman who was seen in our Autoimmune Neurology Clinic, presenting with these very unique abnormal movements, happening multiple times a day. As you can see in this video, he's having these jerking movements which have a dystonic component to it. Interestingly, not just involving the face and arm consistent with facial brachial dystonic seizures, but you could also see jerking or dystonic movements or tonic movements of the legs. And this is a classic example of faciobrachial crural dystonic seizures, which is seen in association with LGI-1 one antibodies. And this patient was confirmed to be LGI-1 antibody-positive, was treated aggressively with immunotherapy upfront, and had complete resolution of these seizures without any significant cognitive deficit in the long run. And he was able to go back to his pre-treatment baseline.

Neural antibodies can inform clinical course and prognosis10

Whereas another patient who presented with seizures had T2 FLAIR hyperintensity involving the medial temporal lobes like we had seen in some of the examples before. Also had some gadolinium enhancement. Subsequent workup and antibody testing revealed that this patient was positive ANNA-1 or anti-Hu antibodies. And despite aggressive immunotherapy including consideration of cyclophosphamide, he continued to progress both in terms of cognitive impairment and went on to develop chronic epilepsy which was very refractory even to multiple anti-seizure medications and did not show much response to immunotherapy. So this is another example where certain antibodies have a much worse long-term prognosis. And detection of these syndromes early and also the type of antibodies can not only help us understand how aggressive to be upfront, choose the type of immunotherapies, we want to choose in terms of management of these patients but also give us an idea about what the natural course of the disease is going to be, setting up appropriate expectations for the patients as well as the family.

Now, since we've discussed the importance of antibodies, not only diagnosis, but also in terms of prognostication, how can we optimize our day-to-day antibody evaluation? One idea, which I thought of when I was in residency at UT Southwestern, ordering a lot of these panels externally, was coming up with a combinatorial model utilizing clinical and paraclinical features which are initially available to us when we are managing these patients and making the decision whether to send an antibody panel or not. And these include features such as presence of subacute encephalopathy, seizure characterization, seizure semiology, as well as sometimes CSF findings or MRI findings. 

Antibody prevalence in epilepsy and encephalopathy score (APE2 Score)11-12

So using some of these variables and using some of the data from our retrospective study, we came up with a composite model called the APE2 score. These are the variables we had just talked about, including presence of new onset seizures, the subacute progression, which we highlighted is one of the factors associated with autoimmunity. Presence of neuropsychiatric changes, autonomic dysfunction, viral prodrome like we talked about in NMDA or MOG. Certain seizure types like facial brachial dystonic seizures. Certain involuntary movements like facial dyskinesias. And then some of the paraclinical variables including some of the MRI findings I just talked about which can be seen in association with autoimmune encephalitis and the history of cancer. 

So individually, these factors were found to be significant and were found to be helpful. But the utility of combining them in a combinatorial model was that as a group, these factors had a very high sensitivity as well as relatively high specificity for prediction who's going to be antibody positive.

APE2 score11-12

In addition to making this model and doing some of the initial statistics, we also performed a Rasch analysis to demonstrate there was utility in combining these different variables together in a model. These individual variables had an independent effect on the model and were found to be relevant. And there was clearly a differential prediction for certain features where things like facial brachial dystonic seizures, which were clearly associated with a particular antibody having the highest predictive value. Where some of the CSF findings, which we found in our study, had lower predictive value but were still suggestive of sort of a neural positivity or detection of an underlying antibody. This APE2 score was not just validated for patients with epilepsy or seizures. This was also validated for cognitive dysfunction or patients suspected to have encephalopathy or dementia in terms of when to order autoimmune encephalopathy, autoimmune dementia panel.

APE2: Cognitive dysfunction

So this is sort of a flow diagram from our study where we reviewed about 424 patient charts. Characterized APE2 scores based on the information available, not just the clinical but some of the MRI and CSF data as well. And interestingly what we found was if we categorize patients with presence of antibodies versus no antibodies, those who had high APE2 scores, most of them fell into the group with neural-specific antibodies of clinical relevance. So, these were the patients who had antibodies which could be clearly associated with autoimmune encephalitis. We had also detected certain number of patients with nonspecific antibodies. These are the low-titer ganglionic receptor antibodies which were previously being tested for autoimmune encephalitis or voltage gated potassium channel antibody without coexisting CASPR2 LGI-1. And most of these cases had low APE2 scores. So what this tells us is that this model can also take out some of the nonspecific ordering practices, which often sometimes can result in nonspecific results and which in turn can cause problems where people can go on doing additional diagnostics to rule in or rule out this idea of autoimmunity.

Autoimmune encephalitis misdiagnosis in adults

And this was recently highlighted by a paper focused on analyzing misdiagnosis of autoimmune encephalitis at multiple centers with expertise in managing these cases. And it was found that among the 107 patients who were misdiagnosed, one of the most common reasons for misdiagnosis was overinterpretation of positive serum antibodies or nonspecific antibodies. So, utilization of either diagnostic criteria or composite models such as APE2 score, which have a high sensitivity or specificity for autoimmune encephalitis, in general, can prevent overinterpretation of some of these results. And unfortunately, these were the patients who with misdiagnosis, about one-fifth of the cases, or 20% of the cases, also had adverse events from various immunotherapies, which they technically shouldn't have gotten because they did not have autoimmune encephalitis.

APE2 Score: Patient 1

So now we'll talk about a couple of examples of how to use the APE2 score. So, this is a 58-year-old woman who presented to an outside facility. With electrical shock-like sensation involving the right side of her body. With periods of speech arrest along with few episodes of generalized tonic-clonic seizures. And three years later, she continued to have these refractory seizures along with oral lingual dyskinesias and stridor, and that's when she presented to Mayo Clinic. In this particular case, if we apply the APE2 score, what we will see is the patient will get a two for oral lingual dyskinesias, two for having refractory seizures. The one thing which I did not mention in the clinic case vignette was the CSF testing was done and was found to be slightly inflammatory, with two nucleated cells but a protein above 50 milligrams per dL. And then as we can see on the MRI, there's clear tubular hyperintensity involving the medial temporal lobes.

So if we combine this score, it would be a score of eight. So, anything above or equal to a score of four, based on the study, we suggest it should be considered for autoantibody evaluation. And in this case, when the panel was sent, she came back positive for ANNA2 or anti-Ri antibody. And again, this is a typical high-risk paraneoplastic antibody which can be associated not only limbic encephalitis but also autoimmune-associated epilepsy in the long run in some of these patients.

Patient 2

Another patient is a 71-year-old gentleman who presented with six-week history of frequent involuntary movements. You can see the jerking movements involving not just the right arm, but some of them also involving the left arm. And also, some dystonic changes involving the face. This reminds you of the other video we had seen before. This is a typical facial brachial dystonic seizure. In addition to that, he had some significant memory loss. An MRI brain was obtained, which was unremarkable. So, this particular patient ended up getting a score of four. One for neuro-onset seizures, three for facial brachial dystonic seizures, a total of four. And then the antibody as we could have predicted just looking at those movements, came back positive for LGI-1 IgG.

Patient 3

And then the third case, which I think is the most useful case in terms of why APE2 score is useful, is a 72-year-old gentleman presenting with progressive cognitive decline over a period of six years. MRI shows global atrophy including medial temporal lobe atrophy. And then CSF shows marginally high protein at 65 milligrams/dL. In this particular scenario, if you apply the APE2 score, the only changes you'll see are for the CSF, you'll get a score of two. So a total score would be lower than four, a total score of two. In this particular case, unfortunately, the ordering provider did send an antibody panel, but panel was negative. So, this is the right scenario where utilization of APE2 score can lead to avoidance of unnecessary antibody testing. And in some of these cases, if you get a potentially a false positive result such as a result of relatively nonspecific antibody like a voltage-gated potassium channel complex without LGI-1 or CASPR2, it can lead to autoimmune encephalitis misdiagnosis and patient receiving unnecessary treatment. And all this could be avoided by, not just utilization of APE2 score but also sort of thinking about the various diagnostic criteria which have been put forward in the last few years. 

Principles of neural antibody testing in AE13-15

So in terms of principle of neural antibody testing, what I suggest is test only if there's a reasonable clinical suspicion utilizing either composite model, such as APE2, or utilizing the diagnostic criteria. Consider sending panels before the treatment is initiated. There's no point if the patient receives multiple sessions of PLEX. The likelihood that the antibody will still be positive is relatively lower. In general, in most cases where a spinal tap is indicated, I would recommend sending both serum as well as CSF antibodies. I recommend testing a panel of antibodies rather than sending a panel at a time. Depending upon the clinical phenotype or the predominant phenotype, I would suggest choosing a panel. So if it's an encephalitis patient, send autoimmune encephalitis panel. If it's a myelopathy patient, consider sending a myelopathy panel. And confirmation by two different techniques, especially for certain antibodies such as GFAP or some of the high-risk antibodies such as an MPC1 or anti-Ri or ANNA1, usually has a higher clinical specificity. And then if the initial panel is negative, you could consider reaching out to either the lab which is doing the testing to look for research-based antibodies or consider sending to a lab which offers some of these research-based testing as well.

What things to keep in mind while interpreting positive or negative AB results?13-15

So how to interpret this panel or what to keep in mind while interpreting a positive or a negative result. If the panel is seronegative but you're concerned about an autoimmune encephalitis epilepsy, I would suggest, as I said, consider reaching out to the performing lab to see if any unclassified antibody was seen. Things like Kelch11 were considered unclassified up until a few years ago. And oftentimes, the way we communicated with the providers was either calling the providers or the providers calling us. If the specimen was tested, was CSF and turned out negative, consider sending the serum as well, if not already sent. There are certain antibodies like LGI-1, CASPR2 where at least based on the assays we are utilizing, serum seems to have slightly higher sensitivity, whereas the reverse is true for an NMDA or GFAP. As assays being utilized are extremely important if the antibodies are being reported through line blot or dot blot, it can sometimes generate both false negative but also false positive results. And then as I said, try to send antibodies before the treatment or immunotherapy is administered.

And more importantly, if the antibody result is positive, but the clinical phenotype is not suggestive of an autoimmune encephalitis and autoimmune epilepsy, we need to ask ourselves whether the specimen which was sent is appropriate because NMDA can generate a relatively high false positive result and if only serum is tested, I think somewhere in the range of 5% to 7%, what assay is being utilized? So, dot blots for Amphiphysin or immunoblots for Amphiphysin alone. In the absence of tissue, IFA have been shown to generate false positive results. And then certain antibodies don't really have high clinical specificity, example being striational, which have been removed from our panel. And then for certain antibodies like GAD65, titers matter. Only high titers, above 20 nanomoles per liter, tend to have a high specificity for GAD65, neurological autoimmunity such as GAD65-associated epilepsy.

Take-home points 

So, take-home points is one, autoimmune encephalitis incidents and prevalence seems to be similar to infectious encephalitis and it seems to be rising probably because of more recognition. More antibodies are being discovered with the growth of newer technologies such as PhIP-Seq. APE2 scores can aid in diagnosis as well as early treatment initiation for patients with autoimmune encephalitis or appropriate antibody panel sending. The most important thing is if patients with very low APE2 scores may not be an indication for autoimmune encephalitis, autoantibody testing, and this could help optimize some of the testing from various institutions. As we send these panels, it's always important to keep in mind a negative panel does not rule out a diagnosis of autoimmune encephalitis, as close to 40% to 50% of cases may still have autoimmune encephalitis but may be seronegative. And then lastly, if the antibody is positive, but the clinical syndrome doesn't fit, think of the potential alternative reasons. Could it be the assay, could it be the secondary utilized, or could it be sort of what specimen was sent?

Thank You


  1. Banks S, Sechi E, Flanagan E. Autoimmune encephalopathies presenting as dementia of subacute onset and rapid progressions. Ther Adv Neurol Disord. 2021 Mar 19.
  2. Tobin W, Strand E, Clark H et al. NMDA receptor encephalitis causing reversible caudate changes on MRI and PET imaging. Neurology 2014.
  3. Abboud H, Probasco J, Irani S, et al. Autoimmune encephalitis: proposed best practice recommendations for diagnosis and acute management. J Neurol Neurosurg Psychiatry. 2021;92:757-768.
  4. Aksamit A, Weinshenker B, Parisi J. Chronic microglial encephalomyelitis (CME). Ann Neurol. 2012;72(Suppl 16):S110.
  5. Flanagan E, Hinson S, Lennon V, et al. Glial fibrillary acidic protein immunoglobulin G as biomarker of autoimmune astrocytopathy: Analysis of 102 patients. Ann Neurol. 2017 Feb;81(2):298-309.
  6. Flanagan E. Paraneoplastic disorders of the nervous system. Continuum (Minneap Minn). 2020 Dec;26(6):1602-1628. 
  7. Madhavan A, Carr C, Morris P, et al. Imaging review of paraneoplastic neurologic syndromes. AJNR Am J Neuroradiol. 2020 Dec;41(12):2176-2187.
  8. Flanagan E. Paraneoplastic disorders of the nervous system. J Neurol. 2021 Dec;268(12):4899-4907. 
  9. Zhao-Fleming, Valencia Sanchez C, Sechi E, et al. CNS Demyelinating attacks requiring ventilatory support with myelin oligodendrocyte glycoprotein or aquaporin-4 antibodies. Neurology. Sep 2021, 97 (13) e1351-e1358.
  10. Smith K, Britton J, Thakolwiboon S, et al. Seizure characteristics and outcomes in patients with neurological conditions related to high-risk paraneoplastic antibodies. Epilepsia. 2023 Jun 27.
  11. Dubey D, Kothapalli N, McKeon A, et al. Predictors of neural-specific autoantibodies and immunotherapy response in patients with cognitive dysfunction. J Neuroimmunol. 2018 Oct;323:62-72.
  12. Dubey D, Pittock S, McKeon A: Antibody prevalence in epilepsy and encephalopathy score: increased specificity and applicability. Epilepsia. 2019 Feb;60(2):367-369.
  13. Dubey D, Singh J, Britton J, et al: Predictive models in the diagnosis and treatment of autoimmune epilepsy. Epilepsia. 2017 Jul;58(7):1181-1189. 
  14. Flanagan E. Paraneoplastic disorders of the nervous system. Continuum (Minneap Minn). 2020 Dec;26(6):1602-1628. 
  15. Flanagan E. Paraneoplastic disorders of the nervous system. J Neurol. 2021 Dec;268(12):4899-4907. 
  16. Sabater L, Saiz A, Dalmau J, Graus F. Pitfalls in the detection of CV2 (CRMP5) antibodies. J Neuroimmunol. 2016 Jan 15;290:80-3.
  17. Budhram A, Nicolle M, Yang L. The positive predictive value of onconeural antibody testing: a retrospective review. Can J Neurol Sci. 2018 Sep;45(5):577-579.


Contact us:

Image credit: Shutterstock

MCL Education

This post was developed by our Education and Technical Publications Team.