For anti-tau antibodies, researchers are still learning what strategies work best. Sometimes, however, the lessons are unclear. Case in point: Two new papers conflict about which antibody isotype ameliorates tau pathology most effectively. In the May 12 Science Translational Medicine, researchers led by Reina Fuji and Geoffrey Kerchner of Genentech in San Francisco report preclinical and Phase 1 data from the company’s anti-tau antibody, semorinemab. The data explain the rationale for using an IgG4 backbone. In cultures of mouse neurons and microglia, the IgG4 version protected neurons from tau toxicity better than did an IgG1. This may be because the IgG1 isotype binds more strongly to microglial receptors, causing these cells to release pro-inflammatory cytokines that damage neurons, the authors noted.

  • An IgG4 version of semorinemab protected neurons better than did an IgG1 version.
  • For AX004, an IgG1 version best cleared tau oligomers.
  • The optimal isotype may depend on an antibody’s intended mechanism of action.

However, last year, another paper using a different experimental paradigm came to the opposite conclusion. In the May 29, 2020, Acta Neuropathologica Communications, researchers led by Monika Zilkova of Axon Neuroscience, Bratislava, Slovak Republic, and Jeroen Hoozemans of Vrije University, Amsterdam, suggest IgG1 antibodies as the better option for anti-tau therapy. In human microglial cultures, IgG1s promoted more tau uptake than did IgG4s. In contrast to Genentech’s study, neither isotype triggered cytokine release.

What to make of this? To Martin Citron of UCB Pharma, Brussels, the data are not directly comparable. “They are looking at different settings and readouts,” Citron wrote to Alzforum. In general, isotype choice should be driven by the mode of action of a particular antibody, he suggested. Some tau antibodies are designed to stop the protein from spreading, while others are slated to clear tangles. It remains unknown what strategy will work best in the human brain, because no anti-tau antibodies have posted positive efficacy data yet.

Semorinemab was developed by AC Immune and Genentech, the latter being part of the Roche group. The antibody binds to the N-terminus of tau and recognizes all six human isoforms, regardless of their phosphorylation state. It was designed to stop the spread of tau pathology by preventing neurons from taking up extracellular tau. In September 2020, Roche reported top-line results for the Phase 2 TAURIEL study of 457 prodromal AD patients, in which semorinemab neither slowed cognitive decline nor budged tau PET. It found its target, as seen by less tau in cerebrospinal fluid, but did not affect downstream biomarkers of degeneration and inflammation (Mar 2021 conference news). The Phase 2 LAURIET trial in moderate AD continues.

In the new paper, Fuji and colleagues describe the preclinical data underlying their approach. They added recombinant tau oligomers to primary cultures of mouse hippocampal neurons, then treated them with semorinemab that had either an IgG4 or an IgG1 backbone. Either isotype worked equally well to prevent tau uptake and protect neurons from toxicity. When the authors added mouse microglia to the mix, however, things changed. While both isotypes still suppressed tau uptake, the IgG1 version failed to protect neurons. Neuronal damage, as measured by fragmentation of the cytoskeletal protein MAP2, was as high in IgG1-treated cultures as in untreated ones. Based on previous work, the authors believe this is because IgG1 stimulated microglia to release harmful cytokines (Lee et al., 2016). 

Because the goal of semorinemab treatment is to stop neuronal uptake of tau oligomers, it is not necessary for the antibody to also activate microglia, the authors noted. Supporting this, either isotype of semorinemab lowered soluble phospho-tau in P30lL tauopathy model mice. “[This] convinced us that a reduced effector function IgG4 antibody may have the best therapeutic properties on balance overall,” a Genentech spokesperson wrote to Alzforum.

However, Einar Sigurdsson of New York University School of Medicine noted that p-tau was only modestly reduced in these mice. “Based on these findings, it is not clear to me why this antibody, and its effectorless version, was selected for clinical trials,” he wrote to Alzforum. Nor does the published data provide insight into why the treatment failed in trials, he added. “Hopefully, additional analyses of these subjects, including quantitation of various tau species, and the ongoing trial, will shed light on this,” Sigurdsson wrote.

Meanwhile, Zilkova and colleagues studied a new anti-tau antibody, AX004. This is a humanized version of the mouse antibody DC8E8, which was generated using the same tau epitope deployed in Axon Neuroscience’s active tau vaccine AADvac1 (Apr 2020 conference news). AX004 recognizes truncated tau fragments missing their N-termini.

The authors tested IgG1 and IgG4 versions of AX004 in pure microglial cultures derived from postmortem human control or Alzheimer’s disease brain. The scientists added oligomers of truncated tau and, 20 minutes later, assessed how well microglia took them up. AX004-IgG4 boosted microglial tau uptake twofold; AX004-IgG1 2.5-fold. In other words, IgG1 stimulated 50 percent more phagocytosis than did IgG4. The source of the microglia made no difference, with those from AD brain responding similarly to control microglia.

What about inflammatory responses? Human microglia in culture spewed cytokines after being exposed to tau, but neither antibody isotype caused any additional increase. The authors concluded that both isotypes were equally safe, hence the more effective IgG1 would be preferable for human therapy.

Commenters noted several limitations of this study. Cynthia Lemere of Brigham and Women’s Hospital, Boston, cautioned that the Axon Neuroscience group did not directly examine the neurotoxicity of either isotype. One way to do this would be to add conditioned media from the treated microglial cultures to neurons, she suggested. Sigurdsson emphasized that microglia behave differently in a dish than in brain, leaving the significance of in-vitro findings unclear (Jul 2016 conference news; Jun 2017 news). Likewise, Citron wondered if culturing microglia with the growth factor GMCSF, as was done in these experiments, might have caused them to assume a phenotype more akin to dendritic immune cells, which primarily present antigens to B and T cells and lack some of the specialized functions of microglia.

So: Which isotype is better? For anti-Aβ amyloid therapy, where the goal is to clear plaque, IgG1 antibodies such as aducanumab and gantenerumab have had more success in clinical trials than IgG4 antibodies such as crenezumab. For anti-tau immunotherapy, however, this may be less of an issue.

Citron noted that UCB Pharma’s anti-tau antibody beprenemab, currently in Phase 1, is, like semorinemab, intended to prevent tau uptake by neurons, rather than to activate microglia. “We neither envisioned a need for high immune effector function nor a need to completely suppress it, so we chose IgG4 with relatively low immune effector function,” he wrote to Alzforum.

The two anti-tau antibodies gosuranemab and tilavonemab also have IgG4 backbones; both have posted negative Phase 2 results in progressive supranuclear palsy, but continue in AD trials (Jul 2019 news; Dec 2019 news). 

Except for bepranemab, the newer crop of antibodies favors the IgG1 isotype. These include BIIB076, which binds in the N-terminal quarter of the protein, E2814, and JNJ-63733657, which bind the microtubule-binding domain region, and LuAF87908, which recognizes an epitope even further toward the C-terminus. The field may have to await more data to find out which works better in clinical practice.

Sigurdsson noted another, more fundamental, wrinkle. Because most tau is intraneuronal, antibodies may have to be taken up by neurons to dramatically affect accumulation. In that case, an antibody’s electrical charge, which appears to influence its neuronal uptake, may be more important than its isotype or effect on microglial phagocytosis, he suggested (Congdon et al., 2019).—Madolyn Bowman Rogers


  1. These reports by Ayalon and Zilkova are helpful to the field of tau immunotherapy. It is interesting that they do not reach the same conclusion regarding the importance of antibody effector function for microglial phagocytosis. Similar differences between studies have been reported previously for Aβ antibodies and likely relate to the models used and/or experimental design. The argument for considering this issue is also much stronger for Aβ antibodies because most of Aβ pathology is extracellular, whereas most of tau pathology is intracellular. Culture studies are necessary for mechanistic insight, but it is well established that microglia in a dish behave differently than in animals, which makes it difficult to extrapolate such data to any in vivo situation.

    There is not a tremendous difference in the effects of the isotypes in the study by Zilkova and colleagues, and they acknowledge that the strong effect of tau alone on microglial cytokine production makes it difficult to say much about effects of isotype on cytokine release. It would be helpful if the authors could comment on how the tau concentration in the culture assay relates to its extracellular levels in vivo. The authors mention in their article that the justification for Genentech/AC Immune’s claim for enhanced safety of tau antibodies that have no effector function appears to rest on them not causing MAP2 fragmentation in culture. This ideally should be supported by additional assays. It is doubtful that Genentech/AC Immune would have made such an important decision without additional data on other toxicity markers to support it, but it is not included in their current article.

    Given the inherent difficulties in relying on and comparing culture data using different models and/or study design, Ayalon et al. rightly compared the two versions of their mouse antibody in a tauopathy mouse model. The version with effector function was effective at a lower dose than the effector-less antibody, but their clearance of pathological tau was not robust. On brain sections, it reduced one phospho-tau epitope but had no effect on a different soluble phospho-tau epitope on western blots. Possible effects on the insoluble tau fraction seen on western blots do not appear to have been analyzed. Further, comparison of the two antibody versions on astro- and microgliosis did not reveal any effect of either antibody compared to controls, which is in line with prior in vivo studies on other tau antibodies.

    In summary, the murine version of semorinemab was modestly effective in the mice, a version with effector function worked at a lower dose than its effector-less counterpart, and neither one was toxic in these animals. Based on these findings, it is not clear to me why this antibody, and its effector-less version, was selected for clinical trials. Functional studies are not included in the article but I would assume that some behavioral or neuronal activity benefit in animal studies would have been needed to green-light clinical trials, and presumably will be published in due course. The data from the nonhuman primates and the clinical plasma data is informative, but does not provide insight into why the trial failed. Hopefully, additional analyses of these subjects, including quantitation of various tau species, and the ongoing trial with tau PET imaging will shed light on this.

    Comparing the efficacy of different isotypes of tau antibodies is certainly warranted, and is being done in my lab, but it is important to remember that most of pathological tau is inside neurons, where several tau antibodies can target it (for recent reviews see Sandusky-Beltran and Sigurdsson, 2020; Sigurdsson 2019). The electrical charge of the antibody, which appears to influence its neuronal uptake, may be a more important consideration than its effector function for microglial phagocytosis (Congdon et al., 2019). It makes sense to target pathological tau where most of it is located, namely inside neurons.


    . Tau immunotherapies: Lessons learned, current status and future considerations. Neuropharmacology. 2020 Sep 15;175:108104. Epub 2020 Apr 28 PubMed.

    . Alzheimer's therapy development: A few points to consider. Prog Mol Biol Transl Sci. 2019;168:205-217. Epub 2019 Jun 26 PubMed.

    . Tau antibody chimerization alters its charge and binding, thereby reducing its cellular uptake and efficacy. EBioMedicine. 2019 Apr;42:157-173. Epub 2019 Mar 22 PubMed.

  2. Understanding the role of the isotype in developing antibody therapies for Alzheimer’s disease is indeed critical. The work from our group also supports the use of tau antibodies in an IgG4 (low effector function) isotype in clinical trials. In a direct comparison of our 2N tau-specific antibody, RN2N, in a murine IgG2a (high effector function) and murine IgG1 (low effector function) format, we demonstrated that IgG2a, but not IgG1, increased the secretion of pro-inflammatory cytokines in vitro (Bajracharaya et al., 2021). Furthermore, only the IgG1 isotype was able to reduce microgliosis in vivo and demonstrated enhanced efficacy compared to IgG2a following treatment of P301L tau transgenic mice. Together, our work and that of Ayalon et al. demonstrate that high microglial activation is not required for tau antibody efficacy and suggests that enhanced pro-inflammatory cytokine secretion may hinder antibody efficacy in vivo.


    . Tau antibody isotype induces differential effects following passive immunisation of tau transgenic mice. Acta Neuropathol Commun. 2021 Mar 12;9(1):42. PubMed.

  3. This is an interesting report by Ayalon et al. on humanized, anti-tau, N-terminus-specific IgG4 monoclonal antibody (mAb), semorinemab. This mAb did not slow the rate of clinical decline in early AD (prodromal to mild) compared to placebo, but it was safe even at very high doses (~34g/month).

    Data presented by the authors (Fig S5) demonstrated that both IgG1 and IgG4 semorinemab inhibited oligomeric tau uptake by mouse hippocampal neurons. However, in the presence of microglia, only the latter mAb was protective against exogenous tau-mediated toxicity in cultured neurons.

    These data support their previous report demonstrating that effector function of the mAb is not required for the reduction of tau pathology (Lee et al., 2016). However, recent results with another humanized mAb, AX004, demonstrated that both IgG1 and IgG4 variants stimulated abnormal tau protein uptake by human adult primary microglial cells (Zilkova et al., 2020). Importantly, although Zilkova and colleagues found that IgG1 more effectively promoted tau uptake than the IgG4 isotype, neither mAb-tau complex increased production of pro-inflammatory (IL1β, IL6, IFNγ, and TNFα) or anti-inflammatory (IL4 and IL10) cytokines by human primary microglia compared with a tau control. Interestingly, the  authors also showed that Fab fragments of IgG1 and IgG4 AX004 alone do not promote tau uptake by human microglia. 

    It is difficult to compare these studies. Ayalon and coauthors used mouse neurons for tau uptake studies and microglia-neuron co-culture along with humanized IgG1 and IgG4 mAb with human Fc fragments. Of note, it was reported that human IgGs recognize mouse and human FcγR with similar binding affinity. Zilkova et al. generated primary human microglia stimulated in vitro for eight to 14 days with GM-CSF and used them with humanized AX004 IgG1 and IgG4 mAbs.  Importantly, they did not use human neurons co-cultured with these microglia.

    While further studies may help us to understand the significance of full-effector or effector-less mAbs for reduction/clearance of pathological molecules involved in neurodegenerative diseases, I need to mention, regrettably, that all AD immunotherapy trials have been uniformly unsuccessful. Although some clinical trials are still in progress, overall, the field is moving to earlier interventions (i.e., preventive treatment of people at preclinical phase).  

    Unfortunately, due to the complexity, the cost, and the need for monthly intravenous injections of asymptomatic people with high concentrations of mAb, passive vaccination is impractical for preventive treatment. By contrast, active vaccines generating sufficient levels of immune responses may protect the host from overt diseases and have been used as preventive measures for over 100 years. Therefore, it is likely that safe and immunogenic preventive Aβ and tau vaccines, or even dual vaccine targeting both pathological molecules initiated in non-demented subjects based on disease-related blood, brain, and CSF biomarkers, may inhibit/reduce oligomerization of Aβ and tau aggregation and delay downstream pathological processes.


    . Antibody-Mediated Targeting of Tau In Vivo Does Not Require Effector Function and Microglial Engagement. Cell Rep. 2016 Aug 9;16(6):1690-700. Epub 2016 Jul 28 PubMed.

    . Humanized tau antibodies promote tau uptake by human microglia without any increase of inflammation. Acta Neuropathol Commun. 2020 May 29;8(1):74. PubMed.

Make a Comment

To make a comment you must login or register.


Therapeutics Citations

  1. Semorinemab
  2. AADvac1
  3. Gantenerumab
  4. Crenezumab
  5. Bepranemab
  6. Gosuranemab
  7. Tilavonemab
  8. BIIB076
  9. E2814
  10. JNJ-63733657
  11. Lu AF87908

News Citations

  1. N-Terminal Tau Antibodies Fade, Mid-Domain Ones Push to the Fore
  2. Active Tau Vaccine: Hints of Slowing Neurodegeneration
  3. When a Microglia Is No Longer a Microglia
  4. What Makes a Microglia? Tales from the Transcriptome
  5. AbbVie’s Tau Antibody Flops in Progressive Supranuclear Palsy
  6. Gosuranemab, Biogen’s Anti-Tau Immunotherapy, Does Not Fly for PSP

Paper Citations

  1. . Antibody-Mediated Targeting of Tau In Vivo Does Not Require Effector Function and Microglial Engagement. Cell Rep. 2016 Aug 9;16(6):1690-700. Epub 2016 Jul 28 PubMed.
  2. . Tau antibody chimerization alters its charge and binding, thereby reducing its cellular uptake and efficacy. EBioMedicine. 2019 Apr;42:157-173. Epub 2019 Mar 22 PubMed.

Other Citations

  1. aducanumab

Further Reading

Primary Papers

  1. . Antibody semorinemab reduces tau pathology in a transgenic mouse model and engages tau in patients with Alzheimer's disease. Sci Transl Med. 2021 May 12;13(593) PubMed.
  2. . Humanized tau antibodies promote tau uptake by human microglia without any increase of inflammation. Acta Neuropathol Commun. 2020 May 29;8(1):74. PubMed.