What i5 index should I use on the Illumina sample sheet for an unindexed p5 primer?

Question

I have an upcoming run on a HiSeq X and most of the libraries in the pool have both i5 and i7 indices. However, some of the libraries were made with the IS4 p5 oligo and it is unindexed.

The IS4 indexing oligo is from the following paper. The sequence is below.

Meyer, Matthias, and Martin Kircher. "Illumina sequencing library preparation for highly multiplexed target capture and sequencing." Cold Spring Harbor Protocols 2010.6 (2010): pdb-prot5448.

                                 miseq index
                              5' TCTTTCC 3'
                                 >>>>>>>
 5' AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTT 3'
    ^------------------^         ^--------------------^
    flowcell binding seq        adapter-binding seq

When we sequence these libraries dual-indexed on the MiSeq, the machine reads into the adapter and returns the index sequence TCTTTCC.

Question

What will the index sequence be when the HiSeq reads it?

The more fundamental question is: How exactly do MiSeq and HiSeq machines read index sequences?

side note

Here is what an index p5 and unindexed p5 molecule look like:

Update: Sept 2019 - After running into this problem again, an Illumina representative agreed with Devon Ryan and said that setting up a run for single indexing is the best option when there are unindexed p5 ends on one or more libraries.

Answer 1

Illumina machines read the two indices in separate index sequencing reactions where they have sequencing barcodes specific for indices. If you have no index, then you should either not bother with creating either index read or mask that during demultiplexing. If you're mixing indexed samples with an unindexed sample, you can't really rely on any particular sequence being returned by the machine (i.e., just use the "undetermined indices" fastq file). The unindexed adapters tend to not have the full index sequencing primer binding site on them, so reads without index generally don't return much signal. In practice, they're likely to adopt barcode sequences of neighboring clusters. We see that a lot when we spike in PhiX. It has no barcode but you tend to get a small amount assigned to each sample either due to index hopping or (more likely) bleed through of signal from neighboring clusters (or small amounts of those clusters walking between nanowells and then their minority signal dominating during index sequencing).