Hi Golam.

In the first instance and before hypothesising causes, I would look at the correlation function data from your measurements. If the sample is not optimally prepared for DLS, or the measurement is not suitably optimised, it is still possible to get a reported size but there are a number of artefacts that can skew the data and lead to uncertainty in the reported size.

I agree to what Dr. Alexander Malm suggests. Additionally, I would suggest performing the experiment under consistent conditions. That is, to make the measurement at a fixed number of minutes after ultra centrifugation of your sample. Do you filter your sample prior to making a DLS measurement ? Did you use a filter for your measurements where you measured smaller exosome size as compared to the first measurement? Are you sure that your exosome's assembly is stable?

If you can possibly access a regular light scattering instrument by ALV or Brookhaven, I'd strongly suggest performing the experiment on it. To obtain size in that case you'd have to make your measurement at various angles.

Dear Golam.

I agree with what colleagues went to... I think that the conditions for measurement must be fixed, special time and tune sizes

There could be several reasons causing the observed behavior [low scattering signal and number fluctuations due to low concentration, aggregating sample not stable in that condition, re-suspending aggregates, filtration,...] Maybe you can attach correlation function and scattering intensity details for specific advice.

It is very likely that the sample (or sample preparation) is the cause of the trending data. You may also find these links of interest

https://www.researchgate.net/post/Does_anyone_have_experience_with_DLS-measurements_of_exosomes_Zetasizer_NanoZS

http://www.materials-talks.com/blog/2014/11/17/tips-tricks-for-nanoparticle-characterization/#exosomes

Exosome are isolated from body fluids that contain a lot of proteins and/or protein aggregates. Rodinde Hendrickx (now at CalTech) worked on exosome isolation in my lab. We did find out that it was very difficult to isolate exosomes from fetal calf serum due to the presence of a strong protein background. Likewise it was difficult to isolate exosomes from cow's milk. Ultrafiltration prior to your experiment is a must IMHO we used a 0.1 um filter to get reproducible results. (As an aside these results were unfortunately never published). As your target is small and not too abundant DLS can be a real challenge!

Thank you very much all of you. I am following your advice and suggestion. I will let you know if I face any complication.

Dear Rabby, One of disadvantages of ultracentrifugation (use of filter membranes often leads to loss of exosome population) is that it brings large number of debris and impurities with exosome and their sampling at different times expressed differently. To avoid this hassle better to use polymer-based exosome isolation techniques.

Thank you sir for the valuable information.

Dear Alexander Malm Ulf Nobbmann Uma Nudurupati

Please look at my attached files. Still, same exosome sample giving different measurements.

I filtered the sample with .22 μm filter. Exosomes were suspended in 0.1M PBS buffer(pH 7.4).

Would you please provide me any technical suggestion?

The instrument reports say:

Distribution Form : |Standard|

Distribution Form(Dispersity) : Monodisperse

I assume these are settings that you specify for the calculation of the result. If so, you are telling the instrument to give you a monodisperse result (the correct term is monomodal). This is the solid line-of-fit to the raw correlation function. You can see it deviates at the longer time delays. This deviation varies from sample to sample. This suggests to me that your samples are not monomodal and you should use a different setting for Distribution Form. I assume the instrument can do that on existing data.

Also, the count rate seems low so try longer measurements to give the long time delay portion of the correlation function more time to become steady.

Hi Golam,

Thanks for adding the data.

I agree with John's comments but would also add that you appear to have some very large material present based on the very long decay, which may even be beyond the range of reported sizes for your instrument.

Your count rate is very low, typically most instruments would optimise for a count rate of order a few hundred kcps, but this optimisation may be thrown off if a few high scattering, large particles are present.

As well as checking the analysis settings, I would also check the measurement settings to check that the scattering intensity is optimised and not fixed, and also observe the count rate if possible. If this contains lots of spikes, then this indicates that large particles are intermittently transiting the beam.

If this is the case, filtering the sample to remove these large components will allow a more reliable characterisation of the smaller components.

I agree to the suggestions that John Francis Miller and Alexander Malm have. Avoid trying to force fit the data to a single exponent. Do you happen to have the raw data for the g2(T) -1 (the second graph in each of the documents) ? If you could fit the g2(T) -1 / (intercept of the g2(T)-1 on y axis) to say a biexponent or so on you'd might get a better idea about the population range of sizes that you have in you sample. Also,what is the concentration of the exosomes ? The low intensity could be a result of low concentration.

I have another tangential question, once assembled into large sizes do these assemblies settle down to the bottom of the tube? (That could be a possible explanation for the long time decay after filtration along with a low count rate). Hope this helps!

Edit: If you see carefully, the size you observe is 1/2 the size of the pore of the filter you use. Why don't you use a slightly bigger filter? 0.45 um PVDF (hydrophilic) filter. - This will ensure you are not filtering your sample out. (I would suggest doing only for one experiment). Additionally, do you filter your phosphate buffer separately before using it? I would suggest doing the same too - to eliminate dust.

Also, I am not sure if you have access to this data, but can you please display intensity vs time for any one of the above 6 experiments. The g2(T) function you have has very low intercept. Typically, the g2(T) - 1 value must be at least 0.3 or so (instrument dependent of course). But in your case this value is -0.5 (your g2(T) intercept is 0.5, 0.5 -1 = -0.5) . My hunch is there is very little or no scatters in there. I would first try getting DLS data with a g2(T) - 1 where the intercept on the y xis is > 0.3 . This is colloquially considered as an indication of signal to noise ratio. If this value is too low, then the correlation function is considered unsuitable for fitting. As a quick control experiment you could also run a test with just 0.1 M PBS after filtering it with 0.22 um filter. If the g2(T) for the PBS alone matches the signal you obtain for the exosome, its QED that you light scattering sample doesn't contain exosomes. (Whether thats a result of the filter or that of some other process, only you will know best)

Rabbee G. Mahmudunnabi Poor reproducibility often arises from broad distributions and (larger) material settling through the laser beam. Have you verified the performance of the instrument with the appropriate NIST traceable standard before even measuring something else? Are you working in Class 100 conditions? Are your continuous phases and solvents filtered to 0.02 microns or better? A general purpose model needs to be tried at the start rather than a (over-resolved?) narrow or monomodal model. However, it's easy to reanalyze your data - you don't need to take a new measurement. This is also a 90 degree system - can you get light in and out of the cuvette?