Influenza virus haemagglutination: a sticky technique that does a lot of lifting

Figure 1. A 96-well, U-bottom plate used in a haemagglutination assay to test for flu virus. The U-shaped bottom lets the RBCs slide to the bottom…unless they have been agglutinated.

Haemagglutination (hemagglutination if you’re from the US) is the sticking (-agglutination) together of red blood cells (RBCs). Haem is the iron-containing, oxygen-transporting portion of haemoglobin, a molecule found in RBCs.

This laboratory method has been leveraged to identify blood types, the presence of antibody to past infections, current bacterial infection and current infections by certain viruses including mumps virus, measles virus, adenoviruses, arboviruses, poxviruses, polyomaviruses, parainfluenza viruses and influenza (flu) viruses.[1,2,12,13,14]

I’m focussing on flu viruses here and for them, haemagglutination is a result of the reaction between the viral envelope protein, haemagglutinin (shortened to HA; see Figure 2) and its sialic acid receptor molecules found on the glycoproteins and lipids of the RBCs of some animals. Haemagglutination happens because a latticework of virus attached to RBCs forms. This latticework prevents a suspension of RBCs from doing what they’d otherwise do, settling to the bottom of a container over time.

Figure 2. The flu virus haemagglutinin (HA) gene segment codes for the HA glycoprotein. Along with M2 protein and neuraminidase (NA) glycoprotein, HA is  embedded in a double layer of lipid, derived from the host cell.

Haemagglutination methods for free flu virus were in use long before we had today’s more sensitive and highly virus-specific PCR-based methods to detect viruses.[1,5,6] The ‘Hirst phenomenon’, named after the author of the first description,[5] was later found to apply to other viruses.[1] Haemagglutination is still in use today, nearly 80 years later.[9]

Haemagglutination needs intact virus proteins, preferably in virus form, to bind to multiple receptor molecules and create that latticework. This means we need to ‘culture’, ‘isolate’ or ‘grow’ virus from a patient sample using a method usually based on egg or cell culture methods.

Figure 3. Haemagglutination (Panel 1 and 2) and haemagglutination inhibition tests (Panel 3 and 4). 10.6084/m9.figshare.6225218
Click on image to enlarge

We can also leverage haemagglutination to create a slightly more complex lab test; the haemagglutination inhibition (HI) assay (Figure 3). An HI can tell us if something is present that prevents haemagglutination. This can be used in a couple of ways to tell us:

  1. The type or subtype of flu virus (virusX) in a viral culture made from an infected patient’s (patientX) sample. This needs a panel of known laboratory-made/commercial anti-flu virus antibodies (Figure 3, Panel 4). It also needs experience in growing virus using cell culture. This panel is usually of limited size in most laboratories. The minimum panel only identifies flu viruses related to those expected to be circulating. A bigger panel is used if trying to find out what is circulating.
    HI isn’t a primary diagnostic tool for use on a patient swab or blood samples.
    HI also confirms that a virus was successfully grown from a specimen at all.
    Virus culture is insensitive and slow compared to RT-PCR methods but is essential if looking at the immune properties of a flu virus.
  2. The presence and specificity of anti-flu virus antibody in the serum sample of patientX who has generated a response to a flu virus infection (Figure 3 Panel 3). 
    This method uses a pre-made or commercial panel of known virus antigens (the parts recognised by our immune response to an infection) to determine which flu virus patientX was infected by, in the recent or distant past. Once again, the size of the panel depends on the role of the laboratory.

Some laboratories only need to know roughly which flu type (A or B) and subtype (e.g H3N2, H1N1, VICTORIA or YAMAGATA lineage) are present.

The National Influenza Centres (NIC) and World Health Organization Collaborating Centres (WHO CCs) for Influenza – delve more deeply. They also sequence some or all of the gene segments of the flu viruses they receive and conduct in-depth flu virus genetic and antibody studies.

Other labs want to understand how effective a vaccine has been. They will conduct virus and antibody studies to determine if the flu viruses circulating among infected people this season are similar enough to those used in the vaccine. Do they match or is there a mismatch? If they match well, there is a good chance that a vaccinated person will mount an immune response that protects them from disease if they get infected by the virus.

In the rest of this post, I’m going to delve further into No.1; detecting, typing and subtyping flu virus after it’s been grown in culture.

Proving you have flu virus in your suspected culture of flu virus…

Sounds stupid, but before wasting time and money using special antibodies to confirm the type and subtype of the flu virus you isolated/grew from patientX, it’s worth checking you did grow flu virus from that patient’s swab or aspirate in the first place.

Flu virus typing by HAI

Figure 4. An example of how the lab dilutes a suspected virus culture sample then tests it for the presence or absence of flu virus. The dilution allows the lab to also test how much virus there was to start with. The addition of blood cells occurs after the virus is diluted.
Click on image to enlarge

To do this, some of the suspect culture can be added to a well of a microwell plate and diluted in salty water (actually phosphate buffered saline; a common lab dilution liquid, or ‘diluent’).

Half of that dilution is transferred by micropipette to the next well, containing an equal volume of PBS. This is mixed and the process is repeated. This creates a dilution series. In this series, there will be half as much virus in the next well (Figure 4) which are called doubling dilutions.

Half an hour later, RBCs from an appropriate animal [6] are added and incubated for 30-60min.

If there is flu virus in the culture, it will agglutinate the RBCs and the microwell will look like it has a diffuse red coating over the bottom of the microwell (latticework). If there is no flu virus present, and thus no haemagglutination, the RBCs will slide down to the bottom of the well in a heap and look like a red dot (also called a button or pellet).

Example influenza haemagglutination result.

Figure 5. There was no virus in Row F or G. Virus in Row H caused haemagglutination to 12 (1 divided by the last agglutinating dilution); wells H1 and H2 contained enough virus to agglutinate RBCs.

How much of a good thing?

With a little extra work, haemagglutination can tell us how much virus was originally in the culture.

In the real example in Figure 5, patientX’s cultured virus was used in Row H. It agglutinated at the 1:6 and 1:12 dilutions (diffuse coating due to the lattice formation) but the next dilution didn’t agglutinate RBCs (red pellet or button). We say this virus had an HA titre of 12 or that it contained “12 HA units” of a flu virus.

If there had been more flu virus, it would have taken more dilution steps before the agglutination effect was removed. Dilution reduces the number of virus particles because we’ve kept halving their number, replacing the half with more PBS diluent. We say we’ve ‘diluted out’ the effect.

We need to know the HA units of virus in a patient’s culture for the next step – identifying which subtype of flu patientX has been infected with.

Using known reference antibodies to find out which flu virus the patient has

Figure 6. HI typing assay reagents prepared by the Melbourne WHO Collaborating Centre.[4]
Click on image to enlarge

Once the presence and amount of flu virus are confirmed, an HI can be used used to identify the type (FluA or FluB) and subtype (for FluA) or lineage (for FluB) of flu virus present.(Figure 7)

We can also do this using type- and subtype-specific RT-PCRs, but that is for another post.

This experiment uses a fixed amount of the patient’s virus isolate – 4 HA units. In our example, patientX’s cultured virus gave us a 12 HA units, so we’d need to make a 1:3 dilution to get our 4 HA unit working stock. Next, we need known reference antibodies. Each antibody preparation (antisera) binds to different flu virus subtypes.

Figure 7. Typing and subtyping of PatientX flu virus isolated using known antibodies in a haemagglutination inhibition (HI) reaction.
Click on image to enlarge.

These antisera preparations are made by exposing a rabbit or ferret to a purified stock of a FluA subtype or FluB lineage. After a wait, their antibody-containing serum can be collected. A dilution series of each antiserum is made as we described earlier. Each antiserum in the example kit in Figure 6 reacts broadly (so could react weakly or miss subtle variants) with vaccine-like flu strains. Results will be described as FluB/www, FluB/xxx, FluA/yyy or FluA/zzz-“like”, where www, xxx, yyy and zzz represent the most likely 2 FluA and 2 FluB strains to circulate in a given year.


Four HA units of virusX are added to each well of a row followed by the RBCs. This is mixed and left to sit for 45 minutes to allow agglutination – all as we did before. The patterns of agglutination look the same as in Figure 5 but this time they mean different things.

If there is a red dot, as shown in Figure 5, Row F and G, there was no lattice forming since haemagglutination was inhibited. This occurred because virus-specific antibody coated virusX,  preventing it from binding to its receptors on the RBCs. If there was a diffuse coating (a latticework formed) on the bottom of the microwell, then there was no reaction with that antibody, which means that virusX did not react with that particular antiserum and did not belong to that FluA subtype/ or FluB lineage. This time, it’s the absence of agglutination that is the positive reaction.

Some things that can cause problems for HI assays and for flu research

Genetically distinct viruses may not be identifiable as clearly antigenically distinct. These may have other differences which mean they don’t generate equally strong immune responses in humans. Different tests (neutralization assay or plaque reduction) may need to be used.

The way flu viruses are isolated or grown makes a difference to their agglutinating properties.[4] If the agglutination part falls over, then HI will be affected. Flu viruses grown using eggs can haemagglutinate a wide range of RBC species.[6] Flu viruses grown in MDCK and MDCK-SIAT-1 cells may not be as good for agglutinating fowl RBCs as guinea pig, turkey and human group O RBCs.

Recently, FluA/H3N2 isolates have struggled to agglutinate RBCs from fowl or turkey. Guinea pig and human group O RBCs are best for these strains.

Patient serum may contain inhibitors of agglutination, including traces of antivirals drugs which can affect haemagglutination.

Most antibodies produced against a flu infection that neutralize flu virus, target the globular end, or head, of the haemagglutinin. The head is also the most variable region precisely because of its exposure; new flu viruses with mutations that produce slightly different shaped head regions will escape antibodies and be able to reproduce.

Not all neutralizing antibodies are directed toward the head though and these others don’t work well in the HI assay.[8,9]  This is a big reason for why it has taken so long to identify regions for a ‘universal’ flu vaccine. The stalk or stem portion of the haemagglutinin protein is one such candidate region (Figure 8).[9,11] The head, not the stem attached to cells so the HI assay won’t be inhibited by antibodies targetting the stem. In us, antibodies to less variable regions can neutralize not just one strain, but many and different subtypes.[9] 

Figure 8. Influenza A H1 haemagglutinin model. The model highlights the protruding head of one of the 3 HA molecules that form a trimer. The head protrudes from the viral envelope. The HA head is under more pressure from our immune system and changes more frequently than the stalk, or stem. Change helps it evade B-cell produced antibodies. Image adapted from Raymond et al, PNAS (2018) 115(1) 168-173 [10]

In the past, results from HI assays were used as a major correlate of vaccine effectiveness.[7] It is possible that HI assays – if not accompanied by data from a human trial – may both falsely report good vaccine effectiveness and poor vaccine effectiveness if the test cannot capture the absence or presence of an immune response to other important sites on the virus. 

Labs find flu strains that are difficult to type. These are sent to a WHO CC for influenza for more extensive analysis.[4]


  2. Lennette, E. H. (ed.). 1992. Laboratory Diagnosis of Viral Infections, 2nd ed. Marcel Dekker, Inc., New York, N.Y.
  3. Reverse transcription polymerase chain reaction (RT-PCR)…a primer for virus detection
    Prepared by the WHO Collaborating Centre for Reference and Research on Influenza.
  6. The Adsorption of Influenza Virus by Red Cells and a New in Vitro Method of Measuring Antibodies for Influenza Virus
  7. Influenza Hemagglutination-Inhibition Antibody Titer as a Correlate of Vaccine-Induced Protection
  8. Universal epitopes of influenza virus hemagglutinins?
  9. A universal problem
  10. Conserved epitope on influenza-virus hemagglutinin head defined by a vaccine-induced antibody.
  11. A highly conserved neutralizing epitope on group 2 influenza A viruses.
  12. A New Test for the Detection of Weak and “Incomplete” RH Agglutinins
  13. Blood Group Typing: From Classical Strategies to the Application of Synthetic Antibodies Generated by
    Molecular Imprinting
  14. Hæmagglutinins of the Hæmophilus Group


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.